Efficient quality control of WSIs

DDP Thesis | Prof. Amit Sethi, Dept. of Electrical Engineering, IIT Bombay

- Employed an active learning approach to train the HistoROI classifier, effectively categorizing Whole Slide Images (WSIs) into six tissue regions: epithelium, stroma, lymphocytes, adipose, artifacts, and miscellaneous
- Evaluated by comparing the foreground predictions of our deep learning-based HistoROI model against the image processing-based HistoQC tool and outperformed the later with a higher dice score on 70% of the WSIs
- Enhanced HistoROI model performance for WSI segregation by implementing Contrastive Learning methods

[Dual Degree Project Report] [Paper]