Three artificial intelligence data challenges based on CT and MRI

• Published in: Diagnostic and interventional imaging Vol. 101; no. 12; pp. 783 - 788
• Main Authors: Lassau, N., Bousaid, I., Chouzenoux, E., Lamarque, J.P., Charmettant, B., Azoulay, M., Cotton, F., Khalil, A., Lucidarme, O., Pigneur, F., Benaceur, Y., Sadate, A., Lederlin, M., Laurent, F., Chassagnon, G., Ernst, O., Ferreti, G., Diascorn, Y., Brillet, P.Y., Creze, M., Cassagnes, L., Caramella, C., Loubet, A., Dallongeville, A., Abassebay, N., Ohana, M., Banaste, N., Cadi, M., Behr, J., Boussel, L., Fournier, L., Zins, M., Beregi, J.P., Luciani, A., Cotten, A., Meder, J.F.
• Format: Journal Article
• Language: English
• Published: France Elsevier Masson SAS 01.12.2020; Elsevier
• Subjects: Artificial Intelligence; Artificial intelligence (AI); Bioengineering; Computed tomography (CT); Computer Science; Deep learning; Imaging; Life Sciences; Machine learning; Magnetic resonance imaging (MRI); MS; ROC; Magnetic resonance imaging (MRI); AI; Computed tomography (CT); AUC; Deep learning; 3D; CT; 2D; Artificial intelligence (AI); Machine learning; JFR; FLAIR; DICOM; GDPR; MSE; SFR; EDSS
• ISSN: 2211-5684; 2211-5684
• DOI: 10.1016/j.diii.2020.03.006

•Three data challenges with over 1200 general data protection regulation compliant examinations each were organized.•For the Multiple Sclerosis Challenge on 3D FLAIR images, the best score (i.e., mean square error) to predict expanded disability status scale obtained by the winning team was 3.04.•For the Sarcopenia Challenge on CT images, the best score (i.e., combination of similarity index and mean square error) obtained by the winning team was 4.•For the Pulmonary Nodule Challenge on CT images, the best score (i.e., area under the curve) obtained by the winning team was 0.899. The second edition of the artificial intelligence (AI) data challenge was organized by the French Society of Radiology with the aim to: (i), work on relevant public health issues; (ii), build large, multicentre, high quality databases; and (iii), include three-dimensional (3D) information and prognostic questions. Relevant clinical questions were proposed by French subspecialty colleges of radiology. Their feasibility was assessed by experts in the field of AI. A dedicated platform was set up for inclusion centers to safely upload their anonymized examinations in compliance with general data protection regulation. The quality of the database was checked by experts weekly with annotations performed by radiologists. Multidisciplinary teams competed between September 11th and October 13th 2019. Three questions were selected using different imaging and evaluation modalities, including: pulmonary nodule detection and classification from 3D computed tomography (CT), prediction of expanded disability status scale in multiple sclerosis using 3D magnetic resonance imaging (MRI) and segmentation of muscular surface for sarcopenia estimation from two-dimensional CT. A total of 4347 examinations were gathered of which only 6% were excluded. Three independent databases from 24 individual centers were created. A total of 143 participants were split into 20 multidisciplinary teams. Three data challenges with over 1200 general data protection regulation compliant CT or MRI examinations each were organized. Future challenges should be made with more complex situations combining histopathological or genetic information to resemble real life situations faced by radiologists in routine practice.