Physical Color Calibration of Digital Pathology Scanners for Deep Learning Based Diagnosis of Prostate Cancer
Xiaoyi Ji, Richard Salmon, Nita Mulliqi, Henrik Olsson, Lars Egevad, Pekka Ruusuvuori, Martin Eklund, Kimmo Kartasalo
Variation in whole slide image (WSI) across different scanners poses a problem for deep learning algorithms. We apply a color calibration slide to standardize WSIs from different sites and evaluate the effect of calibration on a deep learning model for prostate cancer diagnosis. We show that calibration can significantly improve the accuracy of the model.
Wednesday 6th July
Poster Session 1.1 - onsite 15:20 - 16:20, virtual 11:00 - 12:00 (UTC+2)