Augmenting prion surveillance by immunohistochemistry using artificial intelligence-based image analysis

Chronic wasting disease (CWD) and scrapie are transmissible spongiform encephalopathy diseases caused by prions, infectious forms of the prion protein. Currently, immunohistochemistry (IHC) is the sole approved diagnostic method for confirming these prion infections in formalin-fixed tissues. Evaluation of prion IHC requires specially trained veterinary pathologists to assess multiple quality control parameters as well as characteristic chromogen immunolabeling patterns. This manual slide review creates a significant bottleneck for laboratories needing to rapidly scale up surveillance during periods of increased testing demand. Given the repetitive and standardized nature of prion IHC slide review, this assay represents an ideal candidate for computer-assisted diagnostics. To address this challenge, we developed a deep learning-based image analysis approach tailored to review slides from large-scale veterinary prion disease surveillance. Our training dataset included 143 prion IHC whole-slide images containing a total of 3296 annotations. Annotated images were segmented into nonoverlapping tiles and used to fine-tune a pretrained convolutional neural network, enhancing the model's ability to recognize prion-specific quality control parameters and labeling features. When tested on a separate, blinded testing dataset of 50 CWD IHC slides, the model achieved 100% concordance for chromogenic labeling when compared with evaluation by a trained veterinary pathologist. The overarching objective of this project is to automate the initial review of prion IHC slides using deep learning-based image analysis to substantially reduce the time needed for evaluation. Implementation of this technology should enhance diagnostic consistency, improve efficiency, and provide scalable capabilities essential for comprehensive prion surveillance throughout the veterinary diagnostic laboratory network.