A Deep Neuro-Fuzzy Systems for Effective and Interpretable Medical Decision-Making

Abstract

Accurate and interpretable models play a vital role in supporting clinical decision-making for cancer diagnosis. While conventional deep learning models, such as convolutional neural networks (CNNs), exhibit strong classification performance, their lack of transparency limits their applicability in healthcare. To overcome this challenge, this study proposes the Attentive Convolutional Neuro-Fuzzy Network (AConvNFC), a deep neuro-fuzzy system that integrates fuzzy inference with an attention mechanism to dynamically focus on the most significant tumor features. Utilizing a convolutional framework, the model optimizes feature selection, reduces the complexity of fuzzy rules, and enhances interpretability. By evaluating the Colorectal surgery (CRC) dataset, the model demonstrates exceptional performance in distinguishing benign from malignant tumours while offering explicit and actionable insights into the reasoning behind its predictions. This work underscores the potential of AConvNFC as a robust and interpretable solution for medical decision-making tasks.