Abstract
Histopathologic evaluation is essential for categorizing and studying neuromuscular disorders. However, experienced specialists and pathologists are limited, especially in underserved areas. Although new technologies, such as artificial intelligence, are expected to improve medical reach, their use in rare diseases is challenging because of the limited availability of training datasets. To address this knowledge gap, we developed an algorithm based on deep convolutional neural networks that used data from microscopic images of hematoxylin-and-eosin-stained pathology slides. Our algorithm differentiated idiopathic inflammatory myopathies (mostly treatable) from hereditary muscle diseases (mostly non-treatable) and achieved better sensitivity and specificity than real physicians' diagnoses. Furthermore, it successfully and accurately classified four subtypes of the abovementioned muscular conditions. These results suggest that our algorithm can be safely used in a clinical setting. We established the similarity between the algorithm's and physicians' predictions using visualization technology, and clarified the validity of the predictions.
All Related Versions