Medical image classification is critical, where reliability and transparency are crucial for the safe and accurate diagnosis of diseases. Deep Convolutional Neural Networks (DCNNs) are widely used in medical image classification due to their high performance. However, they are often considered black-boxes because they offer little insight into decision-making. Therefore, improving the interpretability of DCNNs is crucial for their adoption in medical diagnoses. This paper proposes a novel three-tier self-interpretable DCNN (TS-CNN) architecture for multi-region medical image classification, which improves classification performance while being inherently interpretable. The proposed TS-CNN architecture is well-suited for medical images with multiple regions, such as images with scattered and randomly shaped lesions. The proposed architecture has three branches: a global branch that learns the relevant patterns from the raw input image; an attention branch that selects the important regions and discards the irrelevant parts for the local branch to learn; and a fusion branch that distills knowledge from both the global and local branches for classification. The proposed architecture is flexible in terms of the backbone CNNs used for classification and post-hoc interpretability methods used for attention capture. We demonstrate the flexibility and generalization of the architecture through a series of experiments involving multiple state-of-the-art CNN architectures such as DenseNet-121, Inception, Xception, and ResNet-50 as the global/local branches, each paired with GradCAM and Saliency maps as attention modules. The proposed architecture outperformed the backbone model in classification tasks on two datasets: a custom-made blob dataset and a publicly available skin lesion PAD-UFES-20 dataset, demonstrating its potential for improving accuracy in medical image classification tasks. The code related to this work can be found at: https://github.com/sikha2552/TS-CNN-A-Three-Tier-Self-Interpretable-CNN-for-Medical-Image-Classification-Empowered-with-Post-hoc.git .

Image segmentation is a critical step in computational biomedical image analysis, typically evaluated using metrics like the Dice coefficient during training and validation. However, in clinical settings without manual annotations, assessing segmentation quality becomes challenging, and models lacking reliability indicators face adoption barriers. To address this gap, we propose a novel framework for predicting segmentation quality without requiring ground truth annotations during test time. Our approach introduces two complementary frameworks: one leveraging predicted segmentation and uncertainty maps, and another integrating the original input image, uncertainty maps, and predicted segmentation maps. We present Bayesian adaptations of two benchmark segmentation models-SwinUNet and Feature Pyramid Network with ResNet50-using Monte Carlo Dropout, Ensemble, and Test Time Augmentation to quantify uncertainty. We evaluate four uncertainty estimates-confidence map, entropy, mutual information, and expected pairwise Kullback-Leibler divergence-on 2D skin lesion and 3D liver segmentation datasets, analyzing their correlation with segmentation quality metrics. Our framework achieves an R² score of 93.25 and Pearson correlation of 96.58 on the HAM10000 dataset, outperforming previous segmentation quality assessment methods. For 3D liver segmentation, Test Time Augmentation with entropy achieves an R² score of 85.03 and a Pearson correlation of 65.02, demonstrating cross-modality robustness. Additionally, we propose an aggregation strategy that combines multiple uncertainty estimates into a single score per image, offering a more robust and comprehensive assessment of segmentation quality compared to evaluating each measure independently. The proposed uncertainty-aware segmentation quality prediction network is interpreted using gradient-based methods such as Grad-CAM and feature embedding analysis through UMAP. These techniques provide insights into the model's behavior and reliability, helping to assess the impact of incorporating uncertainty into the segmentation quality prediction pipeline. The code is available at: https://github.com/sikha2552/Uncertainty-Aware-Segmentation-Quality-Prediction-Bayesian-Modeling-with-Comprehensive-Evaluation-.