Tropical Medicine and
Infectious Disease

Article

Enhancing the Interpretability of Malaria and Typhoid Diagnosis
with Explainable AI and Large Language Models
Kingsley Attai 1, * , Moses Ekpenyong 2,3 , Constance Amannah 4 , Daniel Asuquo 5 , Peterben Ajuga 6 ,
Okure Obot 2 , Ekemini Johnson 1 , Anietie John 1 , Omosivie Maduka 7 , Christie Akwaowo 8
and Faith-Michael Uzoka 9
1

2

3

4

5

6

7
8
9

*

Citation: Attai, K.; Ekpenyong, M.;
Amannah, C.; Asuquo, D.; Ajuga, P.;
Obot, O.; Johnson, E.; John, A.;
Maduka, O.; Akwaowo, C.; et al.
Enhancing the Interpretability of
Malaria and Typhoid Diagnosis with
Explainable AI and Large Language
Models. Trop. Med. Infect. Dis. 2024, 9,
216. https://doi.org/10.3390/
tropicalmed9090216
Academic Editor: John Frean
Received: 7 August 2024
Revised: 13 September 2024
Accepted: 14 September 2024
Published: 16 September 2024

Copyright: © 2024 by the authors.
Licensee MDPI, Basel, Switzerland.

Department of Mathematics and Computer Science, Ritman University, Ikot Ekpene 530101, Nigeria;
eke5461@gmail.com (E.J.); aniettejohn5@gmail.com (A.J.)
Department of Computer Science, Faculty of Computing, University of Uyo, Uyo 520103, Nigeria;
mosesekpenyong@uniuyo.edu.ng (M.E.); okureobot@uniuyo.edu.ng (O.O.)
Science, Technology, Engineering and Mathematics (STEM) Centre, University of Uyo and Centre for
Research, University of Uyo, Uyo 520103, Nigeria
Department of Computer Science, Ignatius Ajuru University of Education, Port Harcourt 500102, Nigeria;
aftermymsc@gmail.com
Department of Information Systems, Faculty of Computing, University of Uyo, Uyo 520103, Nigeria;
danielasuquo@uniuyo.edu.ng
Department of Computer Engineering, Faculty of Engineering, Gregory University, Uturu 441106, Nigeria;
ajugapeterben@gmail.com
University of Port Harcourt Teaching Hospital, Port Harcourt 500102, Nigeria; omosivie.maduka@gmail.com
University of Uyo Teaching Hospital, Uyo 520103, Nigeria; christieakwaowo@uniuyo.edu.ng
Department of Mathematics and Computing, Mount Royal University, Calgary, AB T3E 6K6, Canada;
fuzoka@mtroyal.ca
Correspondence: attai.kingsley@ritmanuniversity.edu.ng; Tel.: +234-8101250218

Abstract: Malaria and Typhoid fever are prevalent diseases in tropical regions, and both are exacerbated by unclear protocols, drug resistance, and environmental factors. Prompt and accurate
diagnosis is crucial to improve accessibility and reduce mortality rates. Traditional diagnosis methods
cannot effectively capture the complexities of these diseases due to the presence of similar symptoms.
Although machine learning (ML) models offer accurate predictions, they operate as “black boxes”
with non-interpretable decision-making processes, making it challenging for healthcare providers to
comprehend how the conclusions are reached. This study employs explainable AI (XAI) models such
as Local Interpretable Model-agnostic Explanations (LIME), and Large Language Models (LLMs)
like GPT to clarify diagnostic results for healthcare workers, building trust and transparency in
medical diagnostics by describing which symptoms had the greatest impact on the model’s decisions
and providing clear, understandable explanations. The models were implemented on Google Colab
and Visual Studio Code because of their rich libraries and extensions. Results showed that the
Random Forest model outperformed the other tested models; in addition, important features were
identified with the LIME plots while ChatGPT 3.5 had a comparative advantage over other LLMs.
The study integrates RF, LIME, and GPT in building a mobile app to enhance the interpretability and
transparency in malaria and typhoid diagnosis system. Despite its promising results, the system’s
performance is constrained by the quality of the dataset. Additionally, while LIME and GPT improve
transparency, they may introduce complexities in real-time deployment due to computational demands and the need for internet service to maintain relevance and accuracy. The findings suggest
that AI-driven diagnostic systems can significantly enhance healthcare delivery in environments with
limited resources, and future works can explore the applicability of this framework to other medical
conditions and datasets.

This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://

Keywords: malaria diagnosis; typhoid diagnosis; machine learning; XAI; LIME; GPT; BERT; ChatGPT;
Gemini; perplexity; explainability; interpretability

creativecommons.org/licenses/by/
4.0/).

Trop. Med. Infect. Dis. 2024, 9, 216. https://doi.org/10.3390/tropicalmed9090216

https://www.mdpi.com/journal/tropicalmed

Trop. Med. Infect. Dis. 2024, 9, 216

2 of 23

1. Introduction
Typhoid fever and malaria are two of the most prevalent febrile diseases in the world,
presenting serious public health issues, especially in tropical and subtropical areas. Typhoid
and malaria are common in these areas due to the high humidity, temperatures, inadequate
healthcare facilities, and the shortage of qualified healthcare providers [1]. Despite these
diseases being caused by different pathogens and transmitted by diverse vectors, they share
several similarities as regards epidemiology, clinical manifestation, and co-infection. Their
prevalence is attributed to environmental and healthcare factors, including a warm and humid
climate, rapid urbanization without adequate infrastructure, which results in crowded living
conditions and poor sanitation, limited access to high-quality healthcare, a lack of preventive
measures, and weak disease surveillance systems in these regions. Typhoid fever and malaria
continue to be the leading causes of morbidity and mortality [2]. Salmonella enterica serotype
Typhi is the bacteria that causes typhoid fever or enteric fever, which affects millions of people
worldwide and can have serious consequences if left untreated [3–5]. Malaria, on the other
hand, is caused by plasmodium parasites that are transmitted by Anopheles mosquito bites,
infecting millions of people and claiming the lives of hundreds of thousands every year [6–8].
Malaria remains one of the world’s most serious health problems [9] and the second most
studied disease according to a systematic review [10]; this is due to its widespread prevalence,
high mortality rate, drug resistance, and environmental factors such as climate change in
tropical regions. The prompt and effective diagnosis of these febrile diseases is essential
for efficient treatment and care, but current diagnostic techniques often face limitations in
accessibility, specificity, and sensitivity. Blood smear examination (microscopy) and rapid
diagnostic tests (RDTs) are the current diagnostic techniques for malaria while the Widal test
and blood culture are the tests for typhoid fever. Since blood smear microscopy is low-cost,
effective, and capable of differentiating between malaria species and quantifying parasites, it is
the gold standard for diagnosing malaria. However, it does require a functional infrastructure
and skilled, qualified microscopists. RDTs identify malaria antigens in a small volume of
blood by using monoclonal antibodies that are directed against the target parasite antigen and
impregnated on a test strip but may be less sensitive to identify mixed or non-Plasmodium
falciparum infections [11]. The Widal test detects typhoid fever in patients’ serum using a
suspension of dead Salmonella enterica as an antigen. Still, it has low specificity and sensitivity,
which can result in incorrect diagnosis and treatment. In contrast, blood culture has high
specificity but can have compromised sensitivity due to low bacterial loads or previous
antibiotic use [12,13].
Machine learning (ML) algorithms are frequently used in the healthcare sector to help
decision-makers make well-informed decisions [14,15]. Medical diagnostics has found ML to
be a potent tool that can improve the efficiency and accuracy of diagnosis, but to guarantee that
medical professionals can rely on and comprehend the judgments made by these models, the
use of ML models in clinical settings demands a high level of interpretability and transparency.
Studies have applied numerous ML techniques in diagnosing malaria [16–18] and typhoid
fever [19], as well as both conditions together [20–22]. Even though ML models are frequently
used to diagnose diseases, the lack of integrated explainability in previous research makes it
difficult for medical professionals to have high confidence in the predictions. According to
Anderson and Thomas [23], concerns about ML algorithms’ lack of interpretability frequently
impede their acceptance in the healthcare sector. Since the healthcare sector is highly regulated,
there is a high demand for accountability and transparency in the decision-making processes
for ML models before their adoption [24]. Healthcare practitioners must be able to comprehend
and interpret the predictions made by ML models to be used safely as these models are used
to supplement clinical decision-making. Their capacity to comprehend and interpret the
choices made by ML models is critical in this sector, as decisions can have a significant impact
on patient outcomes. To address this challenge, an explainable AI (XAI) technique like Local
Interpretable Model-agnostic Explanations (LIME) offers insights into how models arrive at
their predictions, thereby promoting trust and aiding in clinical decision-making by healthcare
professionals. XAI is becoming increasingly important in the healthcare sector, where making

Trop. Med. Infect. Dis. 2024, 9, 216

3 of 23

decisions has extremely high stakes because it is challenging for healthcare professionals
to trust and comprehend the decisions made by traditional machine learning models. In
clinical settings, where comprehension of the reasoning behind a diagnosis is critical for
patient safety, regulatory compliance, and ethical considerations, the lack of interpretability
may impede the adoption of AI [25]. Therefore, XAI offers solutions to these problems by
facilitating AI models’ transparent and intelligible decision-making process. XAI techniques
such as LIME are widely utilized to clarify the inner workings of complex models. LIME
operates by using an interpretable model local to the prediction to approximate the blackbox model. It modifies the input data, tracks how the predictions change as a result, and
then fits a straightforward, understandable model to these modified samples [26,27]. In
situations where individual case explanations are required, LIME is especially helpful as it
helps determine which characteristics are most important for a particular prediction. The
interpretability of ML models in the healthcare industry is greatly enhanced by LIME, which
allows physicians to better comprehend and rely on AI-driven insights, and their capacity
to offer concise, useful explanations improves the usefulness of AI systems in the processes
of diagnosis and treatment planning. LIME has been applied in several healthcare settings
such as in diagnosing diabetes [28], classification of co-morbidities associated with febrile
diseases in children and pregnant women [29], and transparent health predictions [30]. To
further improve accuracy and explainability, incorporating large language models (LLMs)
into diagnostic processes seems promising in combination with XAI techniques. LLMs are
advanced AI systems built using deep learning techniques and trained on vast amounts of
data to accomplish a wide range of natural language processing (NLP) tasks. These models
can bridge the gap between complex ML algorithms and clinical understanding. They are
trained on a wealth of medical data and can provide distinctive interpretations and generate
detailed, contextually relevant explanations for diagnostic outcomes.
The use of LLMs in medical contexts has advanced significantly thanks to projects
like Generative Pre-trained Transformer (GPT) and Bidirectional Encoder Representations
from Transformers (BERT). These models can produce human-like text and comprehend
intricate linguistic patterns because they have been trained on enormous volumes of text
data. The applications of BERT go beyond identifying pandemic illnesses; it can also be
used to process electronic medical records and evaluate the results of goals-of-care talks
in clinical trials [30–33]. GPT has proven to be remarkably adept at producing coherent
and contextually relevant text in various domains [34]. GPT can help in the healthcare
industry by delivering comprehensive patient reports, producing justifications for medical
diagnoses, and offering assistance during clinical decision-making processes [35]. The
accuracy and explainability of diagnostic systems can be greatly improved by integrating
these LLMs and they can produce thorough narratives that clarify the reasoning behind
diagnostic predictions, which facilitates clinician comprehension and validation of AI
recommendations. This ability is essential for bridging the gap between cutting-edge AI
models and real-world, routine clinical use, raising the standard of healthcare delivery as
a whole.
Several other studies have integrated ML and XAI in diagnosis such as predicting the
risk of hypertension [36], preventing breast cancer [37], differentiating bipolar disorder [38],
predicting hepatitis C [39], and modeling comorbidity in patients with febrile diseases [29].
Other studies have proposed LLMs for healthcare purposes such as the prediction of
potential diseases [40], multimodal diagnosis [41], answering cardiology and vascular
pathologies questions [42], and answering questions on health diagnosis [43], but there
appears to be a gap in the literature regarding the combined use of all three methods (ML,
XAI, and LLMs) in diagnosing febrile diseases such as malaria and typhoid fever.
This study aims to enhance the interpretability of typhoid and malaria diagnosis using
ML techniques like Extreme Gradient Boost (XGBoost), Random Forest (RF), and Support
Vector Machine (SVM) with LIME, and LLMs such as GPT, Gemini, and Perplexity. RF
reduces the chance of overfitting and produces a robust result by combining multiple decision trees. The XGBoost algorithm is incredibly scalable and effective, capable of effectively

Trop. Med. Infect. Dis. 2024, 9, 216

4 of 23

managing both linear and nonlinear relationships while SVM can generalize well to new
data, making it a dependable tool for diagnosing diseases with similar symptoms. The XIA
tool gives healthcare workers concise explanations for every diagnosis, assisting them in
determining which symptoms had the greatest influence on the diagnosis. The LLMs further improve the output and increase the tool’s usability for non-specialists by converting
complex technical explanations into plain language. This study emphasizes the potential
of integrating these tools to interpret and contextualize medical data, hence bridging the
gap between healthcare worker comprehension and complex ML diagnoses. A dataset
consisting of patients’ symptoms and diagnoses of malaria and typhoid was collected from
healthcare facilities across the Niger Delta region of Nigeria. By leveraging these advanced
tools, we seek to develop a diagnostic model that delivers precise diagnoses and provides
transparent and understandable insights into their decision-making processes. This research holds significant potential to improve diagnostic practices, ultimately contributing
to better patient outcomes and advancing the field of medical diagnostics. This study can
advance the field of diagnostic medicine and enhance diagnostic procedures, which will
ultimately lead to better patient outcomes. This study’s primary contributions are:

•

•
•

•

•

•

The consideration of multiple diseases (typhoid fever and malaria) allows for a thorough evaluation of the patient’s health, which is essential for managing co-infection
and comorbidity.
Using real-world data ensures that the models are trained and validated on clinical
cases, thereby enhancing the practical relevance and applicability of our findings.
The black-box nature of many ML models is addressed by the integration of an
XAI method, which gives medical professionals transparent and comprehensible
insights into how each feature influences the diagnosis, ensuring that diagnostic
results are presented in a way that is meaningful for easier interpretation. By focusing
on interpretability, healthcare workers can make more accurate and timely diagnoses.
LLMs give the diagnosis process an extra layer of context-aware understanding and
incorporating them makes it possible to better understand and analyze complex
medical outcomes.
The combination of LLMs and conventional ML models enables a thorough comparison of various diagnosis strategies. This not only demonstrates the models’ efficacy but
also the advantages and disadvantages of each approach to managing medical data.
The integration of XAI, LLMs, and ML puts this work at the forefront of medical AI
research. It demonstrates the viability and benefits of using a hybrid approach to
address difficult diagnostic problems, establishing a standard for further study in
the area.

The study is prepared as follows: The methodology is presented in Section 2, including
data collection, preprocessing, and the application of XAI and ML models, along with the
incorporation of LLMs for improved diagnostic interpretability. The results are discussed in
Section 3, evaluating the effectiveness of various algorithms and illustrating how XAI offers
insights into model decisions, along with the implications for clinical practice. Section 4
concludes the study, highlighting its limitations, and offering recommendations for further
research to advance diagnostic techniques.
2. Methodology
2.1. Malaria and Typhoid Diagnosis Framework
The proposed diagnosis framework for malaria and typhoid fever is presented in
Figure 1. The major components of the framework include a healthcare worker, medical
experts, and a mobile device for the collection, processing, and storage of information
locally and on a cloud-based storage for decision making. Patient data were obtained from
medical experts and pre-processed into a format suitable for machine learning modeling
and processing by large language models. Pre-processing ensures data quality, selects and
encodes pertinent features, balances the dataset, and normalizes inputs, which contributes
to the model’s ability to make more dependable predictions. The proposed model can

Trop. Med. Infect. Dis. 2024, 9, 216

experts, and a mobile device for the collection, processing, and storage of information
locally and on a cloud-based storage for decision making. Patient data were obtained from
medical experts and pre-processed into a format suitable for machine learning modeling
and processing by large language models. Pre-processing ensures data quality, selects and
5 of 23
encodes pertinent features, balances the dataset, and normalizes inputs, which contributes
to the model’s ability to make more dependable predictions. The proposed model can be
utilized in the diagnoses of typhoid fever and malaria with enhanced accuracy and
be utilized in
diagnosesworker
of typhoid
fever
and malaria
enhanced
accuracy and
explainability
bythe
a healthcare
through
a mobile
device.with
Through
the user-friendly
explainability
by
a
healthcare
worker
through
a
mobile
device.
Through
the
interface, healthcare workers can input patient’s vitals and symptoms using user-friendly
dropdown
interface,
workers
can
input
patient’s
vitals can
andprocess
symptoms
using
menus
and healthcare
sliders. After
the data
are
entered,
the model
them
and dropdown
instantly
menus and
sliders.as
After
thehaving
data are
entered,
themalaria,
model can
process
them and instantly
diagnose
the patient
likely
typhoid
fever,
neither,
or both.
diagnose the patient as likely having typhoid fever, malaria, neither, or both.

Figure
1. 1.
Malaria
and
Typhoid
Fever
Diagnosis
Framework.
Figure
Malaria
and
Typhoid
Fever
Diagnosis
Framework.

2.2.
Description
Dataset
Used
Study
2.2.
Description
of of
thethe
Dataset
Used
forfor
thethe
Study
The
New
Frontiers
Research
Fund
project’s
dataset
instrument,designed
designedbyby
a team
The
New
Frontiers
inin
Research
Fund
project’s
dataset
instrument,
a team
medical
experts
in the
field
of febrile
diseases
and computer
scientists,
was in
used
ofof
medical
experts
in the
field
of febrile
diseases
and computer
scientists,
was used
thisin
this study.
The dataset,
comprising
patient
records,
organized
into
sections,
study.
The dataset,
comprising
4870 4870
patient
records,
waswas
organized
into
sixsix
sections,
includingdemographic
demographicdata,
data, patient
patient symptoms,
information
[44].
including
symptoms, risk
riskfactors,
factors,and
anddiagnosis
diagnosis
information
The
first
section
contains
the
patient
demographics
as
shown
in
Table
1
and
the
diagnosing
[44]. The first section contains the patient demographics as shown in Table 1 and the
physician’s
information.
The second
the patient’s
symptoms
on a fivediagnosing
physician’s
information.
Thesection
secondcontains
section contains
the patient’s
symptoms
point
scale
(1
=
absent;
2
=
mild;
3
=
moderate;
4
=
severe;
5
=
very
severe),
along
the
on a five-point scale (1 = absent; 2 = mild; 3 = moderate; 4 = severe; 5 = very severe),with
along
doctor’s
level
of
confidence
(a
numerical
rating
scale
from
1
to
10).
The
five-point
symptom
with the doctor’s level of confidence (a numerical rating scale from 1 to 10). The five-point
scale is based
reality,
where
symptoms
in severity,
and
ensures that
symptom
scale on
is clinical
based on
clinical
reality,
wherevary
symptoms
vary
initseverity,
anddata
it
collection
is
consistent
across
various
doctors
and
cases,
reducing
variability
and
potential
ensures that data collection is consistent across various doctors and cases, reducing
bias. The patient’s
degreebias.
of susceptibility
to the
otherofnon-clinical
risk to
factors
was listed
variability
and potential
The patient’s
degree
susceptibility
the other
non-in
the third section, and the doctor’s initial diagnosis was listed in the fourth. The confirmed
clinical risk factors was listed in the third section, and the doctor’s initial diagnosis was
diagnosis was included in the sixth section of the dataset after further investigations such
listed in the fourth. The confirmed diagnosis was included in the sixth section of the
as full blood count, blood film, and serology were conducted on the patient in the fifth
dataset after further investigations such as full blood count, blood film, and serology were
section. A linguistic scale (1 = absent; 2 = very low; 3 = low; 4 = moderate; 5 = high; 6 = very
conducted on the patient in the fifth section. A linguistic scale (1 = absent; 2 = very low; 3
high) was used to rate the intensity of attack for both preliminary and confirmed diagnoses
= low; 4 = moderate; 5 = high; 6 = very high) was used to rate the intensity of attack for
(Sections 4 and 6), along with the doctor’s degree of confidence (1–10) in each case. The
both preliminary and confirmed diagnoses (Sections 4 and 6), along with the doctor’s
dataset contained malaria, typhoid, HIV, respiratory tract infection, urinary tract infection,
degree of confidence (1–10) in each case. The dataset contained malaria, typhoid, HIV,
tuberculosis, lassa fever, yellow fever, and dengue fever with a total of 50 symptoms.
respiratory tract infection, urinary tract infection, tuberculosis, lassa fever, yellow fever,
and
dengue fever with a total of 50 symptoms.
Table 1. Statistics of male and female patients in the dataset.
Table
1. Statistics
of male and female
in the dataset.
Patient
Age (Years)
<5 patients
5–12
13–19
Male
Female
Total

534
419
953

346
323
669

150
213
363

20–64

≥65

Total

1012
1605
2617

133
135
268

2175
2695
4870

2.3. Data Preprocessing and Oversampling
The collected dataset comprised columns with both numeric and string data types,
along with a few missing values. Missing values are a common problem in datasets and
can cause bias, reduce model accuracy, and complicate data preprocessing, all of which can
negatively affect ML model performance.

Trop. Med. Infect. Dis. 2024, 9, 216

6 of 23

Data preprocessing was conducted, including feature selection, feature scaling, and
data cleaning. Records with missing features, irrelevant data, and columns that were not
needed were eliminated during the data-cleaning process. Records with missing symptoms
were likewise eliminated to maintain the integrity of the dataset. Because the patient
consultation tool did not include symptoms for patients under the age of five, records
of those patients were removed. This is because patients in this age group may not be
able to accurately express certain symptoms, leaving them to rely entirely on their parents’
interpretation. A selection of the most pertinent and significant features for modeling
febrile illness (malaria and typhoid fever symptoms) was made to carry out the feature
process. The dataset was reduced to 3914 records, including only the malaria7 and
Trop. Med. Infect. Dis. 2024, 9, x FOR selection
PEER REVIEW
of 30
typhoid fever confirmed diagnoses and their twelve (12) symptoms. These two diseases
with their 12 symptoms were selected from the list of symptoms because the rest of the
diseases were underrepresented in the dataset, leading to an imbalanced dataset. The
A patient’s symptoms, the intensity of each symptom, and confirmed diagnoses
scope was narrowed to these two diseases to enhance the model’s ability to detect and
(malaria
andbetween
typhoidthese
fever)two
are diseases
all included
ineffectively.
the processed dataset. The list of symptoms
differentiate
more
and A
diseases
with
abbreviations
is presented
in Table
2. As shown
in Figurediagnoses
2, custom
patient’s
symptoms,
the intensity
of each
symptom,
and confirmed
mapping
was
used
to
map
the
disease
severity
Absent’
(1)
to
binary
0
and
Very-low’
to
(malaria and typhoid fever) are all included in the processed dataset. The list of symptoms
Very-severe’
(2
to
6)
to
binary
1.
and diseases with abbreviations is presented in Table 2. As shown in Figure 2, custom
mapping was used to map the disease severity ‘Absent’ (1) to binary 0 and ‘Very-low’ to
Table 2. Symptoms and diseases with abbreviations.
‘Very-severe’ (2 to 6) to binary 1.
Symptom/Disease
Abbreviation
Table 2. Symptoms and diseases with abbreviations.
Abdominal pains
ABDPN
Bitter taste in mouth
BITAIM
Symptom/Disease
Abbreviation
Chills and rigors
CHLNRIG
Abdominal pains
ABDPN
Bitter tasteConstipation
in mouth
BITAIMCNST
Chills and Fatigue
rigors
CHLNRIG
FTG
Constipation
CNST FVR
Fever
Fatigue
FTG
Generalized body pain
GENBDYPN
Fever
FVR
Headaches
HDACH
Generalized body pain
GENBDYPN
High-grade
fever
HGGDFVR
Headaches
HDACH
High-grade
fever
HGGDFVR
Lethargy
LTG
Lethargy
LTG
Muscle
and body pain
MSCBDYPN
Muscle and body pain
MSCBDYPN
Stepwise
rise fever
SWRFVR
Stepwise
rise fever
SWRFVR
Malaria
Malaria
MAL MAL
Typhoid
fever/Enteric
fever fever
ENFVR
Typhoid
fever/Enteric
ENFVR

Figure2.2.Pre-processed
Pre-processeddataset.
dataset.
Figure

After
Afterfurther
furtheranalysis,
analysis,we
wenoticed
noticedthat
thatof
ofthe
the3914
3914patients,
patients,1088
1088patients
patientshad
hadneither
neither
malaria
1669
had
only
malaria,
107107
hadhad
onlyonly
typhoid
fever,fever,
and 1050
had
malarianor
nortyphoid
typhoidfever,
fever,
1669
had
only
malaria,
typhoid
and 1050
had both diseases. The Synthetic Minority Oversampling Technique (SMOTE) was
employed to handle the class imbalance. SMOTE has several advantages and when
compared to just replicating minority class instances, it lowers the chance of overfitting
by creating synthetic samples. It improves model performance, is compatible with most
ML techniques, and is useful for various types of data. SMOTE identified minority class

Trop. Med. Infect. Dis. 2024, 9, 216

7 of 23

both diseases. The Synthetic Minority Oversampling Technique (SMOTE) was employed to
handle the class imbalance. SMOTE has several advantages and when compared to just
replicating minority class instances, it lowers the chance of overfitting by creating synthetic
samples. It improves model performance, is compatible with most ML techniques, and
is useful for various types of data. SMOTE identified minority class instances, selected
k-nearest neighbors, and generated and added synthetic samples to the original dataset,
Trop. Med. Infect. Dis. 2024, 9, x FOR as
PEER
REVIEW in Figure 3. The oversampled dataset contains 6676 patient records with
8 of
30
presented
the
class labels 0 (No disease), 1 (Typhoid only), 2 (Malaria only), and 3 (Both diseases) in the
‘Disease’ column.

Figure3.3.Oversampled
Oversampleddataset
datasetwith
withSMOTE.
SMOTE.
Figure

2.4.
2.4.Diagnostic
DiagnosticModels
Modelsand
andModel
ModelOptimization
Optimization
We
Weused
usedGoogle
GoogleColaboratory
Colaboratory(Colab),
(Colab),aafree
freecloud-based
cloud-basedplatform
platformfrom
fromGoogle
Googlethat
that
offers
programming
environment
with quick
robust graphics
offersa Python
a Python
programming
environment
with access
quicktoaccess
to robustprocessing
graphics
unit
(GPU) resources
and
ML libraries.
Additionally,
the platform provides
a CPU
runtimea
processing
unit (GPU)
resources
and ML
libraries. Additionally,
the platform
provides
and
easily
integrates
for storage.
Python
and libraries
as NumPy,
CPU
runtime
and Google
easily Drive
integrates
Google
Drivepackages
for storage.
Pythonsuch
packages
and
Pandas,
and Matplotlib,
which are necessary
for creating
classification
models,
librariesScikit-Learn,
such as NumPy,
Pandas, Scikit-Learn,
and Matplotlib,
which
are necessary
for
were
utilized.
The
ML
algorithms
used
in
building
our
diagnostic
models
and
the
performance
creating classification models, were utilized. The ML algorithms used in building our
metrics
are presented
in the
incorporating
tuning,
known as
diagnostic
models and
thesubsection
performance
metrics hyperparameter
are presented in
the subsection
grid
search
cross-validation
(GridSearchCV),
which
is
used
to
increase
the
precision
of the
incorporating hyperparameter tuning, known as grid search cross-validation
diagnosis.
GridSearchCV
expanded
method
optimizing
hyperparameters
by enabling
(GridSearchCV),
which is
isan
used
to increase
thefor
precision
of the
diagnosis. GridSearchCV
customized
search
spacesfor
foroptimizing
each hyperparameter,
using designated
The hyperis an expanded
method
hyperparameters
by enablingranges.
customized
search
parameter
setting
used
was:
SVM
(‘C’:
[0.1,
1,
10,
100],
‘gamma’:
[‘scale’,
‘auto’,
0.001,
0.01,
spaces for each hyperparameter, using designated ranges. The hyper-parameter setting
0.1],
‘kernel’:
[‘rbf’,
‘linear’]).
‘C’
is
the
parameter
that
controls
the
trade-off
between
achieving
used was: SVM ( C’: [0.1, 1, 10, 100], gamma’: [ scale’, auto’, 0.001, 0.01, 0.1], kernel’:
a[ low
on theC’
training
and minimizing
the model
complexity,
Gamma
definesahow
rbf’,error
linear’]).
is the data
parameter
that controls
the trade-off
between
achieving
low
far
the
influence
of
a
single
training
example
reaches,
while
the
Kernel
function
transforms
error on the training data and minimizing the model complexity, Gamma defines howthe
far
data
into a higher-dimensional
spaceexample
to make reaches,
them easier
to separate
using
a lineartransforms
boundary.
the influence
of a single training
while
the Kernel
function
XGBoost
[3, 4, 5, 6], ‘learning_rate’:
0.2],to
‘n_estimators’:
[100,
200,
the data (‘max_depth’:
into a higher-dimensional
space to make[0.01,
them0.1,
easier
separate using
a linear
300],
‘colsample_bytree’:
[0.3, 0.7]), where
determines
the maximum
depth of
boundary.
XGBoost ( max_depth’:
[3–6], max_depth
learning_rate’:
[0.01, 0.1,
0.2], n_estimators’:
the trees, learning_rate controls how much the model’s weights are adjusted to the loss
[100, 200, 300], colsample_bytree’: [0.3, 0.7]), where max_depth determines the maximum
gradient, n_estimators indicate the number of trees to be built, and colsample_bytree defines
depth of the trees, learning_rate controls how much the model’s weights are adjusted to
the subsample ratio of columns when constructing each tree. RF(‘n_estimators’: [100, 200,
the loss gradient, n_estimators indicate the number of trees to be built, and
300], ‘max_depth’: [None, 10, 20, 30], ‘min_samples_split’: [2, 5, 10], ‘min_samples_leaf’: [1, 2,
colsample_bytree defines the subsample ratio of columns when constructing each tree.
4], ‘bootstrap’: [True, False]), where min_samples_split determines the minimum number of
RF( n_estimators’: [100,200,300], max_depth’: [None, 10, 20, 30], min_samples_split’: [2,
samples required to split an internal node, min_samples_leaf specifies the minimum number
5, 10], min_samples_leaf’: [1,2,4], bootstrap’: [True, False]), where min_samples_split
of samples required to be at a leaf node, and bootstrap determines whether bootstrap samples
determines the minimum number of samples required to split an internal node,
min_samples_leaf specifies the minimum number of samples required to be at a leaf node,
and bootstrap determines whether bootstrap samples are used when building trees. Each
of these hyperparameters aids in fine-tuning the behavior of the model, enhancing its
functionality and ability to diagnose the febrile illnesses considered in this study with
good generalization. These hyperparameters were derived from built-in functions of the

Trop. Med. Infect. Dis. 2024, 9, 216

8 of 23

are used when building trees. Each of these hyperparameters aids in fine-tuning the behavior
of the model, enhancing its functionality and ability to diagnose the febrile illnesses considered
in this study with good generalization. These hyperparameters were derived from built-in
functions of the corresponding ML algorithms. Our local laptops utilized for this study were
a Dell Latitude 7480, Core i5-7200U CPU @ 2.50 GHz (4 CPUs), ~2.7 GHz with 16 GB RAM
for the ML and XAI modeling while a Samsung 950QDB, Core i7-1165G7 @ 2.80Ghz (8 CPUs)
~ 2.8 GHz with 16 GB RAM was used for the large language modeling. We used Visual
Studio Code, a free coding editor that supports several extensions and allows for quick coding
initiation. LLMs are easily accessible thanks to the Python foundation of our development
environment. The process was automated by utilizing core Python packages and libraries,
such as pandas, numpy, flet, matplotlib, flask, flask-sqlalchemy, seaborn, sk-learn, and joblib
for loading models. The information extractor comprises a prompt generator, automator, and
interpreter. The Malaria and Typhoid Diagnosis System interacts with various application
programming interfaces (APIs) for database communication and diagnosis management. It
has two main components: the front-end, built using Flet with Matplotlib and Seaborn for
visualizations, and the back-end, powered by Flask for API integration and MySQL database
management via Flask-SQLAlchemy. The prompt generator converts data into a readable
format, saves prompts in a JSON file, and organizes the patient’s symptoms and severity into
manageable prompts. The prompt used by Caruccio et al. [45] mimics the conversation of a
physician when seeking assistance in diagnosing a patient based on particular symptoms. The
template is “The patient has these symptoms: [S] Tell me which of the following diagnoses
is most related to the symptoms: [D]? [H]”. Where [S] is all of the symptoms listed in the
prompt, [D] the diagnoses the LLM must decide on, and [H] the answer or diagnoses provided
by the LLM. This template was modified to arrive at our prompt: “The patient has these
symptoms with severity levels, listed in the table below. (create a table with only the diagnosis
column filled in), the output should be in CSV format, diagnosis [Malaria, Typhoid Fever, Both,
None]?”. The automator manages data flow by retrieving outputs and storing them in a JSON
feeds these prompts into the large language models (GPT, Gemini, and Perplexity).
Trop. Med. Infect. Dis. 2024, 9, x FOR file.
PEERIt
REVIEW
10 of 30
After that, the interpreter transforms the JSON output into an Excel file so that reporting and
analysis of the data can be carried out. The link to the scripts can be found in this GitHub
account https://github.com/FebrileDiseasesDiagnoses/Auto_tool.git (2 August 2024).
2.4.1. Random Forest
2.4.1. Random
Random Forest algorithm is an ensemble ML technique with robust resistance to
over-fitting
combines
several
to increase
prediction
accuracy to
[46],
as
Randomthat
Forest
algorithm
is an decision
ensembletrees
ML technique
with
robust resistance
overshownthat
in Figure
4. RFseveral
trains predictions
concurrently,
well
on large
datasets,
and
fitting
combines
decision trees
to increase operates
prediction
accuracy
[46],
as shown
is
good
at
estimating
missing
data
[47].
RF
can
easily
resolve
high-dimensional
and
in Figure 4. RF trains predictions concurrently, operates well on large datasets, and is
complex
problemsmissing
such asdata
the [47].
prediction
disease
conditions
[48–50]. By
combining
good
at estimating
RF can of
easily
resolve
high-dimensional
and
complex
individual
tree
predictions
via
voting,
the
final
prediction
is
produced.
This
approach
problems such as the prediction of disease conditions [48–50]. By combining individual
increases
the model’s
robustness,
decreases
overfitting,
and can
in diagnosing
febrile
tree
predictions
via voting,
the final
prediction
is produced.
Thisaid
approach
increases
the
diseases.robustness, decreases overfitting, and can aid in diagnosing febrile diseases.
model’s

Figure4.4.Random
RandomForest
Forestschematic
schematicdiagram.
diagram.
Figure

2.4.2. Extreme Gradient Boost
XGBoost algorithm is a component of the gradient boosting framework, which can
be applied to regression or classification predictive modeling issues. Figure 5 depicts the
computation process used by XGBoost as it introduces weak learners into the ensemble,
focusing each new learner on correcting the mistakes made by the previous ones. Because

Trop. Med. Infect. Dis. 2024, 9, 216

9 of 23

Figure 4. Random Forest schematic diagram.

2.4.2. Extreme
ExtremeGradient
GradientBoost
Boost
2.4.2.
XGBoost algorithm
algorithm is
is aacomponent
component of
ofthe
thegradient
gradientboosting
boostingframework,
framework, which
whichcan
can
XGBoost
be
applied
to
regression
or
classification
predictive
modeling
issues.
Figure
5
depicts
the
be applied to regression or classification predictive modeling issues. Figure 5 depicts the
computation
process
used
by
XGBoost
as
it
introduces
weak
learners
into
the
ensemble,
computation process used by XGBoost as it introduces weak learners into the ensemble,
focusingeach
eachnew
newlearner
learneron
oncorrecting
correctingthe
themistakes
mistakesmade
madeby
bythe
theprevious
previousones.
ones.Because
Because
focusing
of
its
reputation
for
managing
structured
data,
XGBoost
is
extensively
utilized
in
of its reputation for managing structured data, XGBoost is extensively utilized in numerous
numerous applications,
including
theof
prediction
of disease [51].
applications,
including the
prediction
disease [51].

Figure5.5.Extreme
Extremegradient
gradientboosting
boostingschematic
schematicdiagram.
diagram.
Figure

2.4.3.
2.4.3. Support
Support Vector
VectorMachine
Machine
SVM
SVM isiswell-known
well-known for
forworking
working well
wellin
inhigh-dimensional
high-dimensional spaces
spaces and
andfor
forhandling
handling
non-linearly
non-linearly separable
separable data
data by
by utilizing
utilizing kernel
kernel functions.
functions. It
It seeks
seeks to
to determine
determine which
which
hyperplane
best
divides
the
data
into
distinct
classes.
The
margin
is
the
distance
between
hyperplane best divides the data into distinct classes. The margin is the distance between
the
and
that
thehyperplane
hyperplane
and the
theclosest
closestobservations,
observations, and
and the
the support
support vectors
vectors are
are the
thepoints
points
that
Trop. Med. Infect. Dis. 2024, 9, x FOR PEER
REVIEW
11 of
30
are
closest
to
it,
as
shown
in
Figure
6.
SVM
uses
little
memory,
performs
well
with
a
wide
are closest to it, as shown in Figure 6. SVM uses little memory, performs well with a wide
range
rangeof
offeatures,
features,and
andcan
canbe
betailored
tailoredwith
withvarious
variouskernel
kernelfunctions
functionsfor
forintricate
intricatedecision
decision
boundaries.
SVM
is
resistant
to
overfitting
and
can
handle
high-dimensional
data
as
boundaries.
SVM
is
resistant
to
overfitting
and
can
handle
high-dimensional
data
aswell
well
as
an effective
as binary
binary and
and multi-class
multi-class classification
classification issues
issues in
in medical
medical diagnosis,
diagnosis, making
making it
it an
effective
tool
diseases [52].
[52].
tool for
for diagnosing
diagnosing diseases

Figure 6. Support Vector Machine diagram.

2.5. Interpretability
Interpretability and
and Explainability
Explainability Methods
Methods
2.5.
Local Interpretable
Interpretable Model-agnostic
Model-agnostic Explanations
Explanations (LIME)
(LIME) approximates
approximates the
the complex
complex
Local
model near
near aa particular
particular prediction
prediction with
with an
an interpretable
interpretable model
model such
such as
as aa linear
model to
to
model
linear model
provide local explanations. The integration of LIME into our model follows these key steps:
provide local explanations. The integration of LIME into our model follows these key
(i) Instance Selection: LIME was applied to each instance in the test dataset, generating
steps: (i) Instance Selection: LIME was applied to each instance in the test dataset,
localized explanations for the model’s predictions on a case-by-case basis; (ii) Feature
generating localized explanations for the model’s predictions on a case-by-case basis; (ii)
Contribution Analysis: LIME produces visualizations that indicate the contribution of each
Feature Contribution Analysis: LIME produces visualizations that indicate the
contribution of each feature to the prediction. Features that positively influence the
likelihood of a specific disease are displayed on the right side of the plot, while those that
decrease the likelihood are shown on the left; (iii) Global Insight Aggregation: By
aggregating LIME explanations across multiple instances, we can identify patterns and
key features that consistently influence the model’s decisions, providing a broader

Trop. Med. Infect. Dis. 2024, 9, 216

10 of 23

feature to the prediction. Features that positively influence the likelihood of a specific
disease are displayed on the right side of the plot, while those that decrease the likelihood
are shown on the left; (iii) Global Insight Aggregation: By aggregating LIME explanations
across multiple instances, we can identify patterns and key features that consistently
influence the model’s decisions, providing a broader understanding of the model’s behavior
across the dataset.
Generative Pre-trained Transformer (GPT) is pre-trained using unsupervised learning
on a large corpus of text data, where it learns to predict the word that will appear next in
a sequence based on every word that has come before it. This pre-training gives GPT a
thorough grasp of language syntax, semantics, and context. When GPT is fine-tuned on
particular tasks, like text generation, question answering, or text completion, it uses its
learned representations to produce outputs that make sense and are relevant to the context.
GPT is an effective tool for NLP applications because it can produce text similar to that of a
human being and manage a wide range of linguistic tasks.
Bidirectional Encoder Representations from Transformers (BERT) is a transformerbased model that is trained to predict missing words in both directions with the help of
masking certain words in the input and making predictions about them using both left and
right context. Thanks to this bidirectional training, it can capture more complex contextual
meanings and relationships within text, producing more accurate language representations.
BERT can comprehend subtleties in language and performs well on a variety of natural
language understanding tasks, including named entity recognition, sentiment analysis,
and machine translation, thanks to its large-scale pre-training tasks. BERT is a flexible and
potent model for a range of NLP applications. Its performance can be further improved by
fine-tuning it for particular tasks.
2.6. Model Performance Metrics
The dataset used for this study initially contained 4870 patient records with symptoms
of febrile diseases. After preprocessing, the records were reduced to 3914, and after
oversampling, 6676 patient records with relevant features were retained for ML modeling.
In total, 80% of the dataset was used for training and 20% for testing. GridSearchCV
was employed to optimize model performance and StratifiedKFold was used for crossvalidation, dividing the dataset into five stratified folds and shuffling the data before
splitting to ensure robust and unbiased results. The experimental models were evaluated
using key performance metric components. True Positives (TP) are cases where the model
correctly predicts the positive class, represented by the diagonal elements of the matrix,
while True Negatives (TN) are cases where the model correctly predicts the negative class.
TN is the sum of all the cells that are neither in the row nor the column corresponding
to the class being considered. False Positives (FP) are cases where the model incorrectly
predicts the positive class while False Negatives (FN) are cases where the model incorrectly
predicts the negative class. When evaluating the sensitivity and specificity of diagnostic
tests, these metric components are helpful. The evaluation metrics listed below were used
in this paper.
Accuracy is a measurement of how well a model predicts all labels linked to each data
point in a dataset. Datasets with a balanced distribution of positive and negative samples
are good candidates for accuracy. For unbalanced datasets, it is less helpful because they
can be deceptive.
Accuracy =

True Positives + True Negatives
True Positives + True Negatives + False Positives + False Negatives

(1)

Precision is a metric that expresses how accurately a model predicts positive outcomes;
it measures the model’s capacity to correctly identify true positive instances while avoiding
false positives. When false positives come at a high cost, accuracy matters. In the context of
medical diagnosis, for instance, a false positive may result in needless treatments.

Trop. Med. Infect. Dis. 2024, 9, 216

11 of 23

Precision =

True Positives
True Positives + False Positives

(2)

Recall is a metric used to assess a model’s capacity to locate every positive instance
in a dataset. It measures how sensitive the model is to True Positive cases. When the cost
of false negatives is significant, as in medical screenings, recall is crucial because it can be
crucial to miss a positive case (false negative).
Recall =

True Positives
True Positives + False Negatives

(3)

F1-Score is a metric that represents the harmonic mean of Rrecall and Pprecision. The
F1-score is limited to a range of binary values, where 1 denotes that every class’s data point
was correctly predicted and 0 denotes that any class’s data point was incorrectly predicted.
When you must strike a balance between Rrecall and Pprecision, the F1 Score can be helpful,
particularly when your class distribution is not uniform.


Precision ∗ Recall
F1 = 2
(4)of 30
Trop. Med. Infect. Dis. 2024, 9, x FOR PEER REVIEW
18
Precision + Recall
Log Loss is a measure of the probability of a prediction’s accuracy and it penalizes
the difference between the expected probabilities and the actual class labels. Log loss is
instance being in the positive class, and <!-- MathType@Translator@5@5@MathML2 (no
helpful when one needs a metric that can handle probabilistic model outputs and penalizes
namespace).tdl@MathML 2.0 (no namespace)@ -->
incorrect predictions more severely.

<math><mrow><mrow><mrow><mi
mathvariant=ʺnormalʺ>log</mi></mrow><mo>⁡</mo><mrow><mfenced
1 N
(5)
Logloss = − ∑i=1 [yi log( pi ) + (1 − yi )log(1 − pi )]
separators=ʺ&#x007C;ʺ><mrow><msub><mrow><mi>p</mi></mrow><mrow><mi>i</mi>
N
</mrow></msub></mrow></mfenced></mrow></mrow></mrow></math>
where N is the total number of samples in the dataset, yi is the actual label of the i − th
<!-MathType@End@5@5@
-->
instance,
pi is the predicted probability
of the i − th instance being in the positive class, and
is
the
natural
logarithm
of
theofpredicted
probability
for the
positive
class
log( p ) is the natural logarithm
the predicted
probability
for the
positive
class
i

3.
and Discussion
Discussion
3. Results
Results and
performance
areare
shown
in this
section,
The results
results of
ofour
ourassessment
assessmentofofthe
themodels’
models’
performance
shown
in this
section,
including the XAI method
including
method adopted
adoptedas
aswell
wellas
asthe
theexperimental
experimentalassessment
assessmentofofthe
theLLMs
LLMs of
of Malaria
and
TyphoidFever
Fever diagnoses.
diagnoses. Figures
thethe
confusion
matrices,
an an
Malaria
and
Typhoid
Figures7–9
7–9present
present
confusion
matrices,
essential
instrument
for
assessing
how
well
a
classification
ML
model
performs.
essential instrument for assessing how well a classification ML model performs.

Figure 7.
Figure
7. XGBoost
XGBoostAlgorithm
AlgorithmConfusion
ConfusionMatrix.
Matrix.

Trop. Med. Infect. Dis. 2024, 9, 216

12 of 23

Figure 7. XGBoost Algorithm Confusion Matrix.

Trop. Med. Infect. Dis. 2024, 9, x FOR PEER REVIEW

19 of 30

Figure 8.
8. RF
RF Algorithm
Algorithm Confusion
Confusion Matrix.
Matrix.
Figure

Figure 9.
9. SVM
SVMAlgorithm
AlgorithmConfusion
ConfusionMatrix.
Matrix.
Figure

Table 33 presents
presents values
values of
of these
these metrics
metrics and
and the
the computation
computation time
time of
of each
each model
model
Table
while
Figure
10
is
a
pictorial
representation
of
the
model’s
performance
based
on
the
while Figure 10 is a pictorial representation of the model’s performance based on the
considered metrics. The result shows that RF (accuracy = 71.99%, precision = 71.29%,
considered metrics. The result shows that RF (accuracy = 71.99%, precision = 71.29%, recall
recall = 71.99%, F1-Score = 71.45%) demonstrates superior performance, outperforming
= 71.99%, F1-Score = 71.45%) demonstrates superior performance, outperforming XGBoost
XGBoost (accuracy = 71.29%, precision = 70.56%, recall = 71.29%, F1-Score = 70.66%) and
(accuracy = 71.29%, precision = 70.56%, recall = 71.29%, F1-Score = 70.66%) and SVM
SVM (accuracy = 68.60%, precision = 68.65%, recall = 68.60%, F1-Score = 68.21%). High
(accuracy = 68.60%, precision = 68.65%, recall = 68.60%, F1-Score = 68.21%). High recall
recall and precision are essential for diagnosing diseases like typhoid and malaria by
and precision are essential for diagnosing diseases like typhoid and malaria by
guaranteeing that the majority of real cases are identified. In this case, high precision helps
guaranteeing that the majority of real cases are identified. In this case, high precision helps
prevent needless treatments for illnesses that are not present. Because both XGBoost and
prevent needless treatments for illnesses that are not present. Because both XGBoost and
RF do a good job of balancing these metrics, they are better suited for clinical applications
RF do a good job of balancing these metrics, they are better suited for clinical applications
where false positives and false negatives can have detrimental effects. Also, XGBoost has a
where false positives and false negatives can have detrimental effects. Also, XGBoost has
smaller log loss, which suggests more accurate and well-calibrated probability estimates as
a smaller log loss, which suggests more accurate and well-calibrated probability estimates
well as stronger diagnosis confidence. This may be critical in medical diagnostics, where
as
well asis stronger
diagnosis
confidence.inThis
may be of
critical
in medical
diagnostics,
accuracy
not as important
as confidence
the presence
a disease.
In medical
scenarios
where
not as important
as confidence
the presence
of a disease.
In medical
where accuracy
treatmentisdecisions
are influenced
by the in
certainty
of a diagnosis,
lower
log loss
scenarios
treatment
are influenced
by the
of a diagnosis,
lower
values forwhere
XGBoost
indicatedecisions
that its probability
estimates
arecertainty
more reliable.
Because of
RF’s
log
loss
values
for
XGBoost
indicate
that
its
probability
estimates
are
more
reliable.
higher log loss, probability estimates are less trustworthy, which could cause uncertainty
Because
of RF’sdecisions.
higher logSVM
loss, performs
probability
estimates
trustworthy,
could
when making
worse
than are
the less
other
two modelswhich
in terms
of
cause
uncertainty
when
making
decisions.
SVM
performs
worse
than
the
other
two
performance metrics and computation time (running time exceeds one hour), implying
models
in terms
performance
metrics
and computation
(running
time
exceeds
one
that it might
notof
work
as well for
diagnosing
typhoid andtime
malaria
in this
specific
dataset.
hour),
implying
that
it
might
not
work
as
well
for
diagnosing
typhoid
and
malaria
in
this
Therefore, ensemble techniques (XGBoost and Random Forest) may be better at capturing
specific
dataset.
Therefore, between
ensemblesymptoms
techniquesand
(XGBoost
andthan
Random
Forest)
may RF
be
the intricate
relationships
diseases
the SVM
model.
better at capturing the intricate relationships between symptoms and diseases than the
SVM model. RF combines the predictions of multiple decision trees to make a final
prediction, which results in a slightly higher accuracy but at the cost of increased
computational complexity, while XGBoost optimizes each tree by minimizing errors from
the previous ones, leading to faster convergence and efficient model optimization. The

Trop. Med. Infect. Dis. 2024, 9, 216

13 of 23

combines the predictions of multiple decision trees to make a final prediction, which results
in a slightly higher accuracy but at the cost of increased computational complexity, while
XGBoost optimizes each tree by minimizing errors from the previous ones, leading to faster
convergence and efficient model optimization. The moderate F1-scores in these models are
a result of typhoid fever and malaria having very similar symptoms, making it challenging
for the models to differentiate between the two illnesses. This overlap can impair the
model’s predictive accuracy, especially concerning recall and precision.
Table 3. Diagnostics model performance.
Algorithm

Accuracy

XGBoost
0.7129
RF
Trop. Med. Infect. Dis. 2024, 9, x FOR PEER REVIEW 0.7199
SVM
0.6860

Precision

Recall

F1-Score

Log Loss

Computation Time

0.7056
0.7129
0.6865

0.7129
0.7199
0.6860

0.7066
0.7145
0.6821

0.7808
1.0548
0.8016

2 min, 32 s
14 min, 9 s
20 of 30
1 h, 22 min, 7 s

1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Accuracy

Precision

XGBoost

Recall

RF

F1-score

Log Loss

SVM

Figure10.
10.Performance
PerformanceEvaluation
Evaluationof
ofthe
theMachine
MachineLearning
LearningModels.
Models.
Figure

TheLIME
LIME
plots
(Figures
11–13)
provide
global
viewtheoffeatures
how the
features
The
plots
(Figures
11–13)
provide
a globala view
of how
(symptoms)
(symptoms)
contribute
the model’s
diagnoses
across
entireidentifying
test dataset,
identifying
contribute
to the
model’stodiagnoses
across
the entire
test the
dataset,
features
with
features
with
the highest
average contributions,
both
positively
and across
negatively,
across all
the
highest
average
contributions,
both positively
and
negatively,
all diagnoses.
diagnoses.
The
XGBoost
LIME
Figuresymptoms
11 shows symptoms
such as HDACH,
SWRFVR,
The
XGBoost
LIME
diagram
in diagram
Figure 11inshows
such as SWRFVR,
HDACH,
CNST, asby
specified
by theircontributions
negative contributions
the
the
and
CNST,and
as specified
their negative
on the left on
side
ofleft
theside
plot,ofsuggesting
that the lower
levels
or levels
absence
of these of
symptoms
are associated
with a lower
plot, suggesting
that the
lower
or absence
these symptoms
are associated
with a
likelihood
of a patient
typhoid.
Meanwhile,
symptoms
such as BITAIM,
lower likelihood
of a having
patient malaria
having and
malaria
and typhoid.
Meanwhile,
symptoms
such as
LTG,
CHLNRIG,
MSCBDYPN,
and FVR are
most
BITAIM,
LTG, CHLNRIG,
MSCBDYPN,
andthe
FVR
areinfluential
the most symptoms
influential constantly
symptoms
contributing
to the diagnoses
malaria and
typhoid
across
numerous
patients. patients.
constantly contributing
to theof
diagnoses
of malaria
and
typhoid
across numerous
The RF LIME diagram in Figure 12 also points out that the same symptoms (SWRFVR,
HDACH, and CNST) are associated with a lower likelihood of having malaria and typhoid,
whereas BITAIM, CHLNRIG, ABDPN, LTG, GENBDYPN, MSCBDYPN, FTG, and HGGDFVR are influential symptoms that contribute to the diagnoses of malaria and typhoid
among patients.
Figure 13 shows the SVM LIME diagram, indicating that CHLNRIG has the highest
feature importance, followed by MSCBDYPN, LTG, ABDPN, BITAIM, FTG, and CNST as
the influential symptoms that contribute to the diagnoses of malaria and typhoid among
patients. Meanwhile, GENBDYPN, SWRFVR, FVR, HGGDFVR, and HDACH are associated
with a lower likelihood of having malaria and typhoid.

diagnoses. The XGBoost LIME diagram in Figure 11 shows symptoms such as SWRFVR,
HDACH, and CNST, as specified by their negative contributions on the left side of the
plot, suggesting that the lower levels or absence of these symptoms are associated with a
lower likelihood of a patient having malaria and typhoid. Meanwhile, symptoms such as
Trop. Med.
Med. Infect.
Infect. Dis.
9, 216
x FOR PEER REVIEW
21 of 30
Trop.
Dis. 2024,
2024, 9,
14 of 23
BITAIM, LTG, CHLNRIG, MSCBDYPN, and FVR are the most influential symptoms
constantly contributing to the diagnoses of malaria and typhoid across numerous patients.

Trop. Med. Infect. Dis. 2024, 9, x FOR PEER REVIEW

Figure 12. RF Algorithm LIME diagram.
Figure 11. XGBoost Algorithm LIME diagram.

21 of 30

Figure 13 shows the SVM LIME diagram, indicating that CHLNRIG has the highest
Theimportance,
RF LIME followed
diagram by
inMSCBDYPN,
Figure 12 also
outBITAIM,
that theFTG,
same
symptoms
feature
LTG,points
ABDPN,
and
CNST as
(SWRFVR,
HDACH,
and
CNST)
are
associated
with
a
lower
likelihood
of
having
the influential symptoms that contribute to the diagnoses of malaria and typhoid malaria
among
and typhoid,
whereas GENBDYPN,
BITAIM, CHLNRIG,
ABDPN,
GENBDYPN,
MSCBDYPN,
patients.
Meanwhile,
SWRFVR,
FVR, LTG,
HGGDFVR,
and HDACH
are
FTG, and HGGDFVR
arelikelihood
influentialofsymptoms
that contribute
to the diagnoses of malaria
associated
with a lower
having malaria
and typhoid.
and typhoid
among patients.
It is observed
that medical experts should focus on the following influential
symptoms for the diagnosis of malaria and typhoid fever in patients: BITAIM,
CHNLNRIG, LTG, ABDPN, MSCBDYPN, FVR, GENBDYPN, FTG, and HGGDFVR. This
is consistent with the results of Asuquo et al. [53], where GENBDYPN, CHNLNRIG,
ABPN, FVR, FTG, and HGDFVR were observable symptoms. LIME has numerous
advantages. It explains the individual diagnosis in a form that is relatively easy for
humans to comprehend, aiding healthcare workers to understand why a model made a
specific diagnosis. LIME can be applied to many ML models and this versatility makes it
suitable for various medical diagnostic systems. In addition, LIME is suitable for
generating explanations using local approximations [54]. The limitation of LIME is that it
is computationally intensive and expensive to generate explanations for individual
RF
Algorithmfor
LIME
diagram.
Figure 12. RF
Algorithm
LIME
diagram.
diagnoses,
especially
large
datasets and complex models.
Figure 13 shows the SVM LIME diagram, indicating that CHLNRIG has the highest
feature importance, followed by MSCBDYPN, LTG, ABDPN, BITAIM, FTG, and CNST as
the influential symptoms that contribute to the diagnoses of malaria and typhoid among
patients. Meanwhile, GENBDYPN, SWRFVR, FVR, HGGDFVR, and HDACH are
associated with a lower likelihood of having malaria and typhoid.
It is observed that medical experts should focus on the following influential
symptoms for the diagnosis of malaria and typhoid fever in patients: BITAIM,
CHNLNRIG, LTG, ABDPN, MSCBDYPN, FVR, GENBDYPN, FTG, and HGGDFVR. This
is consistent with the results of Asuquo et al. [53], where GENBDYPN, CHNLNRIG,
ABPN, FVR, FTG, and HGDFVR were observable symptoms. LIME has numerous
advantages. It explains the individual diagnosis in a form that is relatively easy for
humans to comprehend, aiding healthcare workers to understand why a model made a
specific diagnosis. LIME can be applied to many ML models and this versatility makes it
suitable for various medical diagnostic systems. In addition, LIME is suitable for
generating
explanations
using
local approximations [54]. The limitation of LIME is that it
SVM
Algorithm LIME
LIME diagram.
diagram.
Figure 13. SVM
Algorithm
is computationally intensive and expensive to generate explanations for individual
It is observed
thatfor
medical
experts
should
focus
on
the following
influential
diagnoses,
especially
large datasets
and complex
models.
Three
sets of experiments
were conducted
to evaluate
the performance
of symptoms
ChatGPT,
for the diagnosis
of malaria
and typhoid
fever
in typhoid.
patients:In
BITAIM,
CHNLNRIG,
LTG,
Gemini,
and Perplexity
in diagnosing
malaria
and
Experiment
1, one prompt
ABDPN,
MSCBDYPN,
FVR,
GENBDYPN,
FTG,
and
HGGDFVR.
This
is
consistent
at a time was sent to the LLMs for the first 100 patients in the dataset, recordingwith
the

Trop. Med. Infect. Dis. 2024, 9, 216

15 of 23

the results of Asuquo et al. [53], where GENBDYPN, CHNLNRIG, ABPN, FVR, FTG, and
HGDFVR were observable symptoms. LIME has numerous advantages. It explains the
individual diagnosis in a form that is relatively easy for humans to comprehend, aiding
healthcare workers to understand why a model made a specific diagnosis. LIME can be
applied to many ML models and this versatility makes it suitable for various medical
diagnostic systems. In addition, LIME is suitable for generating explanations using local
approximations [54]. The limitation of LIME is that it is computationally intensive and
expensive to generate explanations for individual diagnoses, especially for large datasets
and complex models.
Three sets of experiments were conducted to evaluate the performance of ChatGPT,
Gemini, and Perplexity in diagnosing malaria and typhoid. In Experiment 1, one prompt at
a time was sent to the LLMs for the first 100 patients in the dataset, recording the outputs
in a CSV format to see how they performed with a single set of prompts. Experiment 2
involved sending 100 prompts from the first 100 patients in the dataset to the LLMs and
storing the outputs in a CSV format to observe their responses to a series of prompts. In
Experiment 3, 100 unique prompts were sent to the models repeatedly until the entire
dataset was exhausted in order to assess how the models performed when given large sets
of unique prompts. Table 4 presents the results of the three experiments. In Exp 1, ChatGPT
3.5 has a slightly better performance with the highest F1-score (30.99%); F1-score is crucial
as it balances recall and precision, providing a comprehensive measure of the model’s
performance. Although better accuracy and recall are achieved by ChatGPT 3.5 and Gemini
(30%), Perplexity is better at minimizing false positives with its highest precision (38.90%).
In Exp 2, Perplexity performs better, with the highest F1-score (26.29%), accuracy (28%),
and recall (28%). Because it provides a comprehensive measure of the model’s performance
by balancing recall and precision, the F1-score is especially significant. ChatGPT 3.5 is
better at reducing false positives with the highest precision, while Gemini has the lowest
performance. In Exp 3, ChatGPT 3.5 has better accuracy, precision, and recall, followed by
Gemini and Perplexity. Although the ChatGPT model may have trouble striking a balance
between minimizing false positives and identifying true positives, the model’s relatively
low F1-score suggests that there may be an imbalance between precision and recall.
Table 4. Large language models’ performance.
Experiment

Algorithm

Accuracy

Precision

Recall

F1-Score

Exp 1

Chat GPT 3.5
Gemini
Perplexity

0.3000
0.3000
0.2600

0.35562
0.3449
0.3890

0.3000
0.3000
0.2600

0.30999
0.2908
0.28736

Exp 2

Chat GPT 3.5
Gemini
Perplexity

0.2600
0.2700
0.2800

0.2909
0.2607
0.2524

0.2600
0.2700
0.2800

0.2615
0.2296
0.2629

Exp 3

Chat GPT 3.5
Gemini
Perplexity

0.3297
0.2895
0.2632

0.3324
0.2709
0.1957

0.3297
0.2895
0.2632

0.2926
0.2728
0.2171

Although LLMs have a broad range of knowledge, they may not be specialized in
diagnosing complicated medical conditions like ML models that have been specifically
trained in this area. The low F1-score in Table 4 may be related to LLMs’ limitations
in handling medical diagnosis tasks, particularly diseases with similar or overlapping
symptoms. The three experiments were carried out to test how the LLMs perform in
different scenarios. Exp 1 tests the consistency and reliability of the LLMs in diagnosing
diseases when a single prompt is used at a time. Exp 2 tests the LLMs’ capacity to manage
more inputs concurrently because healthcare systems frequently handle several cases at
the same time. Exp 3 tests the LLMs’ capacity to identify illnesses across a larger dataset
through repeated exposure to various inputs. ChatGPT is an innovative technological tool
for comprehending and processing natural language, making it suitable for interpreting and

Trop. Med. Infect. Dis. 2024, 9, 216

16 of 23

summarizing complex up-to-date information. Gemini is an adaptable tool that can handle
various data types such as images and text, making it suitable for diagnostic purposes.
Perplexity is specialized in comprehending and generating complex queries as well as
maintaining context that can be vital for the retrieval and analysis of medical research.
These LLMs lack specialized knowledge and are also capable of producing inaccurate
answers, which can be critical in a medical context. They require high computational power
to generate and process responses, which could limit real-time systems. Data security
and patient privacy are concerns when handling sensitive medical data and they require
proper validation and regulatory approval before they can be trusted and adopted for
clinical use. To facilitate healthcare professionals’ comprehension of the reasoning behind a
diagnostic output, LLMs integrate and analyze large amounts of medical data and produce
human-readable explanations for their decisions.
The overall ML models’ performance in the study was moderate, suggesting the need
for a sufficient dataset to enhance the diagnostic models. While the traditional SMOTE
aided in balancing the dataset, employing an advanced oversampling method may help in
improving the model performance. Even with GridSearchCV, the hyperparameters might
still be improved, particularly for SVM. Better configurations could result from investigating alternative parameter tuning techniques like RandomizedSearchCV or Bayesian
Optimization. To improve the results of the LLMs, the LLMs will be fine-tuned with a
larger dataset, and an ensemble method will be employed to combine the strengths of
different LLMs.
To integrate ML, XAI, and LLM techniques into an app, we propose two methods.
Method 1: Separate Training and Validation for ML and LLM
1. Train, test, and validate an ML model to diagnose malaria and typhoid based on the
patient dataset
2. Apply LIME to explain the ML models’ diagnoses and how each symptom contributed
to the diagnoses
3. Train, test, and validate an LLM model independently for generating explanations based
on the patient dataset
4. Integrate the outputs from ML, LIME, and LLM to provide a comprehensive and
interpretable diagnosis.
The advantage of method 1 is that it might lead to higher diagnostic performance
considering the specific training of the two models (ML and LLM) for this task. The
disadvantage is that the training and validation process of two independent models would
increase the computational complexity of the diagnostic system, especially in combining
the results to ensure consistency and coherence.
Method 2: Integrated ML, LIME, and LLM Process
1. Train, test, and validate an ML model to diagnose malaria and typhoid based on the
patient dataset
2. Apply LIME to explain the ML models’ diagnoses and how each symptom contributed
to the diagnoses
3. Use LLM for further explainability by passing the patient symptoms and ML results
(with LIME explanations) through the model to generate diagnostic explanations in natural
language.
The advantage of method 2 is its simplicity because an integrated pipeline reduces
complexity, making the system easier to develop, test, and maintain, which we have implemented. Performance will be increased and computational overhead could be decreased by
streamlining the procedure into a single pipeline. The explanations produced by LIME are
directly considered by the LLM, which results in more logical and contextually appropriate
explanations. The disadvantage is that the quality of the initial ML and LIME outputs
determines the quality of the explanations provided by the LLM.
The Malaria and Typhoid Fever Diagnosis System is a mobile app that healthcare
workers can use to diagnose typhoid fever and malaria with an easy-to-use interface. The

Trop. Med. Infect. Dis. 2024, 9, 216

17 of 23

system comprises user authentication, a User main dashboard, and a Patient dashboard.
The basic app requirement is an Android OS version 4.0 and above, 4 GB RAM: 2 GB
minimum, ROM: 8 GB minimum, Display Layout: Portrait, and Internet connection. The
user login is shown in Figure 14 and the User main dashboard is shown in Figure 15. The
healthcare worker can register a patient, view a list of patients, and set up appointments for
patients. Figure 16 is the patient registration form while Figure 17 is the patient dashboard
where the patient vitals can be entered as well as the history taking and examination in
Figure 18. The patient’s provisional diagnosis with XAI results is shown in Figure 19
with
Trop. Med. Infect. Dis. 2024, 9, x FOR PEER REVIEW
24 of 30
LIME
plot displaying the symptoms and how they influenced the model’s decision
and
Trop. Med. Infect. Dis. 2024, 9, x FOR the
PEER
REVIEW
24 of
30
the explanation by the ChatGPT LLM.

Figure14.
14.User
UserLogin.
Login.
Figure
Figure 14. User Login.

Figure 15. User Main Dashboard.
Figure15.
15.User
UserMain
MainDashboard.
Dashboard.
Figure

Trop.
Trop.Med.
Med.Infect.
Infect.Dis.
Dis.2024,
2024,9,9,216
x FOR PEER REVIEW
Trop. Med. Infect. Dis. 2024, 9, x FOR PEER REVIEW

Figure 16. Patient
Patient Registration.
Figure
Figure16.
16. Patient Registration.
Registration.

Figure 17. Patient Account Dashboard.
Figure17.
17. Patient
Patient Account
Account Dashboard.
Dashboard.
Figure

25
30
18 of 23
25 of 30

Trop.Med.
Med.Infect.
Infect.Dis.
Dis.2024,
2024,9,9,216
x FOR PEER REVIEW
Trop.
Trop. Med. Infect. Dis. 2024, 9, x FOR PEER REVIEW

26
of 23
30
19 of
26 of 30

Figure18.
18.History
HistoryTaking
Takingand
andExamination.
Examination.
Figure
Figure 18. History Taking and Examination.

Figure 19. XAI Diagnosis Results.
Figure19.
19.XAI
XAIDiagnosis
DiagnosisResults.
Results.
Figure

Previous studies [20,55,56] applied ML models for diagnosing malaria and typhoid
Previous
studies
[20,55,56]
applied
MLmodels
models
fordiagnosing
diagnosing
malaria
andtyphoid
typhoid
Previous
[20,55,56]
applied
ML
for
and
fever,
thoughstudies
these studies
lacked
appropriate
interpretability
in malaria
the decision-making
fever,
though
these
studies
lacked
appropriate
interpretability
in
the
decision-making
fever,
though
these
studies
lacked
appropriate
interpretability
in
the
decision-making
process, which often results in medical experts having difficulties in comprehending the
process,
which
often
resultsin
in
medical
experts
havingdifficulties
difficulties
inand
comprehending
the
process,
which
often
results
medical
experts
in
comprehending
the
reasoning
behind
diagnostic
results.
This
studyhaving
integrated
ML, XAI,
LLM to enhance
reasoning
behind
diagnostic
results.in
This
study
integrated
ML,XAI,
XAI,
and
LLMwith
toenhance
enhance
reasoning
behind
results.
This
integrated
ML,
and
LLM
to
transparency
anddiagnostic
interpretability
thestudy
diagnostic
processes
that
align
global
transparency
and
interpretability
in
the
diagnostic
processes
that
align
with
global
healthtransparency
and
interpretability
in
the
diagnostic
processes
that
align
with
global
healthcare goals. The use of LIME for feature importance analyses and ChatGPT
for
care
goals.
The
use
of
LIME
for
feature
importance
analyses
and
ChatGPT
for
generating
healthcare
goals.
The
use
of
LIME
for
feature
importance
analyses
and
ChatGPT
for
generating context-aware explanations have distinguished the present study. Several
context-aware
explanations explanations
have distinguished
the present study.
Several
factors
can
congenerating
context-aware
have
distinguished
the
present
study.
Several
factors can contribute to the low performance scores in Table 4. These include: (1) the
tribute
to thecontribute
low performance
scores
in Table 4. These
include:
(1)4.the
dataset
used during
factors
to training
the low
performance
scores
Table
These
include:
(1) the
dataset can
used during the
is
limited in size
andindiversity,
affecting
the models’
dataset used during the training is limited in size and diversity, affecting the models’

Trop. Med. Infect. Dis. 2024, 9, 216

20 of 23

the training is limited in size and diversity, affecting the models’ ability to generalize to
unseen cases; (2) LLMs may require further fine-tuning and optimization, as the complexity
of the diseases being diagnosed may overlap with other illnesses, thereby challenging the
models to accurately differentiate between them. Furthermore, LLMs did not show high
domain tolerance to the investigated illnesses, hence fine-tuning them on domain-specific
data can significantly improve their performance.
4. Conclusions
This study creates a medical diagnostic framework for Malaria and Typhoid fever by
integrating XAI, LLMs, and ML models. This approach aims to demystify the black-box
nature of ML models, offering transparent insights into how each feature or symptom
affects the diagnosis. The RF model showed superior prediction performance in terms of
accuracy, recall, precision, and F1-score compared to XGBoost and SVM. The high recall
and precision values in RF are crucial for accurately diagnosing these diseases and for
making appropriate treatment decisions. However, XGBoost exhibited the lowest log loss
and fastest computation time. Further analysis indicates that SVM performs worse than
the other two models, making it less suitable for this dataset. The study suggests that
ensemble techniques like RF and XGBoost better capture the complex relationships between
symptoms and diseases. The XAI analysis identified BITAIM, CHNLNRIG, LTG, ABDPN,
MSCBDYPN, FVR, GENBDYPN, FTG, and HGGDFVR as key features for predicting
Malaria and Typhoid. Among LLMs, ChatGPT 3.5 performed slightly better than Gemini
and Perplexity. This study has shown how RF, LIME, and GPT can be used effectively to
diagnose typhoid fever and malaria using a mobile-based system that meets the crucial
requirements of interpretability and transparency, improving the diagnostic process’s
acceptability and understanding among medical professionals. Future research should
examine the application of various machine learning models, XAI techniques, and LLMs
on a variety of datasets and across other medical conditions, such as in the diagnosis of
diabetes, cardiovascular diseases, and cancer detection, to further validate and generalize
the findings of this study. The validity of AI-driven diagnostics can be strengthened by
extending its application to additional medical conditions. This will ultimately improve
patient outcomes in a range of healthcare domains.
Author Contributions: Conceptualization, F.-M.U. and K.A.; methodology, K.A., C.A. (Constance
Amannah), D.A. and M.E.; validation, F.-M.U., O.O., K.A., C.A. (Constance Amannah), D.A. and
M.E.; formal analysis, K.A., P.A. and D.A.; data curation, K.A. and P.A.; writing—original draft
preparation, K.A., D.A., P.A., E.J., A.J. and M.E.; writing—review and editing, K.A., M.E., D.A.,
C.A. (Constance Amannah), O.O., C.A. (Christie Akwaowo), O.M. and F.-M.U. supervision, F.-M.U.,
C.A. (Constance Amannah), D.A., O.O. and M.E.; project administration, F.-M.U. and O.O.; funding
acquisition, F.-M.U. All authors have read and agreed to the published version of the manuscript.
Funding: This research was funded by the New Frontier Research Fund, grant number NFRFE-201901365 between April and March 2024.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: The raw data supporting the conclusions of this article will be made
available by the authors on request.
Conflicts of Interest: The authors declare no conflicts of interest.

References
1.
2.

Asuquo, D.; Attai, K.; Obot, O.; Ekpenyong, M.; Akwaowo, C.; Arnold, K.; Uzoka, F.M. Febrile Disease Modeling and Diagnosis
System for Optimizing Medical Decisions in Resource-Scarce Settings. Clin. eHealth 2024, 7, 52–76. [CrossRef]
Sohanang-Nodem, F.S.; Ymele, D.; Fadimatou, M.; Fodouop, S.C. Malaria and Typhoid Fever Coinfection among Febrile Patients
in Ngaoundéré (Adamawa, Cameroon): A Cross-Sectional Study. J. Parasitol. Res. 2023, 2023, 5334813. [CrossRef] [PubMed]

Trop. Med. Infect. Dis. 2024, 9, 216

3.
4.
5.
6.
7.
8.
9.
10.
11.

12.
13.
14.
15.
16.

17.
18.
19.

20.

21.
22.
23.
24.
25.

26.

27.

21 of 23

Galán, J.E. Typhoid Toxin Provides a Window into Typhoid Fever and the Biology of Salmonella Typhi. Proc. Natl. Acad. Sci. USA
2016, 113, 6338–6344. [CrossRef]
Gashaw, T.; Jambo, A. Typhoid in Less Developed Countries: A Major Public Health Concern. In Hygiene and Health in Developing
Countries—Recent Advances; IntechOpen: Rijeka, Croatia, 2022. [CrossRef]
Alhumaid, N.K.; Alajmi, A.M.; Alosaimi, N.F.; Alotaibi, M.; Almangour, T.A.; Nassar, M.S.; Tawfik, E.A. Reported Bacterial
Infectious Diseases in Saudi Arabia: Overview and Recent Advances. Res. Sq. 2023, 1–39. [CrossRef]
Paton, D.G.; Childs, L.M.; Itoe, M.A.; Holmdahl, I.E.; Buckee, C.O.; Catteruccia, F. Exposing Anopheles Mosquitoes to Antimalarials
Blocks Plasmodium Parasite Transmission. Nature 2019, 567, 239–243. [CrossRef]
Sato, S. Plasmodium—A Brief Introduction to the Parasites Causing Human Malaria and Their Basic Biology. J. Physiol. Anthropol.
2021, 40, 1. [CrossRef]
Carson, B.B., III. Mosquitos and Malaria Take a Toll. In Challenging Malaria: The Private and Social Incentives of Mosquito Control;
Springer International Publishing: Cham, Switzerland, 2023; pp. 15–25. [CrossRef]
Bria, Y.P.; Yeh, C.H.; Bedingfield, S. Significant Symptoms and Nonsymptom-Related Factors for Malaria Diagnosis in Endemic
Regions of Indonesia. Int. J. Infect. Dis. 2021, 103, 194–200. [CrossRef] [PubMed]
Attai, K.; Amannejad, Y.; Vahdat Pour, M.; Obot, O.; Uzoka, F.M. A Systematic Review of Applications of Machine Learning and
Other Soft Computing Techniques for the Diagnosis of Tropical Diseases. Trop. Med. Infect. Dis. 2022, 7, 398. [CrossRef]
Bosco, A.B.; Nankabirwa, J.I.; Yeka, A.; Nsobya, S.; Gresty, K.; Anderson, K.; Mbaka, P.; Prosser, C.; Smith, D.; Opigo, J.; et al.
Limitations of Rapid Diagnostic Tests in Malaria Surveys in Areas with Varied Transmission Intensity in Uganda 2017–2019:
Implications for Selection and Use of HRP2 RDTs. PLoS ONE 2020, 15, e0244457. [CrossRef] [PubMed]
Mohan, F.R.; Jaber, A.S. Role of the Widal Test in Diagnosing Typhoid Fever Compared with Culture at Teaching Al-Hussein
Hospital in Nasiriyah. Peerian J. 2024, 26, 72–78.
Mather, R.G.; Hopkins, H.; Parry, C.M.; Dittrich, S. Redefining Typhoid Diagnosis: What Would an Improved Test Need to Look
Like? BMJ Glob. Health 2019, 4, e001831. [CrossRef] [PubMed]
Boina, R.; Ganage, D.; Chincholkar, Y.D.; Wagh, S.; Shah, D.U.; Chinthamu, N.; Shrivastava, A. Enhancing Intelligence Diagnostic
Accuracy Based on Machine Learning Disease Classification. Int. J. Intell. Syst. Appl. Eng. 2023, 11, 765–774.
Asuquo, D.E.; Umoren, I.; Osang, F.; Attai, K. A Machine Learning Framework for Length of Stay Minimization in Healthcare
Emergency Department. Stud. Eng. Technol. J. 2023, 10, 1–17. [CrossRef]
Muhammad, B.; Varol, A. A Symptom-Based Machine Learning Model for Malaria Diagnosis in Nigeria. In Proceedings of the
2021 9th International Symposium on Digital Forensics and Security (ISDFS), Elazig, Turkey, 28–29 June 2021; IEEE: Piscataway,
NJ, USA, 2021; pp. 1–6. [CrossRef]
Barracloug, P.A.; Were, C.M.; Mwangakala, H.; Fehringer, G.; Ohanya, D.O.; Agola, H.; Nandi, P. Artificial Intelligence System for
Malaria Diagnosis. Int. J. Adv. Comput. Sci. Appl. 2024, 15, 100806. [CrossRef]
La-Ariandi, H.; Setyanto, A.; Sudarmawan, S. Classification of Malaria Types Using Naïve Bayes Classification. J. Indones. Sos.
Teknol. 2024, 5, 2311–2327. [CrossRef]
Bhuiyan, M.A.; Rad, S.S.; Johora, F.T.; Islam, A.; Hossain, M.I.; Khan, A.A. Prediction of Typhoid Using Machine Learning and
ANN Prior to Clinical Test. In Proceedings of the 2023 International Conference on Computer Communication and Informatics
(ICCCI), Coimbatore, India, 23–25 January 2023; IEEE: Piscataway, NJ, USA, 2023; pp. 1–7. [CrossRef]
Awotunde, J.B.; Imoize, A.L.; Salako, D.P.; Farhaoui, Y. An Enhanced Medical Diagnosis System for Malaria and Typhoid Fever
Using Genetic Neuro-Fuzzy System. In Proceedings of the International Conference on Artificial Intelligence and Smart Environment, Errachidia, Morocco, 24–26 November 2022; Springer International Publishing: Cham, Switzerland, 2022; pp. 173–183.
[CrossRef]
Odion, P.O.; Ogbonnia, E.O. Web-Based Diagnosis of Typhoid and Malaria Using Machine Learning. Niger. Def. Acad. J. Mil. Sci.
Interdiscip. Stud. 2022, 1, 89–103.
Apanisile, T.; Ayeni, J.A. Development of an Extended Medical Diagnostic System for Typhoid and Malaria Fever. Artif. Intell.
Adv. 2023, 5, 28–40. [CrossRef]
Anderson, J.; Thomas, J. Interpretable Machine Learning Models for Healthcare Applications; EasyChair: Manchester, UK, 2024;
p. 12358.
Kiseleva, A.; Kotzinos, D.; De Hert, P. Transparency of AI in Healthcare as a Multilayered System of Accountabilities: Between
Legal Requirements and Technical Limitations. Front. Artif. Intell. 2022, 5, 879603. [CrossRef]
Albahri, A.S.; Duhaim, A.M.; Fadhel, M.A.; Alnoor, A.; Baqer, N.S.; Alzubaidi, L.; Deveci, M. A Systematic Review of Trustworthy
and Explainable Artificial Intelligence in Healthcare: Assessment of Quality, Bias Risk, and Data Fusion. Inf. Fusion 2023,
96, 156–191. [CrossRef]
Tan, L.; Huang, C.; Yao, X. A Concept-Based Local Interpretable Model-Agnostic Explanation Approach for Deep Neural Networks
in Image Classification. In Proceedings of the International Conference on Intelligent Information Processing, Shenzhen, China,
3–6 May 2024; Springer Nature: Cham, Switzerland, 2024; pp. 119–133. [CrossRef]
Thombre, A. Comparison of Decision Trees with Local Interpretable Model-Agnostic Explanations (LIME) Technique and MultiLinear Regression for Explaining Support Vector Regression Model in Terms of Root Mean Square Error (RMSE) Values. arXiv
2024, arXiv:2404.07046. [CrossRef]

Trop. Med. Infect. Dis. 2024, 9, 216

28.

29.

30.

31.

32.

33.
34.

35.
36.
37.
38.

39.
40.
41.
42.
43.

44.
45.

46.
47.
48.
49.
50.

51.

22 of 23

Okay, F.Y.; Yıldırım, M.; Özdemir, S. Interpretable Machine Learning: A Case Study of Healthcare. In Proceedings of the 2021
International Symposium on Networks, Computers and Communications (ISNCC), Dubai, United Arab Emirates, 31 October–2
November 2021; IEEE: Piscataway, NJ, USA, 2021; pp. 1–6. [CrossRef]
Attai, K.; Akwaowo, C.; Asuquo, D.; Esubok, N.E.; Nelson, U.A.; Dan, E.; Uzoka, F.M. Explainable AI Modelling of Comorbidity
in Pregnant Women and Children with Tropical Febrile Conditions. In Proceedings of the International Conference on Artificial
Intelligence and Its Applications, Mauritius, East Africa, 9–10 November 2023; pp. 152–159. [CrossRef]
Ashraf, K.; Nawar, S.; Hosen, M.H.; Islam, M.T.; Uddin, M.N. Beyond the Black Box: Employing LIME and SHAP for Transparent
Health Predictions with Machine Learning Models. In Proceedings of the 2024 International Conference on Advances in
Computing, Communication, Electrical, and Smart Systems (iCACCESS) , Dhaka, Bangladesh, 8–9 March 2024; IEEE: Piscataway,
NJ, USA, 2024; pp. 1–6. [CrossRef]
Li, F.; Jin, Y.; Liu, W.; Rawat, B.P.S.; Cai, P.; Yu, H. Fine-Tuning Bidirectional Encoder Representations from Transformers
(BERT)–Based Models on Large-Scale Electronic Health Record Notes: An Empirical Study. JMIR Med. Inform. 2019, 7, e14830.
[CrossRef] [PubMed]
Nakamura, Y.; Hanaoka, S.; Nomura, Y.; Nakao, T.; Miki, S.; Watadani, T.; Abe, O. Automatic Detection of Actionable Radiology
Reports Using Bidirectional Encoder Representations from Transformers. BMC Med. Inform. Decis. Mak. 2021, 21, 262. [CrossRef]
[PubMed]
Gorenstein, L.; Konen, E.; Green, M.; Klang, E. BERT in Radiology: A Systematic Review of Natural Language Processing
Applications. J. Am. Coll. Radiol. 2024, 21, 914–941. [CrossRef]
Yenduri, G.; Ramalingam, M.; Selvi, G.C.; Supriya, Y.; Srivastava, G.; Maddikunta, P.K.R.; Gadekallu, T.R. GPT (Generative
Pre-Trained Transformer)–A Comprehensive Review on Enabling Technologies, Potential Applications, Emerging Challenges,
and Future Directions. IEEE Access 2024, 12, 54608–54649. [CrossRef]
Wang, Z.; Guo, R.; Sun, P.; Qian, L.; Hu, X. Enhancing Diagnostic Accuracy and Efficiency with GPT-4-Generated Structured
Reports: A Comprehensive Study. J. Med. Biol. Eng. 2024, 44, 144–153. [CrossRef]
Islam, M.M.; Alam, M.J.; Maniruzzaman, M.; Ahmed, N.F.; Ali, M.S.; Rahman, M.J.; Roy, D.C. Predicting the Risk of Hypertension
Using Machine Learning Algorithms: A Cross Sectional Study in Ethiopia. PLoS ONE 2023, 18, e0289613. [CrossRef]
Silva-Aravena, F.; Núñez Delafuente, H.; Gutiérrez-Bahamondes, J.H.; Morales, J. A Hybrid Algorithm of ML and XAI to Prevent
Breast Cancer: A Strategy to Support Decision Making. Cancers 2023, 15, 2443. [CrossRef]
Zhu, T.; Liu, X.; Wang, J.; Kou, R.; Hu, Y.; Yuan, M.; Zhang, W. Explainable Machine-Learning Algorithms to Differentiate
Bipolar Disorder from Major Depressive Disorder Using Self-Reported Symptoms, Vital Signs, and Blood-Based Markers. Comput.
Methods Programs Biomed. 2023, 240, 107723. [CrossRef]
Fan, Y.; Lu, X.; Sun, G. IHCP: Interpretable Hepatitis C Prediction System Based on Black-Box Machine Learning Models. BMC
Bioinf. 2023, 24, 333. [CrossRef]
Jin, M.; Yu, Q.; Zhang, C.; Shu, D.; Zhu, S.; Du, M.; Meng, Y. Health-LLM: Personalized Retrieval-Augmented Disease Prediction
Model. arXiv 2024, arXiv:2402.00746.
Panagoulias, D.P.; Virvou, M.; Tsihrintzis, G.A. Evaluating LLM-Generated Multimodal Diagnosis from Medical Images and
Symptom Analysis. arXiv 2024, arXiv:2402.01730.
Hariri, W. Analyzing the Performance of ChatGPT in Cardiology and Vascular Pathologies. arXiv 2023, arXiv:2307.02518.
Kusa, W.; Mosca, E.; Lipani, A. “Dr LLM, What Do I Have?”: The Impact of User Beliefs and Prompt Formulation on Health
Diagnoses. In Proceedings of the Third Workshop on NLP for Medical Conversations, Nusa Dua, Indonesia, 1 November 2023;
pp. 13–19.
University of Uyo Teaching Hospital; Mount Royal University. NFRF Project Patient Dataset with Febrile Diseases [Data Set]; Zenodo:
Bern, Switzerland, 2024. [CrossRef]
Caruccio, L.; Cirillo, S.; Polese, G.; Solimando, G.; Sundaramurthy, S.; Tortora, G. Can ChatGPT Provide Intelligent Diagnoses? A
Comparative Study Between Predictive Models and ChatGPT to Define a New Medical Diagnostic Bot. Expert Syst. Appl. 2024,
235, 121186. [CrossRef]
Han, H.; Zhang, Z.; Cui, X.; Meng, Q. Ensemble Learning with Member Optimization for Fault Diagnosis of a Building Energy
System. Energy Build. 2020, 226, 110351. [CrossRef]
Zhu, L.; Qiu, D.; Ergu, D.; Ying, C.; Liu, K. A Study on Predicting Loan Default Based on the Random Forest Algorithm. Procedia
Comput. Sci. 2019, 162, 503–513. [CrossRef]
Ghosh, D.; Cabrera, J. Enriched Random Forest for High Dimensional Genomic Data. IEEE/ACM Trans. Comput. Biol. Bioinform.
2021, 19, 2817–2828. [CrossRef]
Jackins, V.; Vimal, S.; Kaliappan, M.; Lee, M.Y. AI-Based Smart Prediction of Clinical Disease Using Random Forest Classifier and
Naive Bayes. J. Supercomput. 2021, 77, 5198–5219. [CrossRef]
Palimkar, P.; Shaw, R.N.; Ghosh, A. Machine Learning Technique to Prognosis Diabetes Disease: Random Forest Classifier
Approach. In Advanced Computing and Intelligent Technologies: Proceedings of ICACIT 2021; Springer: Singapore, 2022; pp. 219–244.
[CrossRef]
Asselman, A.; Khaldi, M.; Aammou, S. Enhancing the Prediction of Student Performance Based on the Machine Learning XGBoost
Algorithm. Interact. Learn. Environ. 2023, 31, 3360–3379. [CrossRef]

Trop. Med. Infect. Dis. 2024, 9, 216

52.
53.
54.
55.
56.

23 of 23

Devikanniga, D.; Ramu, A.; Haldorai, A. Efficient Diagnosis of Liver Disease Using Support Vector Machine Optimized with
Crows Search Algorithm. EAI Endorsed Trans. Energy Web 2020, 7, e10. [CrossRef]
Asuquo, D.E.; Attai, K.F.; Johnson, E.A.; Obot, O.U.; Adeoye, O.S.; Akwaowo, C.D.; Uzoka, F.M.E. Multi-Criteria Decision
Analysis Method for Differential Diagnosis of Tropical Febrile Diseases. Health Inform. J. 2024, 30, 1–41. [CrossRef]
Salih, A.M.; Raisi-Estabragh, Z.; Galazzo, I.B.; Radeva, P.; Petersen, S.E.; Lekadir, K.; Menegaz, G. A Perspective on Explainable
Artificial Intelligence Methods: SHAP and LIME. arXiv 2024, arXiv:2305.02012. [CrossRef]
Maidabara, A.H.; Ahmadu, A.S.; Malgwi, Y.M.; Ibrahim, D. Expert System for Diagnosis of Malaria and Typhoid. Comput. Sci. IT
Res. J. 2021, 2, 1–15. [CrossRef]
Mariki, M.; Mkoba, E.; Mduma, N. Combining Clinical Symptoms and Patient Features for Malaria Diagnosis: Machine Learning
Approach. Appl. Artif. Intell. 2022, 36, 2031826. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.