Explainable Hybrid Deep Learning for Automated Diagnosis of Canine Mammary Tumors

Automatic classification of canine mammary tumors (CMT) plays a vital role in ensuring accurate diagnosis, reliable predictive evaluation, and reducing manual intervention in in the diagnostic process.This paper introduces an Explainable Artificial Intelligence (XAI) system for automated CMT classification, which combines the DenseNet201 deep learning (DL) architecture for feature extraction with a Random Forest (RF) classifier to distinguish between benign and malignant tumors in CMT images.The proposed system follows a structured four-phase pipeline: (1) data acquisition and Laplacian filter preprocessing to enhance tumor edges; (2) feature extraction using DenseNet201 and classification with a Random Forest (RF) model; (3) testing and evaluation of the proposed system; and (4) interpretability analysis using the Shapley Additive exPlanations (SHAP) XAI technique.Experimental results demonstrate that the DenseNet201-RF hybrid model achieves an accuracy of 93.9%, surpassing benchmark classifiers while offering interpretability for clinical validation.The proposed system enables accurate automatic classification of canine mammary tumors while integrating SHAP-based XAI for interpretability.