Deep learning-based fully automatic segmentation of wrist cartilage in MR images.

Autor: Brui E; Department of Physics and Engineering, University of Information Technology, Mechanics and Optics, St Petersburg, Russia., Efimtcev AY; Department of Physics and Engineering, University of Information Technology, Mechanics and Optics, St Petersburg, Russia.; Federal Almazov North-West Medical Research Center, St Petersburg, Russia., Fokin VA; Department of Physics and Engineering, University of Information Technology, Mechanics and Optics, St Petersburg, Russia.; Federal Almazov North-West Medical Research Center, St Petersburg, Russia., Fernandez R; APHM, Service de Radiologie, Hôpital de la Conception, Marseille, France., Levchuk AG; Federal Almazov North-West Medical Research Center, St Petersburg, Russia., Ogier AC; Aix-Marseille Universite, CNRS, Centre de Résonance Magnétique Biologique et Médicale, UMR, Marseille, France., Samsonov AA; Department of Radiology, University of Wisconsin-Madison, Madison, Wisconsin, US., Mattei JP; Aix-Marseille Universite, CNRS, Centre de Résonance Magnétique Biologique et Médicale, UMR, Marseille, France.; Assistance Publique Hôpitaux de Marseille, Institut de l'appareil locomoteur, Service de Rhumatologie, Hôpital Sainte Marguerite, Marseille, France., Melchakova IV; Department of Physics and Engineering, University of Information Technology, Mechanics and Optics, St Petersburg, Russia., Bendahan D; Aix-Marseille Universite, CNRS, Centre de Résonance Magnétique Biologique et Médicale, UMR, Marseille, France., Andreychenko A; Department of Physics and Engineering, University of Information Technology, Mechanics and Optics, St Petersburg, Russia.; Department of Health Care of Moscow, Research and Practical Clinical Center of Diagnostics and Telemedicine Technologies, Moscow, Russia.
Jazyk: angličtina
Zdroj: NMR in biomedicine [NMR Biomed] 2020 Aug; Vol. 33 (8), pp. e4320. Date of Electronic Publication: 2020 May 11.
DOI: 10.1002/nbm.4320
Abstrakt: The study objective was to investigate the performance of a dedicated convolutional neural network (CNN) optimized for wrist cartilage segmentation from 2D MR images. CNN utilized a planar architecture and patch-based (PB) training approach that ensured optimal performance in the presence of a limited amount of training data. The CNN was trained and validated in 20 multi-slice MRI datasets acquired with two different coils in 11 subjects (healthy volunteers and patients). The validation included a comparison with the alternative state-of-the-art CNN methods for the segmentation of joints from MR images and the ground-truth manual segmentation. When trained on the limited training data, the CNN outperformed significantly image-based and PB-U-Net networks. Our PB-CNN also demonstrated a good agreement with manual segmentation (Sørensen-Dice similarity coefficient [DSC] = 0.81) in the representative (central coronal) slices with a large amount of cartilage tissue. Reduced performance of the network for slices with a very limited amount of cartilage tissue suggests the need for fully 3D convolutional networks to provide uniform performance across the joint. The study also assessed inter- and intra-observer variability of the manual wrist cartilage segmentation (DSC = 0.78-0.88 and 0.9, respectively). The proposed deep learning-based segmentation of the wrist cartilage from MRI could facilitate research of novel imaging markers of wrist osteoarthritis to characterize its progression and response to therapy.
(© 2020 John Wiley & Sons, Ltd.)
Databáze: MEDLINE
Externí odkaz: