Development of an ensemble CNN model with explainable AI for the classification of gastrointestinal cancer

Sep 21, 2026

Uploaded by Violet

AI-assisted cancer detection in clinical settings faces hurdles driven by limited explainability of model mechanisms, which makes medical practitioners wary of adopting opaque diagnostic decisions. This research investigates Shapley additive explanations (SHAP) and local interpretable model-agnostic explanations (LIME) to explain predictions using an ensemble of three CNNs: InceptionV3, InceptionResNetV2, and VGG16. Trained on the Kvasir dataset of gastrointestinal pathology images, the model achieves 96.89% accuracy and 96.877% F1-scores. Post-training, SHAP and LIME identify deterministic image features influencing class predictions, supporting XAI progress for gastrointestinal cancer detection.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages23

Reading Time58 min

AccessFree to download

Abstract

1 Introduction

Gastrointestinal cancer background and impact

Risk factors and sex/age differences

Prognosis and importance of early detection

Methods and model overview

Ensemble architecture and training dataset

Explainability with SHAP and LIME

Results and discussion

Accuracy and F1 performance

Interpretable feature analysis

Conclusion

PLOS ONE OPEN ACCESS Citation: Auzine MM, Heenaye-Mamode Khan M, Baichoo S, Gooda Sahib N, Bissoonauth-Daiboo P, Gao X, et al. (2024) Development of an ensemble CNN model with explainable AI for the classification of gastrointestinal cancer. PLoS ONE 19(6): e0305628 . [https://doi.org/10.1371/journal](https://doi.org/10.1371/journal). pone.0305628 Editor: Ali Mohammad Alqudah, University of Manitoba, CANADA Received: January 24, 2024 Accepted: June 2, 2024 Published: June 25, 2024 Copyright: © 2024 Auzine et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: All files used in this study are publicly available from the ‘KVASIR: A Multi-Class Image Dataset for Computer Aided Gastrointestinal Disease Detection’ database (DOI: [https://doi.org/10.1145/3193289/](https://doi.org/10.1145/3193289/)) . Funding: • M. HMK • TB-2021-01 • Higher Education Commission (HEC)• [https://www.hec](https://www.hec). mu/ • NO The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. RESEARCH ARTICLE Development of an ensemble CNN model with explainable AI for the classification of gastrointestinal cancer Muhammad MuzzammilAuzine1 *, Maleika Heenaye-Mamode Khan1, Sunilduth Baichoo1, Nuzhah Gooda Sahib1, Preeti Bissoonauth-Daiboo1, Xiaohong Gao2, Zaid Heetun3 1 Department of Software and Information Systems, University of Mauritius, Reduit, Mauritius, 2 Department of Computer Science, Middlesex University London, London, United Kingdom, 3 Center for Gastroenterology and Hepatology, Dr Abdool Gaffoor Jeetoo Hospital, Port Louis, Mauritius * [mmuzzammil.auzine@gmail.com](mmuzzammil.auzine@gmail.com) Abstract The implementation of AI assisted cancer detection systems in clinical environments has faced numerous hurdles, mainly because of the restricted explainability of their elemental mechanisms, even though such detection systems have proven to be highly effective. Medical practitioners are skeptical about adopting AI assisted diagnoses as due to the latter’sinability to be transparent about decision making processes. In this respect, explainable artificial intelligence (XAI) has emerged to provide explanations for model predictions, thereby overcoming the computational black box problem associated with AI systems. In this particular research, the focal point has been the exploration of the Shapley additive explanations (SHAP) and local interpretable model-agnostic explanations (LIME) approaches which enable model prediction explanations. This study used an ensemble model consisting of three convolutional neural networks(CNN): InceptionV3, InceptionResNetV2 and VGG16, which was based on averaging techniques and by combining their respective predictions. These models were trained on the Kvasir dataset, which consists of pathological findings related to gastrointestinal cancer. An accuracy of 96 .89% and F1-scores of 96 .877% were attained by our ensemble model. Following the training of the ensemble model, we employed SHAP and LIME to analyze images from the three classes, aiming to provide explanations regarding the deterministic features influencing the model’s predictions. The results obtained from this analysis demonstrated a positive and encouraging advancement in the exploration of XAI approaches , specifically in the context of gastrointestinal cancer detection within the healthcare domain. 1 Introduction The gastrointestinal tract comprises of organs that form the digestive system. Mutation of cells lining at least one of these organs induces the production of tumors which eventually leads to the development of gastrointestinal cancer. It is noteworthy that gastrointestinal (GI) cancers Competing interests: The authors ha

FAQs

Why is explainable AI important for gastrointestinal cancer diagnosis in this study?

Medical practitioners are cautious because many AI systems cannot transparently explain how decisions are made. Explainable AI provides explanations for model predictions and addresses the black-box concern.

What ensemble model is used for classification?

The approach combines three convolutional neural networks: InceptionV3, InceptionResNetV2, and VGG16. Their predictions are combined using averaging techniques.

How do SHAP and LIME contribute to the analysis after training?

After training the ensemble model, SHAP and LIME are applied to images from the three classes to explain which image features influence the model’s predictions.

Sep 21, 2026

Uploaded by Violet