An automated method of identifying incorrectly labelled images based on the sequences of loss functions of deep learning networks
arXiv:2607.02594v1 Announce Type: cross Abstract: Deep learning is widely applied in medical image analysis, but up to 10% of manually labelled images may be incorrect, degrading model performance. This paper proposes an automated method to identify incorrectly labelled medical images by analyzing sequences of loss functions from deep learning classification networks over multiple training epochs. Identified images can be reviewed and relabelled by experts, improving dataset quality and model performance. Two experiments validate the method on a fundus image dataset for referable diabetic reti
The increasing reliance on deep learning in critical fields like medicine, coupled with the known prevalence of data labeling errors, makes this method timely for improving AI robustness.
Incorrectly labeled data significantly degrades AI model performance and trust, especially in high-stakes applications; this method offers a path to mitigate a fundamental weakness in current AI systems.
The ability to automatically detect and correct erroneous data labels opens the door to more reliable and accurate AI models, reducing manual review burdens and improving diagnostic capabilities.
- · AI developers
- · Healthcare providers
- · Patients
- · Data annotation companies (potentially for validation)
- · AI models trained on noisy datasets
- · Medical institutions using unvalidated AI systems
Improved accuracy and reliability of medical AI diagnostic systems.
Accelerated adoption of AI in healthcare as trust in data quality and model performance increases.
Reduced costs and improved outcomes in medical diagnoses, potentially extending to other data-intensive fields.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI