The 3D Image Block Ranking Method Using Axial, Coronal, and Sagittal 2D Image Patches’ Rankings for Explainable Medical Image Classification

A great challenge for interpretable medical imaging is discovering the relationship among 3D image blocks, relevant top-ranked features such as biomarkers, associated regions, and a machine learning model’s final decisions. Firstly, a new 2D Grad-CAM-based method using feature selection is created to produce explainable 2D heatmaps with a small number of highlighted image patches corresponding to top-ranked features. Secondly, a new 2D image patch ranking algorithm is designed to leverage the newly defined feature matrices and relevant statistical data from numerous heatmaps to reliably rank axial patches, coronal patches, and sagittal patches. Thirdly, a novel 3D image block ranking algorithm is developed to generate a “Block Ranking Map (BRM)" by using the axial patch ranking scores, coronal patch ranking scores, and sagittal patch ranking scores. Lastly, a hybrid 3D image block ranking algorithm is made to generate a reliable hybrid BRM by using different block ranking scores generated by the 3D image block ranking algorithm using different top feature sets. The simulation results indicate that the novel hybrid 3D image block ranking algorithm can identify top-ranked blocks associated with important brain areas related to Alzheimer’s disease diagnosis. A doctor may conveniently use the hybrid BRM with axial, coronal, and sagittal views to better understand the relationship between the top-ranked blocks and medical diagnosis for a rational and explainable medical diagnosis.