Imbalanced Medical Image
Imbalanced medical image classification addresses the challenge of training accurate diagnostic models when datasets contain vastly different numbers of samples per disease category. Current research focuses on improving model robustness to this imbalance using techniques like weighted loss functions, attention mechanisms that prioritize minority classes, and contrastive learning to enhance feature representation. These methods, often implemented within convolutional neural networks (CNNs) or capsule networks (CapsNets), aim to improve diagnostic accuracy, particularly for rare diseases, thereby impacting both clinical decision-making and the development of more equitable AI-driven healthcare systems.
Papers
Medical Knowledge-Guided Deep Learning for Imbalanced Medical Image Classification
Long Gao, Chang Liu, Dooman Arefan, Ashok Panigrahy, Margarita L. Zuley, Shandong Wu
Constrained Deep One-Class Feature Learning For Classifying Imbalanced Medical Images
Long Gao, Chang Liu, Dooman Arefan, Ashok Panigrahy, Shandong Wu