Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

MONICA: Benchmarking on Long-tailed Medical Image Classification

Lie Ju, Siyuan Yan, Yukun Zhou, Yang Nan, Xiaodan Xing, Peibo Duan, Zongyuan Ge

TL;DR

A unified, well-structured codebase called Medical OpeN-source Long-taIled ClassifiCAtion (MONICA), which implements over 30 methods developed in relevant fields and evaluated on 12 long-tailed medical datasets covering 6 medical domains is built.

Abstract

Long-tailed learning is considered to be an extremely challenging problem in data imbalance learning. It aims to train well-generalized models from a large number of images that follow a long-tailed class distribution. In the medical field, many diagnostic imaging exams such as dermoscopy and chest radiography yield a long-tailed distribution of complex clinical findings. Recently, long-tailed learning in medical image analysis has garnered significant attention. However, the field currently lacks a unified, strictly formulated, and comprehensive benchmark, which often leads to unfair comparisons and inconclusive results. To help the community improve the evaluation and advance, we build a unified, well-structured codebase called Medical OpeN-source Long-taIled ClassifiCAtion (MONICA), which implements over 30 methods developed in relevant fields and evaluated on 12 long-tailed medical datasets covering 6 medical domains. Our work provides valuable practical guidance and insights for the field, offering detailed analysis and discussion on the effectiveness of individual components within the inbuilt state-of-the-art methodologies. We hope this codebase serves as a comprehensive and reproducible benchmark, encouraging further advancements in long-tailed medical image learning. The codebase is publicly available on https://github.com/PyJulie/MONICA.

MONICA: Benchmarking on Long-tailed Medical Image Classification

TL;DR

A unified, well-structured codebase called Medical OpeN-source Long-taIled ClassifiCAtion (MONICA), which implements over 30 methods developed in relevant fields and evaluated on 12 long-tailed medical datasets covering 6 medical domains is built.

Abstract

Long-tailed learning is considered to be an extremely challenging problem in data imbalance learning. It aims to train well-generalized models from a large number of images that follow a long-tailed class distribution. In the medical field, many diagnostic imaging exams such as dermoscopy and chest radiography yield a long-tailed distribution of complex clinical findings. Recently, long-tailed learning in medical image analysis has garnered significant attention. However, the field currently lacks a unified, strictly formulated, and comprehensive benchmark, which often leads to unfair comparisons and inconclusive results. To help the community improve the evaluation and advance, we build a unified, well-structured codebase called Medical OpeN-source Long-taIled ClassifiCAtion (MONICA), which implements over 30 methods developed in relevant fields and evaluated on 12 long-tailed medical datasets covering 6 medical domains. Our work provides valuable practical guidance and insights for the field, offering detailed analysis and discussion on the effectiveness of individual components within the inbuilt state-of-the-art methodologies. We hope this codebase serves as a comprehensive and reproducible benchmark, encouraging further advancements in long-tailed medical image learning. The codebase is publicly available on https://github.com/PyJulie/MONICA.
Paper Structure (19 sections, 4 figures, 9 tables)

This paper contains 19 sections, 4 figures, 9 tables.

Table of Contents

  1. Introduction
  2. Supported Tasks, Benchmarks, and Metrics
  3. Problem Definition and Supported Tasks
  4. Benchmarks and Datasets
  5. Dermatology Datasets
  6. Ophthalmology Datasets
  7. Radiology Datasets
  8. Others
  9. MONICA and Supported Methodologies
  10. Codebase Structure
  11. Supported Methodologies
  12. Re-sampling Methodologies
  13. Information Augmentation Methodologies
  14. Module Improvement Methodologies
  15. Experiments
  16. ...and 4 more sections

Figures (4)

  • Figure 1: Overview of MONICA. The benchmark is meticulously designed for the evaluation of the generalization of various long-tailed learning methodologies on medical image classification. We also develop a unified, well-structured codebase integrating over 30 methods developed in relevant fields and evaluate which on 12 long-tailed medical datasets covering 6 medical domains.
  • Figure 2: The performance and weight norms of the model trained from ISIC-2019-LT ($r$=500).
  • Figure 3: The performance of OOD detection methods with LTMIC methods.
  • Figure 4: (a) Performance and the gap between the selected epoch (based on validation set performance) and the best test set epoch across different methods. (b) Performance and the gap between the selected epoch (based on validation set performance) and the final epoch across different methods.