Advancing diagnostic performance and clinical usability of neural networks via adversarial training and dual batch normalization
This work addresses the critical problem of improving the interpretability and clinical usability of diagnostic AI for radiologists, which is a major barrier to clinical adoption.
The authors found that adversarially trained models, especially when combined with dual batch normalization, significantly improved the interpretability of saliency maps for pathology detection as rated by six experienced radiologists across X-ray, CT, and MRI scans. They also showed that these models maintained accuracy comparable to standard models on large datasets, including an external test set of 22,433 X-rays.
Unmasking the decision-making process of machine learning models is essential for implementing diagnostic support systems in clinical practice. Here, we demonstrate that adversarially trained models can significantly enhance the usability of pathology detection as compared to their standard counterparts. We let six experienced radiologists rate the interpretability of saliency maps in datasets of X-rays, computed tomography, and magnetic resonance imaging scans. Significant improvements were found for our adversarial models, which could be further improved by the application of dual batch normalization. Contrary to previous research on adversarially trained models, we found that the accuracy of such models was equal to standard models when sufficiently large datasets and dual batch norm training were used. To ensure transferability, we additionally validated our results on an external test set of 22,433 X-rays. These findings elucidate that different paths for adversarial and real images are needed during training to achieve state of the art results with superior clinical interpretability.