Explainable AI based efficient ensemble model for breast cancer classification using optical coherence tomography

In this paper, we propose a novel framework for breast cancer classification using optical coherence tomography (OCT) images. As the incidence of breast cancer in India continues to rise, early detection is essential for increasing survival rates. For precise classification, an efficient classifier model is required. Due to the variety of datasets and performance metrics, a single classifier may not be sufficient. Thus, we propose an efficient ensemble classifier based on Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) integrated with the Crow Search Algorithm (CSA). TOPSIS, a technique for Multi-Criteria Decision Making (MCDM), ranks and selects the ensemble's base classifiers. The CSA fine-tunes the weights of the base classifiers. In addition, we employ SHAP (SHapley Additive exPlanations) values to visualize feature attributions at the observation level, thereby facilitating model interpretation and comprehension. The proposed classifier performs admirably on the experimental testing dataset, with values of 0.921, 0.921, 0.923, 0.921, 0.846 and 0.846 for precision, recall, F1-score, Kappa, and MCC, respectively. This model's ultimate goal is to reduce reliance on skilled pathologists, reduce interobserver variability, and speed up the processing of breast tissue assessments.