Jin Long

Okyaz Eminaga, Timothy Jiyong Lee, Jessie Ge et al.

Background: The clinical documentation of cystoscopy includes visual and textual materials. However, the secondary use of visual cystoscopic data for educational and research purposes remains limited due to inefficient data management in routine clinical practice. Methods: A conceptual framework was designed to document cystoscopy in a standardized manner with three major sections: data management, annotation management, and utilization management. A Swiss-cheese model was proposed for quality control and root cause analyses. We defined the infrastructure required to implement the framework with respect to FAIR (findable, accessible, interoperable, re-usable) principles. We applied two scenarios exemplifying data sharing for research and educational projects to ensure the compliance with FAIR principles. Results: The framework was successfully implemented while following FAIR principles. The cystoscopy atlas produced from the framework could be presented in an educational web portal; a total of 68 full-length qualitative videos and corresponding annotation data were sharable for artificial intelligence projects covering frame classification and segmentation problems at case, lesion and frame levels. Conclusion: Our study shows that the proposed framework facilitates the storage of the visual documentation in a standardized manner and enables FAIR data for education and artificial intelligence research.

Emma Chen, Andy Kim, Rayan Krishnan et al.

A major obstacle to the integration of deep learning models for chest x-ray interpretation into clinical settings is the lack of understanding of their failure modes. In this work, we first investigate whether there are patient subgroups that chest x-ray models are likely to misclassify. We find that patient age and the radiographic finding of lung lesion, pneumothorax or support devices are statistically relevant features for predicting misclassification for some chest x-ray models. Second, we develop misclassification predictors on chest x-ray models using their outputs and clinical features. We find that our best performing misclassification identifier achieves an AUROC close to 0.9 for most diseases. Third, employing our misclassification identifiers, we develop a corrective algorithm to selectively flip model predictions that have high likelihood of misclassification at inference time. We observe F1 improvement on the prediction of Consolidation (0.008 [95% CI 0.005, 0.010]) and Edema (0.003, [95% CI 0.001, 0.006]). By carrying out our investigation on ten distinct and high-performing chest x-ray models, we are able to derive insights across model architectures and offer a generalizable framework applicable to other medical imaging tasks.

Rikiya Yamashita, Jin Long, Snikitha Banda et al.

Suboptimal generalization of machine learning models on unseen data is a key challenge which hampers the clinical applicability of such models to medical imaging. Although various methods such as domain adaptation and domain generalization have evolved to combat this challenge, learning robust and generalizable representations is core to medical image understanding, and continues to be a problem. Here, we propose STRAP (Style TRansfer Augmentation for histoPathology), a form of data augmentation based on random style transfer from non-medical style source such as artistic paintings, for learning domain-agnostic visual representations in computational pathology. Style transfer replaces the low-level texture content of an image with the uninformative style of randomly selected style source image, while preserving the original high-level semantic content. This improves robustness to domain shift and can be used as a simple yet powerful tool for learning domain-agnostic representations. We demonstrate that STRAP leads to state-of-the-art performance, particularly in the presence of domain shifts, on two particular classification tasks in computational pathology.