Pick-and-Learn: Automatic Quality Evaluation for Noisy-Labeled Image Segmentation

TL;DR

This paper presents an automatic label quality evaluation method for deep neural networks to improve medical image segmentation accuracy amid noisy labels, enhancing robustness and generalization.

Contribution

It introduces a novel automatic label quality assessment strategy and an overfitting control module for training deep networks on noisy biomedical image data.

Findings

Outperforms baseline methods in noisy label scenarios

Maintains high accuracy across different noise levels

Enhances generalization in biomedical image segmentation

Abstract

Deep learning methods have achieved promising performance in many areas, but they are still struggling with noisy-labeled images during the training process. Considering that the annotation quality indispensably relies on great expertise, the problem is even more crucial in the medical image domain. How to eliminate the disturbance from noisy labels for segmentation tasks without further annotations is still a significant challenge. In this paper, we introduce our label quality evaluation strategy for deep neural networks automatically assessing the quality of each label, which is not explicitly provided, and training on clean-annotated ones. We propose a solution for network automatically evaluating the relative quality of the labels in the training set and using good ones to tune the network parameters. We also design an overfitting control module to let the network maximally learn from the precise annotations during the training process. Experiments on the public biomedical image segmentation dataset have proved the method outperforms baseline methods and retains both high accuracy and good generalization at different noise levels.