Innovative Prostate Cancer Classification: Merging Autoencoders, PCA, SHAP, and Machine Learning Techniques

Gavade, Anil and Nerli, Rajendra and Gavade, Priyanka and Kumar, Meshach and Mehta, Utkal V. (2025) Innovative Prostate Cancer Classification: Merging Autoencoders, PCA, SHAP, and Machine Learning Techniques. In: Proceedings of the First International Conference on Advanced Robotics, Control, and Artificial Intelligence. Lecture Notes in Networks and Systems, 1376 . Springer Nature, Singapore, pp. 375-390. ISBN 978-981-96-5372-0

Text - Published Version Restricted to Registered users only Download (3MB)

Abstract

This paper introduces an innovative hybrid method for the classification of Gleason Scores (GS) in prostate cancer using Whole Slide Images. The researchers combine Deep Learning and Machine Learning (ML) techniques to automate the precise classification of GS. They employ a specially designed Variational Autoencoder for feature extraction, utilizing a pre-trained VGG16 Convolutional Neural Network to build the encoder. Principal Component Analysis is then used to reduce the dimensionality of the feature vector to 50 significant features for further Gleason Grade classification. The study uses the SICAPv2 database and evaluates feature importance with Shapley Additive explanations (SHAP). Comparative analysis of five ML techniques and two custom-designed Deep Neural Network (DNN) architectures shows that the Support Vector Machine algorithm with hyperparameter tuning and the custom-designed five-layer DNN architecture achieved accuracies of 84% and 89%, respectively, outperforming other ML models.

Item Type:	Book Chapter
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering > Robotics and Automation
Divisions:	School of Information Technology, Engineering, Mathematics and Physics (STEMP)
Depositing User:	Utkal Mehta
Date Deposited:	12 Jun 2025 00:24
Last Modified:	12 Jun 2025 00:33
URI:	https://repository.usp.ac.fj/id/eprint/14994

Actions (login required)

View Item