Published on:
19 July 2023
Primary Category:
Computer Vision and Pattern Recognition
Paper Authors:
Jia-Xin Zhuang,
Jiabin Cai,
Jianguo Zhang,
Wei-shi Zheng,
Ruixuan Wang
Proposes CARE method to embed attention into CNN training for imbalanced medical image data
Uses bounding boxes and Grad-CAM to drive attention to lesion regions
Improves classification accuracy on minority classes in skin and chest X-ray datasets
Works with different CNN architectures and can combine with existing imbalance approaches
Allows automated lesion localization to reduce annotation needs
Using attention to classify medical images with data imbalance
This paper proposes a method to handle data imbalance in medical image classification, where some disease classes have many more examples than rare classes. It uses an attention mechanism to help models focus on lesion regions when training on rare disease images. This improves classification of minority classes.
Detecting rare events with machine learning
Learning from mislabeled patient data
Learning to classify images incrementally from imbalanced data
Targeted data augmentation mitigates bias in machine learning models
Using Transfer Learning to Classify Breast Cancer
Patching Techniques for Precise Brain Tumor Segmentation
No comments yet, be the first to start the conversation...
Sign up to comment on this paper