Physical Color Calibration of Digital Pathology Scanners for Robust Artificial Intelligence–Assisted Cancer Diagnosis




Ji, Xiaoyi; Salmon, Richard; Mulliqi, Nita; Khan, Umair; Wang, Yinxi; Blilie, Anders; Olsson, Henrik; Pedersen, Bodil Ginnerup; Sørensen, Karina Dalsgaard; Ulhøi, Benedicte Parm; Kjosavik, Svein R.; Janssen, Emilius A.M.; Rantalainen, Mattias; Egevad, Lars; Ruusuvuori, Pekka; Eklund, Martin; Kartasalo, Kimmo.

PublisherElsevier BV

2025

Modern Pathology

Modern Pathology

100715

38

5

0893-3952

DOIhttps://doi.org/10.1016/j.modpat.2025.100715

https://doi.org/10.1016/j.modpat.2025.100715

https://research.utu.fi/converis/portal/detail/Publication/485059263



The potential of artificial intelligence (AI) in digital pathology is limited by technical inconsistencies in the production of whole slide images (WSIs). This causes degraded AI performance and poses a challenge for widespread clinical application, as fine-tuning algorithms for each site is impractical. Changes in the imaging workflow can also compromise diagnostic accuracy and patient safety. Physical color calibration of scanners, relying on a biomaterial-based calibrant slide and a spectrophotometric reference measurement, has been proposed for standardizing WSI appearance, but its impact on AI performance has not been investigated. We evaluated whether physical color calibration can enable robust AI performance. We trained fully supervised and foundation model–based AI systems for detecting and Gleason grading prostate cancer using WSIs of prostate biopsies from the STHLM3 clinical trial (n = 3651) and evaluated their performance in 3 external cohorts (n = 1161) with and without calibration. With physical color calibration, the fully supervised system’s concordance with pathologists’ grading (Cohen linearly weighted κ) improved from 0.439 to 0.619 in the Stavanger University Hospital cohort (n = 860), from 0.354 to 0.738 in the Karolinska University Hospital cohort (n = 229), and from 0.423 to 0.452 in the Aarhus University Hospital cohort (n = 72). The foundation model’s concordance improved as follows: from 0.739 to 0.760 (Karolinska), from 0.424 to 0.459 (Aarhus), and from 0.547 to 0.670 (Stavanger). This study demonstrated that physical color calibration provides a potential solution to the variation introduced by different scanners, making AI-based cancer diagnostics more reliable and applicable in diverse clinical settings.


R.S. received funding from Innovate UK (Future Leaders Fellowship MR/V023314/1). U.K. received funding from University of Turku (graduate school), Finland. A.B. received a grant from the Health Faculty at the University of Stavanger, Norway. B.G.P and K.D.S received funding from Innovation Fund Denmark (8114-00014B) for the Danish branch of the NordCaP project. M.R. received funding from Swedish Research Council and Swedish Cancer Society. P.R. received funding from the Research Council of Finland (341967) and Cancer Foundation Finland. M.E. received funding from Swedish Research Council, Swedish Cancer Society, Swedish Prostate Cancer Society, Nordic Cancer Union, Karolinska Institutet, and Region Stockholm. K.K. received funding from the SciLifeLab & Wallenberg Data Driven Life Science Program (KAW 2024.0159), David and Astrid Hägelen Foundation, Instrumentarium Science Foundation, KAUTE Foundation, Karolinska Institute Research Foundation, Orion Research Foundation, and Oskar Huttunen Foundation. Computations were made possible by the National Academic Infrastructure for Supercomputing in Sweden (NAISS) and the Swedish National Infrastructure for Computing (SNIC) at C3SE, partially funded by the Swedish Research Council (2022-06725 and 2018-05973), by the supercomputing resource Berzelius provided by National Supercomputer Centre at Linköping University and the Knut and Alice Wallenberg foundation, and by CSC—IT Center for Science, Finland. The funders of the study and the providers of computing infrastructure had no role in study design, data collection, data analysis, data interpretation, or writing of the report.


Last updated on 2025-20-03 at 13:09