Bulletin of Electrical Engineering and Informatics

Vol. 13, No. 2, April 2024, pp. 1276~1285

ISSN: 2302-9285, DOI: 10.11591/eei.v13i2.6280  1276

Predicting lung cancer risk using explainable artificial

intelligence

Shahin Shoukat Makubhai, Ganesh R. Pathak, Pankaj R. Chandre

Department of Computer Science and Engineering, MIT School of Computing, MIT Art, Design and Technology University, Loni
Kalbhor, India

Article Info ABSTRACT

Article history: Lung cancer is a lethal disease that claims numerous lives annually, and
early detection is essential for improving survival rates. Machine learning
Received Mar 20, 2023 has shown promise in predicting lung cancer risk, but the lack of
Revised Jul 17, 2023 transparency and interpretability in black-box models impedes the
Accepted Oct 6, 2023 understanding of factors that contribute to risk. Explainable artificial
intelligence (XAI) can overcome this limitation by providing a clear and
understandable approach to machine learning. In this study, we will use a
Keywords: large patient record dataset to train an XAI-based model that considers
various patient information, including lifestyle factors, clinical data, and
Explainable artificial medical history, for predicting lung cancer risk. We will use different XAI
intelligence techniques, including decision trees, partial dependence plots, and feature
Interpretability importance, to interpret the model’s predictions. These methods will provide
Lung cancer healthcare professionals with a transparent and interpretable framework for
Machine learning screening and treatment decisions concerning lung cancer risk.
Prediction
This is an open access article under the CC BY-SA license.
Risk factors

Corresponding Author:
Shahin Shoukat Makubhai
Department of Computer Science and Engineering, MIT School of Computing
MIT Art, Design and Technology University
Loni Kalbhor, India
Email: [Link]@[Link]

1. INTRODUCTION
Lung cancer is one of the most common and deadly forms of cancer worldwide. It is estimated that
lung cancer accounts for 2.09 million new cases and 1.76 million deaths each year. Early detection and
accurate diagnosis of lung cancer are essential for improving patient outcomes and reducing mortality rates [1].
A branch of artificial intelligence (AI) known as explainable AI (XAI) aims to develop systems that are
simple enough for people to understand [2]. XAI is especially significant in the healthcare industry, where
gaining the confidence of physicians and patients requires being able to explain how a machine learning
model is making predictions.
Creating a machine learning model that can estimate a person’s probability of getting lung cancer
based on various risk factors is required to predict lung cancer risk using XAI. Personal traits like age, sex,
smoking history, family history, and risk factors include examples such as being exposed to environmental
toxins [3]. It would be necessary to train a sizable dataset of lung cancer patients and healthy individuals for
the XAI model used to predict the chance of developing lung cancer [4]. To guarantee that the dataset is a
representative of the general population and contains a diverse range of people with varying risk factors, it
would need to be carefully curated.
Once trained, the XAI model can be used to estimate a person’s chance of developing lung cancer
based on their risk factors. Additionally, the XAI model would be able to explain how it got to its prediction [5].

Journal homepage: [Link]

Bulletin of Electr Eng & Inf ISSN: 2302-9285  1277

The most significant risk variables that went into the prediction would be listed or visualized in this
explanation. Gaining the confidence of physicians and patients requires being able to explain how the XAI
model arrived at its forecast [6]. The XAI model can help doctors make informed decisions, and give patients
tailored suggestions based on their risk factors. The XAI model can help patients make wise choices about
their health and take measures to prevent lung cancer and reduce the risk of its development.
The use of XAI to predict the likelihood of developing lung cancer is a novel strategy that has the
potential to significantly advance cancer diagnosis and therapy [7]. Although using AI in healthcare is not
novel, explaining how AI models make decisions can help boost confidence in the outcomes and give
healthcare professionals insightful information. One of the main causes of cancer-related deaths globally, and
a significant public health concern is lung cancer [8]. Finding people who are at a high risk of getting lung
cancer, can help with early diagnosis and possibly enhance patient outcomes [9]. Despite the fact that there
are a number of established risk factors for lung cancer, including smoking and exposure to toxins in the
environment, it is still difficult to identify those who are at high risk.
The development of models that can precisely predict a person’s risk of getting lung cancer based on
various risk factors, such as age, smoking history, and family history of cancer, may be feasible using XAI [10].
Additionally, by comprehending the AI model’s decision-making process, medical professionals can learn
important information about the risk factors more closely linked to lung cancer, which can help guide in
prevention and treatment plans [11]. Overall, the application of XAI in predicting lung cancer risk has the
potential to significantly improve cancer prevention and treatment, ultimately leading to better patient
outcomes [12].
In conclusion, using XAI to forecast lung cancer risk in healthcare is a significant application of
artificial intelligence. Using XAI models, we can accurately predict an individual’s likelihood of getting lung
cancer and explain how the prediction was made. This knowledge can empower doctors and patients to make
informed healthcare decisions that may ultimately save lives.
Within this section, you will find comparable studies addressing intrusion detection and prevention.
Below is another analysis of the existing literature on this subject. An important subject that has received a
great deal of attention recently, is the use of XAI to predict the likelihood of developing lung cancer. In this
literature review, we will explore a range of research papers, journals, and publications pertaining to the
utilization of XAI for forecasting the probability of lung cancer development.
The paper entitled “early lung cancer diagnostic biomarker discovery by machine learning methods”
by Xie et al. [13] proposes the use of machine learning methods to identify potential biomarkers for early
detection of lung cancer. The authors start by outlining the present difficulties and restrictions in lung cancer
diagnosis and stress the importance of early detection for better patient outcomes. The authors then go on to
explain the study’s methodology, which involved using a variety of ML algorithms to analyze gene
expression data from lung cancer patients and healthy controls in order to find possible biomarkers. The
findings demonstrated that possible biomarkers for early detection of lung cancer could be found by using
machine learning techniques. The study, according to the authors, offers a hopeful direction for additional
investigation, and the creation of early-detection tools for lung cancer.
The paper entitled “biomarkers in lung cancer screening: achievements, promises, and challenges”
by Seijo et al. [14] provides an overview of the current state of biomarkers in lung cancer screening. The
article starts by emphasizing the significance of early lung cancer detection, which is frequently challenging
because symptoms frequently do not appear until the disease has advanced to an advanced stage. The author
then goes over the drawbacks of conventional detection techniques like low-dose computerized tomography
(CT) scans and chest X-rays. The idea of biomarkers is then introduced by the author. These are quantifiable
illness indicators that can be found in biological samples like blood, urine, or tissue. Including genetic,
epigenetic, and protein biomarkers, the author also gives an idea of the various categories of biomarkers. The
paper then reviews the state of biomarker research for lung cancer screening today, highlighting some of the
most promising biomarkers found so far. The author talks about how using biomarkers in screening could
increase the likelihood of early discovery and decrease false positives. The article ends with a discussion of
the difficulties confronting the field of lung cancer screening biomarker research. Large-scale validation
studies, the creation of standardized assays, and the incorporation of biomarkers into already-in-use screening
procedures are some of these difficulties.
The paper entitled “lung cancer prediction by deep learning to identify benign lung nodules” by
Heuvelmans et al. [15] targeted to develop a dl model to predict lung cancer, and identify benign lung
nodules. The use of deep learning models in medical imaging, particularly in the diagnosis of lung cancer,
was reviewed in the literature by the author. Deep learning models have the potential to increase lung cancer
detection and decrease false-positive rates, according to the study. The research also covered deep learning
models’ drawbacks, such as the need for sizable datasets and their lack of interpretability. Overall, the
literature review stressed the need for more research in this field and supported the use of deep learning
models in lung cancer prediction.
Predicting lung cancer risk using explainable artificial intelligence (Shahin Shoukat Makubhai)
1278  ISSN: 2302-9285

The paper entitled “artificial intelligence in cancer imaging: clinical challenges and applications” by
Bi et al. [16] analyzes the potential of artificial intelligence in the field of cancer imaging. The study starts
off with an overview of the problems that are currently plaguing cancer imaging, such as the shortcomings of
conventional imaging techniques like CT and MRI and the demand for more precise and effective cancer
detection techniques. The survey then examines how more precise and effective methods of diagnosis and
treatment could revolutionize cancer imaging with the help of AI. The authors discusses various AI methods,
such as computer vision and deep learning, and how they can be used with various cancer imaging
approaches, like mammography and radio mics. The study also covers the difficulties and restrictions
associated with applying AI to cancer imaging, including the need for a significant amount of high-quality
data and the possibility of bias in AI algorithms. The author emphasizes the importance of collaboration
between clinicians and AI researchers to overcome these challenges and develop effective AI-based cancer
imaging systems. Overall, the literature survey imposes the potential of AI in revolutionizing cancer imaging
and the need for continued research and collaboration to realize this potential.
The paper entitled “bias in data-driven artificial intelligence systems-an introductory survey” by
Ntoutsi et al. [17] provides an overview of the issue of bias in AI systems that are trained on large datasets.
The paper defines bias in AI and gives examples of how it can appear in various situations, such as when
making lending and hiring choices. The nature of the data being used, the algorithms used to process the data,
and the societal and cultural contexts in which the AI systems are created and implemented are just a few of
the topics covered in this piece that can cause bias in AI. The author also discusses various strategies that
have been put forth to deal with bias in AI, including the use of fairness measures and the creation of moral
standards for AI development. Overall, the article highlights the importance of understanding and addressing
bias in AI systems, as these systems increasingly play a role in decision-making processes in a wide range of
industries and sectors.
The paper entitled “lung cancer detection and classification by using machine learning and
multinomial Bayesian” by Dwivedi et al. [18] explores the application of machine learning and Bayesian
methods in the detection and classification of lung cancer. The survey starts off by stressing the significance
of early lung cancer detection, which can considerably enhance patient outcomes. The author then gives an
outline of Bayesian and machine learning techniques and talks about how they might be used to identify and
categorize lung cancer. After that, the survey examines a number of studies that have used machine learning
and Bayesian techniques to identify and categorize lung cancer, including studies that have applied deep
learning methods and studies that have applied Bayesian networks. The benefits and shortcomings of these
techniques are also discussed by the author. The findings comes to a concluding with a discussion of the
potential future paths for research in the field of machine learning and Bayesian methods for lung cancer
detection and classification. Although these approaches have shown potential, the author points out that more
study is required before precise models that can be applied in clinical practice can be created.
The paper entitled “a neural network and optimization based lung cancer detection system in CT
images” by Venkatesh et al. [19] proposes a system for lung cancer detection using neural networks and
optimization techniques. The system analyses lung CT images to extract pertinent features using a CNN. The
most significant features from the extracted features are then prioritised using genetic algorithms, and these
are fed into a support vector machine classifier for the final classification of the CT picture as either
cancerous or non-cancerous. The pre-processing of CT images, design and training of the CNN, optimisation
of the features, and implementation of the SVM classifier are just a few of the processes covered in the
article. The article also includes experimental findings on a publicly accessible dataset that demonstrates the
high accuracy of the proposed system in identifying lung cancer in CT images. Overall, the study offers an
intriguing method for detecting lung cancer using a fusion of deep learning and optimization techniques, and
it offers encouraging findings that may help to advance the creation of improved lung cancer screening
techniques.
The paper entitled “explainable machine learning framework for lung cancer hospital length of stay
prediction” by Alsinglawi et al. [8] develops a machine learning framework to predict the length of stay for
patients with lung cancer in hospitals. The goal of the framework is to be explainable, which means that it
can offer concise and understandable justifications for its forecasts, which is crucial for clinical decision-
making. The authors reviewed current machine learning methods for predicting hospital lengths of stay in
addition to explainable machine learning approaches. The authors discovered that although machine learning
models can successfully predict duration of stay with high accuracy, they frequently lack interpretability,
which reduces their usefulness in clinical settings. The authors suggested a hybrid strategy, that combines a
rule-based model and a machine-learning model to handle this problem. The rule-based model is used to
produce explanations for the prediction, while the machine-learning model is used to predict the length of
stay. The machine learning model’s features are used by the rule-based model to produce explanations that
give a precise and understandable account of how the prediction was made. The authors compared their

Bulletin of Electr Eng & Inf, Vol. 13, No. 2, April 2024: 1276-1285
Bulletin of Electr Eng & Inf ISSN: 2302-9285  1279

framework to several other machine learning algorithms using a dataset of lung cancer patients from a
hospital. The outcomes demonstrated that this framework outperformed the other models in terms of
accuracy, and also offered simple, understandable justifications for its forecasts.
The paper entitled “explainable machine learning for lung cancer screening models” by
Kobylińska et al. [20] is a comprehensive review of recent studies that use machine learning (ML) techniques
to develop lung cancer screening models. The authors emphasizes the significance of early lung cancer
detection and the possibility for ML models to enhance screening efficiency and accuracy. The survey
centers on research that makes use of explainable ML methods, which make it possible for clinicians to
comprehend how the models make their predictions. The author explains the various ML algorithms used in
the studies, such as deep learning, random forests, and logistic regression, as well as their advantages and
disadvantages. The authors also looks at different feature engineering and feature selection methods applied
to these works. The survey contains a thorough discussion of the difficulties involved in creating and testing
explainable machine learning models for lung cancer screening, including the lack of readily available data,
class imbalance, and interpretability problems. In order to overcome these difficulties and confirm the
efficacy of explainable ML models for lung cancer screening, the author emphasizes the need for additional
study in his concluding paragraph.
The paper entitled “an explainable AI-driven biomarker discovery framework for non-small cell
lung cancer classification” by Dwivedi et al. [21] presents a literature survey on biomarker discovery and
classification of non-small cell lung cancer (NSCLC) using XAI-driven frameworks. The survey highlights
the importance of biomarkers in early diagnosis and effective treatment of NSCLC. It also discusses various
machine learning algorithms and feature selection techniques used in biomarker discovery for NSCLC
classification. The survey further emphasizes the need for XAI-driven frameworks to ensure transparency,
interpretability, and reproducibility of the biomarker discovery process. The article concludes with the
author’s proposed framework for NSCLC classification using XAI-driven biomarker discovery. Different
applications of machine learning-AI techniques in medical field [22]-[30].
Table 1 shows the summary of predicting lung cancer risk using XAI. The current state of research
and practice in the field would be identified, and then areas, where there is a need for growth or development,
would be identified, as part of a gap analysis for predicting lung cancer risk using XAI. Here are some
measures that could be taken to carry out such an analysis:
− Identify current methods for predicting lung cancer risk: reviewing the existing research on machine
learning and conventional statistical models for predicting the risk of developing lung cancer would be
the first move [31], [32]. The key components or risk factors that are usually incorporated into these
models (such as age, smoking history, and family history) as well as any drawbacks or difficulties with
these methods would need to be identified.
− Evaluate the interpretability of existing models: being able to explain how a model generates its
predictions in a way that is comprehensible and transparent is a crucial component of XAI [33], [34]. The
interpretability of current lung cancer risk prediction algorithms should therefore be assessed. In order to
do this, it may be necessary to evaluate the models’ capacity to offer precise and understandable
justifications for how they arrived at a given risk score, as well as the consistency of these justifications
for various cases.
− Identify areas for improvement: the next stage would be to pinpoint areas that require development or
improvement based on the analysis of the current models [35], [36]. This could entail locating additional
risk variables that could be incorporated into the models, enhancing the readability of current models, or
creating brand-new machine learning techniques that are especially intended to be more understandable.
− Develop and test new models: the next step would be to create and test fresh models for predicting the
risk of developing lung cancer after potential areas for improvement have been found [37], [38]. This
might entail experimenting with new techniques for interpreting and visualising the findings,
incorporating new data sources, or using fresh machine learning algorithms.
− Evaluate the performance of new models: it is essential to evaluate the performance of the newly
developed machine learning models for predicting lung cancer risk by comparing them to existing models
[39], [40]. This comparison should consider various critical factors such as precision, interpretability,
algorithmic effectiveness, and usability [41]. By analyzing these aspects, we can determine the strengths
and weaknesses of each model, identify areas for improvement, and ensure that the new models are
reliable and effective tools for predicting an individual’s risk of developing lung cancer.
Overall, a gap analysis for lung cancer risk prediction using XAI would entail a thorough analysis of
the existing literature, the identification of areas for improvement, and the creation and testing of new
models. By thoroughly assessing current methodologies and establishing novel, more comprehensible
models, it could be feasible to enhance the accuracy and utility of lung cancer risk prediction, benefiting both
patients and healthcare practitioners.

Predicting lung cancer risk using explainable artificial intelligence (Shahin Shoukat Makubhai)
1280  ISSN: 2302-9285

Table 1. Summary of predicting lung cancer risk using XAI

Paper title Main objective Methods Dataset Key findings
DeepXplainer: an interpretable deep Develop a deep Deep learning, NLST Achieved 95.3%
learning based approach for lung explainable neural network SHAP values dataset accuracy and identified
cancer detection using explainable for predicting lung cancer key risk factors
artificial intelligence [42] risk
Explainable machine learning for lung Develop an XAI model to Rule-based NLST Achieved 87% accuracy
cancer screening models. Applied predict lung cancer risk reasoning, dataset and identified key risk
sciences [20] decision trees factors
An explainable machine learning Develop a machine Random NLST Achieved 85.4%
framework for lung cancer hospital learning model with forest, ReliefF dataset accuracy and identified
length of stay prediction [8] explainable feature algorithm key risk factors
selection for lung cancer
risk prediction
A review of explainable deep learning Develop a deep learning Deep learning, NLST Achieved 89.5%
cancer detection models in medical model with explainability LIME, SHAP dataset accuracy and identified
imaging [43] for predicting lung cancer values key risk factors
risk
Optimized deep learning model for Develop an interpretable Deep learning, NLST Achieved 89.9%
lung cancer prediction using ANN deep learning model for LRP method dataset accuracy and identified
algorithm [44] lung cancer prediction key risk factors
Interpretability approaches of Develop a hybrid model Decision tree, NLST Achieved 86.4%
explainable AI in analyzing features for lung cancer prediction fuzzy logic, dataset accuracy and identified
for lung cancer detection [45] using XAI IF-THEN key risk factors
rules
Cancer prediction using graph-based Develop an explainable Genetic NLST Achieved 88.2%
gene selection and explainable genetic algorithm for algorithm, dataset accuracy and identified
classifier [46] predicting lung cancer risk decision tree key risk factors
Artificial intelligence: deep learning Develop an XAI model for Support vector LIDC-IDRI Achieved 86.1%
in oncological radiomics and lung cancer prediction machine, dataset accuracy and identified
challenges of interpretability and data using radiomics features SHAP values key radiomics features
harmonization [47]
A hybrid ensemble feature selection- Develop a hybrid feature Random NLST Achieved 87.5%
based learning model for COPD selection method and forest, ReliefF dataset accuracy and identified
prediction on high-dimensional random forest algorithm algorithm, key risk factors
feature space [48] for lung cancer risk LASSO
prediction regression
End-to-end lung cancer screening Develop an XAI model for Deep learning, LUNA16 Achieved 95.5%
with three-dimensional deep learning lung cancer prediction SHAP values dataset accuracy and identified
on low-dose chest computed from low-dose CT scans key radiomics features
tomography [49]

2. METHOD
Predicting lung cancer risk using XAI can be achieved through the following system methodology:
− Data collection: gather information about lung cancer risk factors, including age, gender, smoking history,
exposure to the environment, family history, medical history, and other variables, from reputable sources.
The information must be varied, impartial, and population-representative.
− Data preprocessing: cleanse, normalise, and prepare the gathered data by converting it into a predictive
model-friendly format. Handling absent or inconsistent data is also crucial.
− Feature selection: determine which elements or factors from the preprocessed data are pertinent and
should be incorporated into the predictive model. Different methods, including statistical methods,
domain expertise, and machine learning algorithms, can be used for feature selection.
− Model development: developing an XAI model that can accurately predict an individual’s probability of
developing lung cancer based on selected features is crucial. The model should be transparent, easy to
comprehend, and able to provide explanations for its conclusions. Popular classification algorithms such
as support vector machines, decision trees, and logistic regression can be used to create this model. The
XAI model will take into account relevant features such as age, sex, smoking history, family history, and
exposure to toxins from the environment. By carefully selecting and curating these features, we can
ensure that the model is robust and effective in predicting an individual’s risk of developing lung cancer.
The model will be designed to provide clear and concise explanations for its conclusions, allowing
healthcare professionals and patients to understand the reasoning behind its predictions. By utilizing a
transparent and interpretable algorithm, we can enhance the trust and adoption of the model in the
medical field, ultimately leading to better healthcare outcomes for patients.

Bulletin of Electr Eng & Inf, Vol. 13, No. 2, April 2024: 1276-1285
Bulletin of Electr Eng & Inf ISSN: 2302-9285  1281

− Model evaluation: evaluate the model’s performance using suitable metrics such as accuracy, precision,
recall, F1-score, and confusion matrix. Additionally, model’s explainability can also be evaluated by
computing feature importance or creating visual explanations.
− Model interpretation: interpret the model to understand the factors that influence lung cancer risk
prediction. Explanations can be generated using techniques such as feature importance, local
explanations, and global explanations.
− Deployment: install the developed model in a setting that will allow for the actual application. Making
sure the deployed model is reliable, secure, and capable of successfully handling new data inputs is
crucial. Establishing ethical and legal criteria for the model’s implementation and use is also crucial.
Through the application of this methodology, it is conceivable to develop a transparent and easily
comprehensible AI model for gauging the probability of lung cancer onset. Such a model can provide
valuable support to healthcare practitioners in their decision-making processes, thereby alleviating the
challenges posed by this disease. Figure 1 shows the system architecture for predicting lung cancer risk using
XAI.

Figure 1. System architecture for predicting lung cancer risk using XAI

The steps for predicting lung cancer risk using XAI are as:
− Data collection: collect relevant data from various sources such as electronic health records, medical
imaging, and clinical trials.
− Data preprocessing: clean, normalize, perform feature engineering, and handle missing data in the
collected data.
− Feature selection: select the most important features that contribute to the prediction of lung cancer risk.
− Machine learning algorithm: build a machine learning model that can predict the risk of lung cancer using
the selected features.
− Explanation generation and visualization: use various XAI techniques to generate and visualize the
explanations for the model predictions.
− Explanation evaluation: evaluate the generated explanations for their usefulness, accuracy, and
comprehensibility.

3. RESULTS AND DISCUSSION

Using machine learning algorithms to analyse data on lung cancer risk factors and forecast the
likelihood of a person getting lung cancer, XAI can predict the risk of developing lung cancer [50], [51].
A subset of AI called XAI seeks to provide interpretable and transparent models so that people can
understand how a model makes predictions.
You would first need to collect information on various variables that may contribute to the
development of lung cancer in order to forecast lung cancer risk using XAI [52], [53]. These variables might
include demographic data like age, sex, and race as well as lifestyle variables like smoking status,
secondhand smoke exposure, and job hazards [54], [55]. Once the data has been gathered, you can evaluate it
and build a creating a predictive model utilizing machine learning algorithms such as logistic regression.

Predicting lung cancer risk using explainable artificial intelligence (Shahin Shoukat Makubhai)
1282  ISSN: 2302-9285

decision trees, and random forests. By learning from labelled examples of lung cancer cases and non-cancer
cases, these models can be trained using supervised learning methods.
Use XAI methods applying techniques such as feature importance analysis, partial dependence
plots, and Shapley values to render the model comprehensible and transparent [56], [57]. These methods
enable you to comprehend the connection between the input variables and the outcome prediction and assist
in pinpointing the factors that have the greatest influence on predicting lung cancer risk.
It’s essential to understand that XAI-based lung cancer risk prediction does not replace professional
medical advice or diagnosis [58]. However, it can be useful in identifying people who might have an
increased susceptibility to lung cancer development and who might profit from early detection and treatment.
Table 2 shows comparison of a dataset for predicting lung cancer risk using XAI.
What is the source of the dataset?
Was it collected from a study or from medical records?
How was the dataset collected?
Were participants recruited through a particular method or were they randomly selected?
Were the participants required to meet certain inclusion criteria?
What is the size of the dataset?
How many samples are included in the dataset?
What are the features in the dataset?
What are the demographic features of the participants?
What are the medical features used to predict lung cancer risk?
Are there any other features included in the dataset?
Are there any missing values in the dataset?
If so, how were they handled?
How was the data preprocessed?
Were any features transformed or scaled?
Were any outliers removed?
Were any features removed due to collinearity or redundancy?
What machine learning algorithm(s) were used to predict lung cancer risk?
What performance metrics were used to evaluate the models?
What features were important for predicting lung cancer risk?
How were feature importances calculated?
Were there any unexpected or counterintuitive features that were found to be important?
How was the model’s interpretability evaluated?
Were any explainability techniques used?
How well were the explainability results received by experts in the medical field?
What are the potential real-world implications of the model’s predictions?
How might the model be used to improve patient outcomes?
Are there any ethical concerns associated with using the model?
These are a series of questions related to a specific dataset used to predict lung cancer risk. The
questions cover topics such as the source and collection of the dataset, participant recruitment and inclusion
criteria, dataset size and features, missing value handling and preprocessing, machine learning algorithms
used, performance metrics and feature importance, interpretability and explainability techniques used,
potential real-world implications of the model’s predictions, and ethical considerations associated with using
the model.

Table 2. Comparison of dataset for predicting lung cancer risk using XAI
Dataset Number of Number of
Type of features Target variable Data source
name instances features
Dataset 1 500 20 Demographic, medical history, Lung cancer Hospital records
lifestyle diagnosis
Dataset 2 1000 15 Demographic, clinical tests, genetic Lung cancer Research study
diagnosis
Dataset 3 750 25 Demographic, medical history, Lung cancer Health survey
environmental exposure diagnosis
Dataset 4 2000 30 Demographic, clinical tests, Lung cancer Electronic health
lifestyle diagnosis records
Dataset 5 1500 18 Demographic, medical history, Lung cancer National cancer
smoking history diagnosis registry

Bulletin of Electr Eng & Inf, Vol. 13, No. 2, April 2024: 1276-1285
Bulletin of Electr Eng & Inf ISSN: 2302-9285  1283

4. CONCLUSION
In conclusion, XAI-based risk prediction for lung cancer, has the potential to significantly improve
early detection and prevention of the disease. By analyzing patient data such as age, gender, smoking history,
and other health factors, machine learning algorithms can identify patterns and risk factors that may be
difficult for humans to detect. XAI is particularly important in healthcare, as it can provide transparent
interpretation for its predictions, which can help patients and clinicians understand the reasoning behind the
algorithm’s suggestions. This could lead to increased trust in the technology, and better health outcomes.
Nevertheless, it is crucial to emphasize that AI must not serve as a substitute for human judgment
and expertise. Instead, it should function as an instrument to enhance and complement clinical decision-
making. Furthermore, it is crucial that AI is developed and deployed in an ethical and responsible manner,
taking into account issues such as data privacy, bias, and transparency. Hence, when employing interpretable
AI for forecasting lung cancer risk, it holds the promise of being a valuable asset in the battle against lung
cancer. Its application should be in tandem with alternative diagnostic and preventive methods, placing a
significant emphasis on ethical and responsible implementation.

REFERENCES
[1] V. K. Raghu et al., “Feasibility of lung cancer prediction from low-dose CT scan and smoking factors using causal models,”
Thorax, vol. 74, no. 7, pp. 643–649, 2019, doi: 10.1136/thoraxjnl-2018-212638.
[2] S. Doppalapudi, R. G. Qiu, and Y. Badr, “Lung cancer survival period prediction and understanding: Deep learning approaches,”
International Journal of Medical Informatics, vol. 148, no. September 2020, p. 104371, 2021, doi:
10.1016/[Link].2020.104371.
[3] M. C. Tammemägi, “Selecting lung cancer screenees using risk prediction models- where do we go from here,” Translational
lung cancer research, vol. 7, no. 3, pp. 243–253, 2018, doi: 10.21037/tlcr.2018.06.03.
[4] M. Avanzo, J. Stancanello, G. Pirrone, and G. Sartor, “Radiomics and deep learning in lung cancer,” Strahlentherapie und
Onkol., vol. 196, no. 10, pp. 879–887, 2020, doi: 10.1007/s00066-020-01625-9.
[5] Y. Li, D. Gu, Z. Wen, F. Jiang, and S. Liu, “Classify and explain: An interpretable convolutional neural network for lung cancer
diagnosis,” ICASSP, CASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP),
pp. 1065–1069, 2020, doi: 10.1109/ICASSP40776.2020.9054605.
[6] Y. Lei, Y. Tian, H. Shan, J. Zhang, G. Wang, and M. K. Kalra, “Shape and margin-aware lung nodule classification in low-dose
CT images via soft activation mapping,” Medical Image Analysis, vol. 60, p. 101628, 2020, doi: 10.1016/[Link].2019.101628.
[7] B. Ramos, T. Pereira, J. Moranguinho, J. Morgado, J. L. Costa, and H. P. Oliveira, “An interpretable approach for lung cancer
prediction and subtype classification using gene expression,” in 2021 43rd Annual International Conference of the IEEE
Engineering in Medicine & Biology Society (EMBC), pp. 1707–1710, 2021, doi: 10.1109/EMBC46164.2021.9630775.
[8] B. Alsinglawi et al., “An explainable machine learning framework for lung cancer hospital length of stay prediction,” Scientific
reports, vol. 12, no. 1, pp. 607, 2022, doi: 10.1038/s41598-021-04608-7.
[9] F. Silva et al., “Pre-training autoencoder for lung nodule malignancy assessment using CT images,” applied sciences, vol. 10, no.
21, pp. 1–13, 2020, doi: 10.3390/app10217837.
[10] C. Venkatesh and P. Bojja, “Development of qualitative model for detection of lung cancer using optimization,” International
Journal of Innovative Technology and Exploring Engineering, vol. 8, no. 9, pp. 3143–3147, 2019, doi:
10.35940/ijitee.i8619.078919.
[11] L. Bertolaccini, P. Solli, A. Pardolesi, and A. Pasini, “An overview of the use of artificial neural networks in lung cancer
research,” Journal of thoracic disease, vol. 9, no. 4, pp. 924–931, 2017, doi: 10.21037/jtd.2017.03.157.
[12] F. Silva et al., “Towards Machine Learning-Aided Lung Cancer Clinical Routines: Approaches and Open Challenges,” Journal of
Personalized Medicine, vol. 12, no. 3, pp. 1–36, 2022, doi: 10.3390/jpm12030480.
[13] Y. Xie et al., “Early lung cancer diagnostic biomarker discovery by machine learning methods,” Translational oncology, vol. 14,
no. 1, 2021, doi: 10.1016/[Link].2020.100907.
[14] L. M. Seijo et al., “Biomarkers in lung cancer screening: achievements, promises, and challenges,” Journal of Thoracic Oncology,
vol. 14, no. 3, pp. 343–357, 2019, doi: 10.1016/[Link].2018.11.023.
[15] M. A. Heuvelmans et al., “Lung cancer prediction by deep learning to identify benign lung nodules,” Lung Cancer, vol. 154, no.
2020, pp. 1-4, 2021, doi: 10.1016/[Link].2021.01.027.
[16] W. L. Bi et al., “Artificial intelligence in cancer imaging: Clinical challenges and applications,” CA: a cancer journal for
clinicians, vol. 69, no. 2, pp. 127–157, 2019, doi: 10.3322/caac.21552.
[17] E. Ntoutsi et al., “Bias in data-driven artificial intelligence systems—An introductory survey,” Wiley Interdisciplinary Reviews:
Data Mining and Knowledge Discovery, vol. 10, no. 3, pp. 1–14, 2020, doi: 10.1002/widm.1356.
[18] M. S. A. Dwivedi, M. R. P. Borse, and M. A. M. Yametkar, “Lung cancer detection and classification by using machine learning
& multinomial bayesian,” IOSR Journal of Electronics and Communication Engineering (IOSR-JECE), vol. 9, no. 1, pp. 69–75,
2014, doi: 10.9790/2834-09136975.
[19] C. Venkatesh, K. Ramana, S. Y. Lakkisetty, S. S. Band, S. Agarwal, and A. Mosavi, “A neural network and optimization based
lung cancer detection system in CT images,” Frontiers in Public Health, vol. 10, no. June, pp. 1–9, 2022, doi:
10.3389/fpubh.2022.769692.
[20] K. Kobylińska, T. Orłowski, M. Adamek, and P. Biecek, “Explainable machine learning for lung cancer screening models,”
Applied Sciences, vol. 12, no. 4, 2022, doi: 10.3390/app12041926.
[21] K. Dwivedi, A. Rajpal, S. Rajpal, M. Agarwal, V. Kumar, and N. Kumar, “An explainable ai-driven biomarker discovery
framework for non-small cell lung cancer classification,” Computers in Biology and Medicine, vol. 153, no. August 2022, 2023,
doi: 10.1016/[Link].2023.106544.
[22] I. Tunali, R. J. Gillies, and M. B. Schabath, “Application of radiomics and artificial intelligence for lung cancer precision
medicine,” Cold Spring Harbor perspectives in medicine, vol. 11, no. 8, 2021, doi: 10.1101/cshperspect.a039537.
[23] M. Marcos et al., “Artificial intelligence in medicine : knowledge representation and transparent and explainable systems,”
Heidelberg: Springer, 2019.

Predicting lung cancer risk using explainable artificial intelligence (Shahin Shoukat Makubhai)
1284  ISSN: 2302-9285

[24] E. Dritsas and M. Trigka, “Lung cancer risk prediction with machine learning models,” Big Data and Cognitive Computing, vol.
6, no. 4, 2022, doi: 10.3390/bdcc6040139.
[25] C. Rampinelli et al., “Exposure to low dose computed tomography for lung cancer screening and risk of cancer: Secondary
analysis of trial data and risk-benefit analysis,” BMJ, vol. 356, pp. 1–6, 2017, doi: 10.1136/bmj.j347.
[26] T. Kadir and F. Gleeson, “Lung cancer prediction using machine learning and advanced imaging techniques,” Translational lung
cancer research, vol. 7, no. 3, pp. 304–312, 2018, doi: 10.21037/tlcr.2018.05.15.
[27] S. Mall, A. Srivastava, B. D. Mazumdar, M. Mishra, S. L. Bangare, and A. Deepak, “Implementation of machine learning
techniques for disease diagnosis,” Materials Today: Proceedings, vol. 51, no. 8, 2022, p. 2198-2201, ISSN 2214-7853, doi:
10.1016/[Link].2021.11.274.
[28] X. Wu, D. Wei, B. P. Vasgi, A. K. Oleiwi, S. L. Bangare, and E. Asenso, “Research on network security situational awareness
based on crawler algorithm,” Security and Communication Networks, vol. 2022, 2022, doi: 10.1155/2022/3639174.
[29] N. Shelke, S. Chaudhury, S. Chakrabarti, S. L. Bangare, G. Yogapriya, and P. Pandey. “An efficient way of text-based emotion
analysis from social media using LRA-DNN,” Neuroscience Informatics, vol. 2, no. 3, 2022, doi: 10.1016/[Link].2022.100048.
[30] S. L. Bangare, G. Pradeepini and S. T. Patil, “Brain tumor classification using mixed method approach,” 2017 International
Conference on Information Communication and Embedded Systems, 2017, pp. 1-4, doi: 10.1109/ICICES.2017.8070748.
[31] S. L. Bangare, G. Pradeepini, and S. T. Patil, “Implementation for brain tumor detection and three dimensional visualization
model development for reconstruction,” ARPN Journal of Engineering and Applied Sciences (ARPN JEAS), vol. 13, no. 2, pp.
467-473, 2018.
[32] S. L. Bangare, “Classification of optimal brain tissue using dynamic region growing and fuzzy min-max neural network in brain
magnetic resonance images,” Neuroscience Informatics, vo. 2, no. 3, September 2022, doi: 10.1016/[Link].2021.100019.
[33] S. L. Bangare, D. Virmani, G. R. Karetla, P. Chaudhary, H. Kaur, S. N. H. Bukhari, and S. Miah, “Forecasting the applied deep
learning tools in enhancing food quality for heart related diseases effectively: a study using structural equation model analysis,”
Journal of Food Quality, vol. 2022, pp. 1-8, 2022, doi: 10.1155/2022/6987569.
[34] V. D. P. Jasti et al., “Image processing and machine learning-based classification and detection of liver tumor,” BioMed Research
International, vol. 2022, pp. 1-7, 2022, doi: 10.1155/2022/3398156.
[35] A. S. Ladkat et al., “Deep Neural Network-Based Novel Mathematical Model for 3D Brain Tumor Segmentation,” Computational
Intelligence and Neuroscience, vol. 2022, pp. 1-8, 2022, doi: 10.1155/2022/4271711.
[36] S. Sari, T. Sutikno, I. Soesanti, and N. A. Setiawan, “A review of convolutional neural network-based computer-aided lung
nodule detection system,” IAES International Journal of Artificial Intelligence (IJ-AI), vol. 12, no. 3, pp. 1044–1061, 2023, doi:
10.11591/ijai.v12.i3.pp1044-1061.
[37] G. Abdikerimova et al., “Detection of chest pathologies using autocorrelation functions,” International Journal of Electrical and
Computer Engineering (IJECE), vol. 13, no. 4, pp. 4526–4534, 2023, doi: 10.11591/ijece.v13i4.pp4526-4534.
[38] R. Mothkur and V. B. Nagendrappa, “An optimal model for classification of lung cancer using grey wolf optimizer and deep
hybrid learning,” Indonesian Journal of Electrical Engineering and Computer Science (IJEECS), vol. 30, no. 1, pp. 406–413,
2023, doi: 10.11591/ijeecs.v30.i1.pp406-413.
[39] Y. Tashtoush, R. Obeidat, A. Al-Shorman, O. Darwish, M. Al-Ramahi, and D. Darweesh, “Enhanced convolutional neural
network for non-small cell lung cancer classification,” International Journal of Electrical and Computer Engineering (IJECE),
vol. 13, no. 1, pp. 1024–1038, 2023, doi: 10.11591/ijece.v13i1.pp1024-1038.
[40] R. S. Shankar, R. S. Chigurupati, P. Voosala, and N. Pilli, “An extensible framework for recurrent breast cancer prognosis using
deep learning techniques,” Indonesian Journal of Electrical Engineering and Computer Science (IJEECS), vol. 29, no. 2, pp.
931–941, 2023, doi: 10.11591/ijeecs.v29.i2.pp931-941.
[41] A. A. Abd Al-Ameer, G. A. Hussien, and H. A. Al Ameri, “Lung cancer detection using image processing and deep learning,”
Indonesian Journal of Electrical Engineering and Computer Science (IJEECS), vol. 28, no. 2, pp. 987–993, 2022, doi:
10.11591/ijeecs.v28.i2.pp987-993.
[42] N. A. Wani, R. Kumar, and J. Bedi, “DeepXplainer: An interpretable deep learning based approach for lung cancer detection
using explainable artificial intelligence,” Computer Methods and Programs in Biomedicine, vol. 243, 2023, doi:
10.1016/[Link].2023.107879.
[43] M. A. Gulum, C. M. Trombley, and M. Kantardzic, “A Review of Explainable Deep Learning Cancer Detection Models in
Medical Imaging,” Applied Sciences, vol. 11, no. 10, pp. 1-21, 2020, doi: 10.3390/app11104573.
[44] D. Rawat, Meenakshi, L. Pawar, G. Bathla and R. Kant, "Optimized Deep Learning Model for Lung Cancer Prediction Using
ANN Algorithm," 2022 3rd International Conference on Electronics and Sustainable Communication Systems (ICESC),
Coimbatore, India, 2022, pp. 889-894, doi: 10.1109/ICESC54411.2022.9885607.
[45] M. Pal, S. Mistry, and D. De, “Interpretability Approaches of Explainable AI in Analyzing Features for Lung Cancer Detection,”
Frontiers of ICT in Healthcare Proceedings of EAIT 2022, Singapore: Springer Nature Singapore, 2023, vol. 519, pp. 277–287,
doi: 10.1007/978-981-19-5191-6_23.
[46] M. Rostami, and M. Oussalah, “Cancer prediction using graph-based gene selection and explainable classifier,” Finnish Journal
of EHealth and EWelfare, vol. 14, no. 1, pp. 61–78, 2022, doi: 10.23996/fjhw.111772.
[47] P. Papadimitroulas et al., “Artificial intelligence: Deep learning in oncological radiomics and challenges of interpretability and
data harmonization,” Physica Medica, vol. 83, pp. 108-121, 2021, doi: 10.1016/[Link].2021.03.009
[48] S. R. B. Banda, and T. R. Babu, “A Hybrid Ensemble Feature Selection-Based Learning Model for COPD Prediction on High-
Dimensional Feature Space,” Data Engineering and Communication Technology Proceedings of 3rd ICDECT-2K19, Singapore:
Springer Singapore, 2020, pp. 663–675, doi: 10.1007/978-981-15-1097-7_55.
[49] D. Ardila et al., “End-to-end lung cancer screening with three-dimensional deep learning on low-dose chest computed
tomography,” Nature Medicine, vol. 25, no. 6, pp. 954-961, 2019, doi: 10.1038/s41591-019-0447-x
[50] M. F. Abdullah, S. N. Sulaiman, M. K. Osman, N. K. A. Karim, S. Setumin, and I. S. Isa, “A new procedure for lung region
segmentation from computed tomography images,” International Journal of Electrical and Computer Engineering (IJECE), vol.
12, no. 5, pp. 4978–4987, 2022, doi: 10.11591/ijece.v12i5.pp4978-4987.
[51] O. R. Kadhim, H. J. Motlak, and K. K. Abdalla, “Computer-aided diagnostic system kinds and pulmonary nodule detection
efficacy,” International Journal of Electrical and Computer Engineering (IJECE), vol. 12, no. 5, pp. 4734–4745, 2022, doi:
10.11591/ijece.v12i5.pp4734-4745.
[52] S. Kothari, S. Chiwhane, S. Jain, and M. Baghel, “Cancerous brain tumor detection using hybrid deep learning framework,”
Indonesian Journal of Electrical Engineering and Computer Science (IJEECS), vol. 26, no. 3, pp. 1651–1661, 2022, doi:
10.11591/ijeecs.v26.i3.pp1651-1661.
[53] O. Erefai, A. Soulaymani, A. Mokhtari, and H. Hami, “Never smokers with lung cancer in Morocco: epidemiology and gender

Bulletin of Electr Eng & Inf, Vol. 13, No. 2, April 2024: 1276-1285
Bulletin of Electr Eng & Inf ISSN: 2302-9285  1285

differences,” International Journal of Public Health Science (IJPHS), vol. 11, no. 2, pp. 417–422, 2022, doi:
10.11591/ijphs.v11i2.21306.
[54] R. D. Abdu-Aljabar and O. A. Awad, “Parallel extreme gradient boosting classifier for lung cancer detection,” Indonesian Journal
of Electrical Engineering and Computer Science (IJEECS), vol. 24, no. 3, pp. 1610–1617, 2021, doi:
10.11591/ijeecs.v24.i3.pp1610-1617.
[55] Z. Rustam, A. Purwanto, S. Hartini, and G. S. Saragih, “Lung cancer classification using fuzzy C-means and fuzzy Kernel C-
means based on CT scan image,” IAES International Journal of Artificial Intelligence (IJ-AI), vol. 10, no. 2, pp. 291–297, 2021,
doi: 10.11591/ijai.v10.i2.pp291-297.
[56] P. Chandre, P. Mahalle, and G. Shinde, “Intrusion prevention system using convolutional neural network for wireless sensor
network,” IAES International Journal of Artificial Intelligence (IJ-AI), vol. 11, no. 2, pp. 504-515, 2022, doi:
10.11591/ijai.v11.i2.pp504-515.
[57] H. F. Kareem, M. S. AL-Husieny, F. Y. Mohsen, E. A. Khalil, and Z. S. Hassan, “Evaluation of SVM performance in the
detection of lung cancer in marked CT scan dataset,” Indonesian Journal of Electrical Engineering and Computer Science
(IJEECS), vol. 21, no. 3, pp. 1731–1738, 2021, doi: 10.11591/ijeecs.v21.i3.pp1731-1738.
[58] S. Makubhai, G. R. Pathak, and P. R. Chandre, “Prevention in Healthcare: An Explainable AI Approach,” International Journal
on Recent and Innovation Trends in Computing and Communication, vol. 11, no. 5, pp. 92–100. 2023, doi:
10.17762/ijritcc.v11i5.6582.

BIOGRAPHIES OF AUTHORS

Shahin Shoukat Makubhai is an Assistant Professor in the Department of

Computer Science and Engineering at MIT School of Computing, MIT ADT, and Pune, India.
She received her B.E. degree in Computer Science and Engineering from DKTE Society’s
Textile and Engineering Institute (an Autonomous Institute), Ichalkaranji, India, and her
[Link]. degree in Computer Engineering with specialization in cloud computing from Vellore
Institute of Technology - VIT Chennai, India in 2020. She is currently pursuing her Ph.D in
Computer Science at MIT-ADT University, Pune, India. She has earned several global
certifications and has also contributed to research through patents and copyrights. Her research
interests include artificial intelligence and cloud computing. She can be contacted at:
[Link]@[Link].

Ganesh R. Pathak received Bachelor of Engineering (B.E. - Computer) from

Walchand Institute of Technology, Maharashtra, India, Master of Engineering (M.E. - CSEIT)
from Savitribai Phule Pune University (formerly University of Pune), Maharashtra, India and
PhD degree in Computer Science and Engineering at Sathyabama Institute of Science and
Technology (Deemed to be University), Chennai, Tamil Nadu, India. He is presently working
as Professor in the Department of Computer Science and Engineering, School of Computing,
MIT Art, Design and Technology University, Pune. His research interests include artificial
intelligence, machine learning, data science, computer networks, wireless communication,
wireless sensor network, especially security in wireless sensor network. His teaching areas
include mobile computing, pervasive computing information assurance and security and
usability engineering. He is a member of IEEE and CSI. He can be contacted at:
[Link]@[Link].

Pankaj R. Chandre has obtained his B.E degree in Information Technology from
Sant Gadge Baba Amravati University, Amravati, India, M.E. degree in Computer Engineering
from from Mumba1 University Maharashtra, India in the year 2011 and Ph.D in Computer
Engineering from Savitribai Phule Pune University, Pune, India in the year 2021. He is currently
working as an Associate Professor in Department of Computer Science and Engineering, MIT
School of Computing, MIT ADT, Pune, India. He has published 60 plus papers at international
journals and conferences. He has guided more than 30 plus under-graduate students and 20 plus
postgraduate students for projects. His research interests are network security and information
security. He can be contacted at: [Link]@[Link].

Predicting lung cancer risk using explainable artificial intelligence (Shahin Shoukat Makubhai)