Publikation
Contrast-Enhancing Lesion Segmentation in Multiple Sclerosis: A Deep Learning Approach Validated in a Multicentric Cohort.
Wissenschaftlicher Artikel/Review - 22.08.2024
Greselin Martina, Lu Po-Jui, Melie-Garcia Lester, Ocampo-Pineda Mario, Galbusera Riccardo, Cagol Alessandro, Weigel Matthias, de Oliveira Siebenborn Nina, Ruberte Esther, Benkert Pascal, Müller Stefanie, Finkener Sebastian, Vehoff Jochen, Giulio Disanto, Findling Oliver, Chan Andrew T, Salmen Anke, Pot Caroline, Bridel Claire, Zecca Chiara, Derfuss Tobias, Lieb Johanna Maria, Diepers Michael, Wagner Franca, Vargas Maria Isabel, Du Pasquier Renaud A, Lalive Patrice H, Pravatà Emanuele, Weber Johannes, Gobbi Claudio, Leppert David, Kim Olaf Chan-Hi, Cattin Philippe C, Hoepner Robert, Roth Patrick, Kappos Ludwig, Kuhle Jens, Granziera Cristina
Bereiche
PubMed
DOI
Kontakt
Zitation
Art
Zeitschrift
Veröffentlichungsdatum
ISSN (Druck)
Kurzbeschreibung/Zielsetzung
The detection of contrast-enhancing lesions (CELs) is fundamental for the diagnosis and monitoring of patients with multiple sclerosis (MS). This task is time-consuming and suffers from high intra- and inter-rater variability in clinical practice. However, only a few studies proposed automatic approaches for CEL detection. This study aimed to develop a deep learning model that automatically detects and segments CELs in clinical Magnetic Resonance Imaging (MRI) scans. A 3D UNet-based network was trained with clinical MRI from the Swiss Multiple Sclerosis Cohort. The dataset comprised 372 scans from 280 MS patients: 162 showed at least one CEL, while 118 showed no CELs. The input dataset consisted of T1-weighted before and after gadolinium injection, and FLuid Attenuated Inversion Recovery images. The sampling strategy was based on a white matter lesion mask to confirm the existence of real contrast-enhancing lesions. To overcome the dataset imbalance, a weighted loss function was implemented. The Dice Score Coefficient and True Positive and False Positive Rates were 0.76, 0.93, and 0.02, respectively. Based on these results, the model developed in this study might well be considered for clinical decision support.