Breast nodule classification with two-dimensional ultrasound using Mask-RCNN ensemble aggregation - 16/11/21
, Caroline Raynaud a, Cybèle Ciofolo-Veit a, Alexandre Popoff a, Thomas Caramella c, Pascal Kbaier d, Corinne Balleyguier e, f, Sana Harguem-Zayani e, Héloïse Dapvril g, Luc Ceugnart h, Michele Monroc i, Foucauld Chamming's j, Isabelle Doutriaux-Dumoulin k, Isabelle Thomassin-Naggara l, Audrey Haquin m, Mathilde Charlot n, Joseph Orabona o, Tiphaine Fourquet p, Imad Bousaid q, Nathalie Lassau f, e, Antoine Olivier a
Bienvenue sur EM-consulte, la référence des professionnels de santé.
Article gratuit.
Connectez-vous pour en bénéficier!
Highlights |
• | Breast nodule classification (benign vs. malignant) on two-dimensional ultrasound images initially marked as BI-RADS 3 and 4 yields an area under the receiver operating characteristic curve of 0.67. |
• | A mask region-based convolutional neural network can be implemented to solve the classification problem. |
• | Despite high variability in original ultrasound images, the neural network can discriminate between characteristics based solely on image features. |
Abstract |
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.
Le texte complet de cet article est disponible en PDF.Keywords : Artificial intelligence, Breast neoplasms, Deep learning, Neural network, Ultrasound
Plan
Vol 102 - N° 11
P. 653-658 - novembre 2021 Retour au numéro
Article précédent
- Minimally invasive loco-regional treatments for painful knee osteoarthritis: The time has come
- Stavros Spiliopoulos, Dimitrios Filippiadis
- Can we use radiomics in ultrasound imaging? Impact of preprocessing on feature repeatability
- Loïc Duron, Julien Savatovsky, Laure Fournier, Augustin Lecler
Bienvenue sur EM-consulte, la référence des professionnels de santé.