Detection of prostate cancer in patch-level Gleason grading using deep learning
Master of Science (M.S.), Drexel University
08 Jun 2021
:
https://doi.org/10.17918/00000390
With more than 1 million diagnoses reported each year, prostate caner becomes one of the most common cancers among males. Whole slide images are examined by pathologists and scored according to the Gleason grading system. In this thesis, a deep learning system focused on patch-level Gleason grading is presented. This system generates patch-level results first and predicts a mask image for each prostate cancer slide. Patch-level classification is done under a decision flow method and a normal multi-class neural network method. The decision flow iteratively filters out areas from benign to most severe using 4 separate models. The generated mask images contains regional information regarding cancer severity for each biopsy. Each predicted mask image is then aggregated into a prognostic grading group that determines clinical decision thresholds of treatments. This work achieves similar slide-level results compared to previous works and is able to predict additional mask images for clinical practice.
44
43
- Detection of prostate cancer in patch-level Gleason grading using deep learning
- Wenhan Tan
- David E. Breen (Advisor)
- Drexel University
- Master of Science (M.S.)
- Drexel University; Philadelphia, Pennsylvania
- xiv, 87 pages
- Thesis
- English
- College of Engineering (1970-2026); Electrical (and Computer) Engineering (1970-2026); Drexel University
- 991015051548904721