Your browser doesn't support javascript.
loading
Breast nodule classification with two-dimensional ultrasound using Mask-RCNN ensemble aggregation.
Evain, Ewan; Raynaud, Caroline; Ciofolo-Veit, Cybèle; Popoff, Alexandre; Caramella, Thomas; Kbaier, Pascal; Balleyguier, Corinne; Harguem-Zayani, Sana; Dapvril, Héloïse; Ceugnart, Luc; Monroc, Michele; Chamming's, Foucauld; Doutriaux-Dumoulin, Isabelle; Thomassin-Naggara, Isabelle; Haquin, Audrey; Charlot, Mathilde; Orabona, Joseph; Fourquet, Tiphaine; Bousaid, Imad; Lassau, Nathalie; Olivier, Antoine.
Afiliación
  • Evain E; Philips Research France, 92150 Suresnes, France; University of Lyon, CREATIS, CNRS UMR5220, Inserm U1044, INSA-Lyon, University of Lyon 1, 69100 Villeurbanne, France. Electronic address: ewan.evain@philips.com.
  • Raynaud C; Philips Research France, 92150 Suresnes, France.
  • Ciofolo-Veit C; Philips Research France, 92150 Suresnes, France.
  • Popoff A; Philips Research France, 92150 Suresnes, France.
  • Caramella T; Riviera Imagerie Médicale, 06800 Cagnes-sur-Mer, France.
  • Kbaier P; Centre d'Imagerie Médicale Toulon Hyeres Littoral, 83000 Toulon, France.
  • Balleyguier C; Department of Medical Imaging, Institut Gustave Roussy, 94800 Villejuif, France; Laboratoire d'Imagerie Biomédicale Multimodale Paris-Saclay. BIOMAPS, UMR 1281. Université Paris-Saclay, Inserm, CNRS, CEA, 94805 Villejuif, France.
  • Harguem-Zayani S; Department of Medical Imaging, Institut Gustave Roussy, 94800 Villejuif, France.
  • Dapvril H; Department of Women's Imaging, CH de Valenciennes, 59300 Valenciennes, France.
  • Ceugnart L; Department of Radiology, Centre Oscar Lambret, 59000 Lille, France.
  • Monroc M; Department of Radiology, Clinique Saint-Antoine, 76230 Bois-Guillaume, France.
  • Chamming's F; Department of Radiology, Institut Bergonié, 33000 Bordeaux, France.
  • Doutriaux-Dumoulin I; Department of Radiology, Institut de Cancérologie de l'Ouest, 44800 Saint-Herblain, France.
  • Thomassin-Naggara I; Department of Radiology, Centre Intercommunal de Créteil, 94000 Créteil, France.
  • Haquin A; Department of Radiology, Hôpital Croix Rousse, 69000 Lyon, France.
  • Charlot M; Department of Radiology, CH Lyon Sud, 69000 Lyon, France.
  • Orabona J; Department of Radiology, CH de Bastia, 20200 Bastia, France.
  • Fourquet T; Department of Radiology, CHRU Lille, 59000 Lille, France.
  • Bousaid I; Département de la Transformation Numérique et du Système d'Information, Gustave-Roussy Cancer Campus, Université Paris-Saclay, 94805 Villejuif, France.
  • Lassau N; Laboratoire d'Imagerie Biomédicale Multimodale Paris-Saclay. BIOMAPS, UMR 1281. Université Paris-Saclay, Inserm, CNRS, CEA, 94805 Villejuif, France; Department of Medical Imaging, Institut Gustave Roussy, 94800 Villejuif, France.
  • Olivier A; Philips Research France, 92150 Suresnes, France.
Diagn Interv Imaging ; 102(11): 653-658, 2021 Nov.
Article en En | MEDLINE | ID: mdl-34600861
PURPOSE: The purpose of this study was to create a deep learning algorithm to infer the benign or malignant nature of breast nodules using two-dimensional B-mode ultrasound data initially marked as BI-RADS 3 and 4. MATERIALS AND METHODS: An ensemble of mask region-based convolutional neural networks (Mask-RCNN) combining nodule segmentation and classification were trained to explicitly localize the nodule and generate a probability of the nodule to be malignant on two-dimensional B-mode ultrasound. These probabilities were aggregated at test time to produce final results. Resulting inferences were assessed using area under the curve (AUC). RESULTS: A total of 460 ultrasound images of breast nodules classified as BI-RADS 3 or 4 were included. There were 295 benign and 165 malignant breast nodules used for training and validation, and another 137 breast nodules images used for testing. As a part of the challenge, the distribution of benign and malignant breast nodules in the test database remained unknown. The obtained AUC was 0.69 (95% CI: 0.57-0.82) on the training set and 0.67 on the test set. CONCLUSION: The proposed deep learning solution helps classify benign and malignant breast nodules based solely on two-dimensional ultrasound images initially marked as BIRADS 3 and 4.
Asunto(s)
Palabras clave

Texto completo: 1 Colección: 01-internacional Base de datos: MEDLINE Asunto principal: Algoritmos / Redes Neurales de la Computación Tipo de estudio: Diagnostic_studies / Prognostic_studies Límite: Humans Idioma: En Revista: Diagn Interv Imaging Año: 2021 Tipo del documento: Article Pais de publicación: Francia

Texto completo: 1 Colección: 01-internacional Base de datos: MEDLINE Asunto principal: Algoritmos / Redes Neurales de la Computación Tipo de estudio: Diagnostic_studies / Prognostic_studies Límite: Humans Idioma: En Revista: Diagn Interv Imaging Año: 2021 Tipo del documento: Article Pais de publicación: Francia