Tissue Artifact Segmentation and Severity Assessment for Automatic Analysis Using WSI

Traditionally, pathological analysis and diagnosis are performed by manually eyeballing glass-slide specimen under a microscope an expert. Whole slide image (WSI) is the digital produced from glass-slide. WSI enabled to be observed on computer-screen led computational pathology where computer-vision artificial intelligence utilized for automated diagnosis. With current advancement, entire can analyzed autonomously without human supervision. However, could fail or lead wrong if affected tissue artifacts such as fold air bubble depending severity. Existing artifact detection methods rely experts severity assessment eliminate artifact-affected regions analysis. This process time-consuming, exhausting undermines goal of removal evaluating their severity, which result in loss diagnostically important data. Therefore, it necessary detect then assess automatically. In this paper, we propose system that incorporates evaluation with utilizing convolutional neural networks (CNN). The proposed uses DoubleUNet segment ensemble network six fine-tuned CNN models determine method outperformed state-of-the-art accuracy 9% segmentation achieved strong correlation 97% pathologist’s assessment. robustness was demonstrated using our heterogeneous dataset practical usability ensured integrating system.