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

Article Article scientifique Article Unleashing the potential of digital pathology data by training computer-aided diagnosis models without human annotations eng 2022-07-22 2022-07-22 author Marini Niccolo 0000-0002-5273-5741 937672 author Marchesin Stefano author Otalora Montenegro Juan Sebastian 878979 author Wodzinski Marek 0000-0002-8076-6246 author Caputo Alessandro author van Rijthoven Mart author Aswolinskiy Witali author Bokhorst John-Melle 0000-0001-7231-8833 author Podareanu Damian author Petters Edyta author Boytcheva Svetla 0000-0002-5542-9168 author Buttafuoco Genziana author Vatrano Simona author Fraggetta Filippo 0000-0002-0040-9987 author van der Laak Jeroen 0000-0001-7982-0754 author Agosti Maristella author Ciompi Francesco author Silvello Gianmaria 0000-0003-4970-4554 author Muller Henning 0000-0001-6800-9878 126188 author Atzori Manfredo 102 10.1038/s41746-022-00635-4 35869179 2398-6352 PMC9307641 https://www.nature.com/articles/s41746-022-00635-4 <p>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.</p> Dewey 616.0757 Dewey 025.063 European Commission EXtreme-scale Analytics via Multimodal Ontology Discovery & Enhancement 825292 EXA MODE EU H2020 13 Centre universitaire d'informatique 0D3 62 Département de radiologie et informatique médicale 227 false 469 Radiologie des urgences 800 https://github.com/ilmaro8/Multiple_Instance_Learning_instance_based https://github.com/ExaNLP/sket/ npj digital medicine 5 1 9eff1f7c-8890-4019-a063-90544d1dc2f8 Article (Published version) Public Article-scientifique-from-crossref.pdf 7698251 application/pdf CC BY-4.0 c92ebc30-3cb7-46fa-876d-34ecf6341f56 Appendix Public 41746_2022_635_MOESM1_ESM.pdf 1576106 application/pdf Reporting summary openAccess Centres et instituts / Centre universitaire d'informatique Faculté de médecine / Section de médecine clinique / Département de radiologie et informatique médicale

Title	Published in	Year	Views	Downloads

Unleashing the potential of digital pathology data by training computer-aided diagnosis models without human annotations	npj digital medicine	2022	172	55
Deep Learning for Histopathology Image Analysis From Heterogeneous and Multimodal Data Sources		2021	506	77

Archive ouverte UNIGE

Otalora Montenegro, Juan Sebastian

account_balance Affiliation entities

Research groups

Affiliation entities