Unleashing the potential of digital pathology data by training computer-aided diagnosis models without human annotations.

Autor: Marini N; Information Systems Institute, University of Applied Sciences Western Switzerland (HES-SO Valais), Sierre, Switzerland. niccolo.marini@hevs.ch.; Centre Universitaire d'Informatique, University of Geneva, Geneva, Switzerland. niccolo.marini@hevs.ch., Marchesin S; Department of Information Engineering, University of Padua, Padua, Italy., Otálora S; Information Systems Institute, University of Applied Sciences Western Switzerland (HES-SO Valais), Sierre, Switzerland.; Centre Universitaire d'Informatique, University of Geneva, Geneva, Switzerland., Wodzinski M; Information Systems Institute, University of Applied Sciences Western Switzerland (HES-SO Valais), Sierre, Switzerland.; Department of Measurement and Electronics, AGH University of Science and Technology, Krakow, Poland., Caputo A; Department of Pathology, Ruggi University Hospital, Salerno, Italy.; Pathology Unit, Gravina Hospital Caltagirone ASP, Catania, Italy., van Rijthoven M; Department of Pathology, Radboud University Medical Center, Nijmegen, The Netherlands., Aswolinskiy W; Department of Pathology, Radboud University Medical Center, Nijmegen, The Netherlands., Bokhorst JM; Department of Pathology, Radboud University Medical Center, Nijmegen, The Netherlands., Podareanu D; SURFsara, Amsterdam, The Netherlands., Petters E; MicroscopeIT, Wrocław, Poland., Boytcheva S; Sirma AI, Sofia, Bulgaria.; Institute of Information and Communication Technologies, Bulgarian Academy of Sciences, Sofia, Bulgaria., Buttafuoco G; Pathology Unit, Gravina Hospital Caltagirone ASP, Catania, Italy., Vatrano S; Pathology Unit, Gravina Hospital Caltagirone ASP, Catania, Italy., Fraggetta F; Pathology Unit, Gravina Hospital Caltagirone ASP, Catania, Italy.; Pathology Unit, Cannizzaro Hospital, Catania, Italy., van der Laak J; Department of Pathology, Radboud University Medical Center, Nijmegen, The Netherlands.; Center for Medical Image Science and Visualization, Linkoping University, Linkoping, Sweden., Agosti M; Department of Information Engineering, University of Padua, Padua, Italy., Ciompi F; Department of Pathology, Radboud University Medical Center, Nijmegen, The Netherlands., Silvello G; Department of Information Engineering, University of Padua, Padua, Italy., Muller H; Information Systems Institute, University of Applied Sciences Western Switzerland (HES-SO Valais), Sierre, Switzerland.; Medical Faculty, University of Geneva, Geneva, Switzerland., Atzori M; Information Systems Institute, University of Applied Sciences Western Switzerland (HES-SO Valais), Sierre, Switzerland.; Department of Neuroscience, University of Padua, Padua, Italy.
Jazyk: angličtina
Zdroj: NPJ digital medicine [NPJ Digit Med] 2022 Jul 22; Vol. 5 (1), pp. 102. Date of Electronic Publication: 2022 Jul 22.
DOI: 10.1038/s41746-022-00635-4
Abstrakt: The digitalization of clinical workflows and the increasing performance of deep learning algorithms are paving the way towards new methods for tackling cancer diagnosis. However, the availability of medical specialists to annotate digitized images and free-text diagnostic reports does not scale with the need for large datasets required to train robust computer-aided diagnosis methods that can target the high variability of clinical cases and data produced. This work proposes and evaluates an approach to eliminate the need for manual annotations to train computer-aided diagnosis tools in digital pathology. The approach includes two components, to automatically extract semantically meaningful concepts from diagnostic reports and use them as weak labels to train convolutional neural networks (CNNs) for histopathology diagnosis. The approach is trained (through 10-fold cross-validation) on 3'769 clinical images and reports, provided by two hospitals and tested on over 11'000 images from private and publicly available datasets. The CNN, trained with automatically generated labels, is compared with the same architecture trained with manual labels. Results show that combining text analysis and end-to-end deep neural networks allows building computer-aided diagnosis tools that reach solid performance (micro-accuracy = 0.908 at image-level) based only on existing clinical data without the need for manual annotations.
(© 2022. The Author(s).)
Databáze: MEDLINE
Externí odkaz: