Quality control stress test for deep learning-based diagnostic model in digital pathology
| Autor: | Lech Nieroda, Birgid Schömig-Markiefka, Junya Fukuoka, Reinhard Büttner, Alexey Pryalukhin, Viktor Achter, Alexander Quaas, Andrey Bychkov, Yuri Tolkach, Wolfgang Hulla, Anant Madabhushi |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2021 |
| Předmět: |
0301 basic medicine
Male Quality Control Pathology medicine.medical_specialty Computer science media_common.quotation_subject Context (language use) Article 030218 nuclear medicine & medical imaging Pathology and Forensic Medicine 03 medical and health sciences 0302 clinical medicine Deep Learning Stress test Diagnostic model medicine Image Processing Computer-Assisted Humans Quality (business) Digitization media_common Artifact (error) Prostate cancer Pathology Clinical business.industry Deep learning Digital pathology Prostatic Neoplasms Reproducibility of Results Pattern recognition 030104 developmental biology Artificial intelligence Neural Networks Computer business Artifacts |
| Zdroj: | Modern Pathology |
| ISSN: | 1530-0285 0893-3952 |
| Popis: | Digital pathology provides a possibility for computational analysis of histological slides and automatization of routine pathological tasks. Histological slides are very heterogeneous concerning staining, sections' thickness, and artifacts arising during tissue processing, cutting, staining, and digitization. In this study, we digitally reproduce major types of artifacts. Using six datasets from four different institutions digitized by different scanner systems, we systematically explore artifacts' influence on the accuracy of the pre-trained, validated, deep learning-based model for prostate cancer detection in histological slides. We provide evidence that any histological artifact dependent on severity can lead to a substantial loss in model performance. Strategies for the prevention of diagnostic model accuracy losses in the context of artifacts are warranted. Stress-testing of diagnostic models using synthetically generated artifacts might be an essential step during clinical validation of deep learning-based algorithms. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full text from SpringerLink