• Home
  • Search
  • Interpretable Deep Learning for Pancreatic Cancer Diagnosis Using TabNet and SHapley Additive Explanations
  • https://doi.org/10.1109/icecst66106.2025.11307265Copy DOI Icon

Interpretable Deep Learning for Pancreatic Cancer Diagnosis Using TabNet and SHapley Additive Explanations

  • Oct 23, 2025
  • G Prasad Babu +5 more
Show More
  • Abstract
  • Literature Map
  • References
  • Similar Papers
Abstract

Pancreatic cancer is still one of the deadliest types of cancer because it doesn't show any symptoms until it's too late for treatment to work. Fewer than <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{1 0 \%}$</tex> of people who are diagnosed with it live for five years. Conventional diagnostic methods often demonstrate insufficient sensitivity and interpretability, consequently reducing their effectiveness in clinical decision-making. This study introduces an interpretable deep learning framework employing TabNet for the early classification of pancreatic cancer using tabular clinical and biochemical datasets, thereby overcoming current limitations. For the best performance of the model, the method includes a lot of preprocessing, feature normalization, stratified data partitioning, and hyperparameter tuning. To address the opacity of conventional deep learning models, SHapley Additive exPlanations (SHAP) are employed to evaluate the contribution of each biological marker to predictive accuracy. The experimental results demonstrate that the proposed model shows better accuracy than basic machine learning methods like Random Forest, Support Vector Machine, and Gradient Boosting. It has a balanced precision, recall, and ROC-AUC score, and its accuracy is about 95 %. SHAP-based interpretability shows that CA19-9, creatinine levels, and liver function tests are all important risk factors. This makes the tests more reliable and makes the diagnosis clearer. The findings indicate that the integration of explainable artificial intelligence with deep learning frameworks yields a reliable, accurate, and transparent decision support system for pancreatic cancer screening. This study lays the foundation for the deployment of clinically aligned AI tools in real-world healthcare environments, with future developments focused on the multimodal integration of imaging and genomic biomarkers to enhance predictive accuracy and customize treatment planning.

Similar Papers
  • Research Article
  • Citations2

Improving Machine Learning Classification Predictions through SHAP and Features Analysis Interpretation.

  • Oct 20, 2025
  • Journal of chemical information and modeling
  • Leonardo Bernal +2
  • Research Article

Establishment and evaluation of a machine learning prediction model for sepsis-related encephalopathy in the elderly

  • Oct 01, 2025
  • Zhonghua wei zhong bing ji jiu yi xue
  • Xiao Yue +8
  • Research Article
  • Citations11

Prediction of gully erosion susceptibility through the lens of the SHapley Additive exPlanations (SHAP) method using a stacking ensemble model.

  • May 01, 2025
  • Journal of environmental management
  • Jeongho Han +2
  • Research Article

Explainable AI Analysis of Cognitive Distortions and Their Predictive Role in Adolescent Major Depressive Episodes

  • Jan 01, 2026
  • Journal of Adolescent and Youth Psychological Studies
  • Lina Khoury +2
  • Research Article
  • Citations2

The future of critical care: AI-powered mortality prediction for acute variceal gastrointestinal bleeding and acute non-variceal gastrointestinal bleeding patients

  • May 16, 2025
  • Frontiers in Medicine
  • Zhou Liu +8
  • PDF
  • Research Article
  • Citations15

Using Artificial Intelligence Approach for Investigating and Predicting Yield Stress of Cemented Paste Backfill

  • Feb 06, 2023
  • Sustainability
  • Van Quan Tran
  • Research Article
  • Citations6

Application of machine learning model in predicting the likelihood of blood transfusion after hip fracture surgery.

  • Sep 21, 2023
  • Aging Clinical and Experimental Research
  • Xiao Chen +3
  • Research Article

An Interpretable Machine Learning Model Based on MRI Features for Predicting Pain Severity in Temporomandibular Disorders.

  • Nov 18, 2025
  • Journal of oral rehabilitation
  • Chuanfang Xu +6
  • Research Article

Factors affecting and identification of key environmental determinants of the Oncomelania hupensis snail density in the Yangtze River Delta based on machine learning models

  • Mar 04, 2026
  • Zhongguo xue xi chong bing fang zhi za zhi = Chinese journal of schistosomiasis control
  • Y Li +6
  • Research Article
  • Citations2

Predicting Fiber Length Characteristics of Recycled Cotton and Cellulose Fiber Blends Using Machine Learning Models

  • May 12, 2025
  • Advanced Theory and Simulations
  • Tuser Tirtha Biswas +2
  • Research Article
  • Citations5

Machine learning analysis of survival outcomes in breast cancer patients treated with chemotherapy, hormone therapy, surgery, and radiotherapy

  • Jul 10, 2025
  • Scientific Reports
  • Eyachew Misganew Tegaw +1
  • Research Article
  • Citations7

Development and validation of a hypoxemia prediction model in middle-aged and elderly outpatients undergoing painless gastroscopy

  • May 23, 2025
  • Scientific Reports
  • Leilei Zheng +8
  • Research Article
  • Citations15

Machine Learning And Artificial Intelligence in Diabetes Prediction And Management: A Comprehensive Review of Models

  • Dec 17, 2024
  • Innovatech Engineering Journal
  • Md Ashraful Alam +3
  • Research Article

Machine Learning-Based Prediction of Institutional Delivery Dropout (IDD) Among Nigerian Women: An Exploratory Study Using SHAP Interpretability.

  • Feb 23, 2026
  • Journal of epidemiology and global health
  • Jamilu Sani +2
  • Research Article

Transforming Breast Cancer Prediction: Advanced Machine Learning Models for Accurate Prediction and Personalized Care

  • Sep 26, 2025
  • International Journal of Statistics in Medical Research
  • Usha Adiga +5
Cactus Communications logo

Copyright 2026 Cactus Communications. All rights reserved.