Deep learning is widely applied in medical image analysis, but up to 10% of manually labeled images may be incorrect, degrading model performance. This paper proposes an automated method to identify incorrectly labeled medical images by analyzing sequences of loss functions from deep learning classification networks over multiple training epochs. Identified images can be reviewed and relabeled by experts, improving dataset quality and model performance.
Two experiments validate the method on a fundus image dataset for referable diabetic retinopathy screening. In the first, 6% (648) of 10,788 gold-standard labels were intentionally flipped. The method identified 75.31% (488) of the flipped samples, with only 4.85% (492) false positives among correctly labeled samples.
In the second, reviewing and correcting the 980 identified samples (9.1% of the dataset) and retraining the model improved best accuracy on an independent test set from 95.93% (with 6% label noise) to 96.50% (with 1.5% noise), approaching the ideal 96.57% (with 0% noise). The results demonstrate the method's effectiveness in improving model performance through automated label quality control.
Blogger's Review: The automated method for identifying incorrectly labeled images is of significant value, especially in the medical field where data labeling quality directly affects deep learning model performance. By reducing the burden of manual reviews, it significantly enhances dataset accuracy, laying a solid foundation for further research. The experimental results also provide strong support for practical applications.