Self-Supervised Radiograph Anatomical Region Classification -- How Clean Is Your Real-World Data?

TL;DR

This paper demonstrates that self-supervised and contrastive learning methods can accurately classify the anatomical region in radiographs, even with minimal labeled data, improving data quality and labeling accuracy in medical imaging workflows.

Contribution

The study shows the effectiveness of self-supervised contrastive methods for anatomical classification in radiographs, achieving high accuracy with limited labels and identifying label errors.

Findings

Achieved 96.6% accuracy with a single model

Ensemble approach reached 97.7% accuracy

Identified 35% incorrect labels in the dataset

Abstract

Modern deep learning-based clinical imaging workflows rely on accurate labels of the examined anatomical region. Knowing the anatomical region is required to select applicable downstream models and to effectively generate cohorts of high quality data for future medical and machine learning research efforts. However, this information may not be available in externally sourced data or generally contain data entry errors. To address this problem, we show the effectiveness of self-supervised methods such as SimCLR and BYOL as well as supervised contrastive deep learning methods in assigning one of 14 anatomical region classes in our in-house dataset of 48,434 skeletal radiographs. We achieve a strong linear evaluation accuracy of 96.6% with a single model and 97.7% using an ensemble approach. Furthermore, only a few labeled instances (1% of the training set) suffice to achieve an accuracy of 92.2%, enabling usage in low-label and thus low-resource scenarios. Our model can be used to correct data entry mistakes: a follow-up analysis of the test set errors of our best-performing single model by an expert radiologist identified 35% incorrect labels and 11% out-of-domain images. When accounted for, the radiograph anatomical region labelling performance increased -- without and with an ensemble, respectively -- to a theoretical accuracy of 98.0% and 98.8%.