A Crux on Deep Clustering Neural Networks for Medical Image Classification

Haesung Oh, Sujeong Han, Jaesung Lee

Abstract: In the domain of image classification, researchers have conducted extensive studies on image classification using deep clustering methods based on public large-scale image datasets. Deep clustering research has garnered increasing attention due to its adaptability across a wide range of application domains and scientific fields. However, its application in medical imaging remains relatively underexplored, largely due to the limited availability of high-quality datasets, inherent data irregularities, and significant inter-patient variability, which complicates model generalization. Despite these challenges, deep clustering exhibits strong potential in the medical domain, particularly for patient stratification and the discovery of both high-level and subtle patterns within complex medical datasets. The experimental comparison results can discover clustering neural networks suitable for medical images and verify their potential in the medical domain.