Explainability for regression CNN in fetal head circumference estimation from ultrasound images
Zhang, Jing; Petitjean, Caroline; Yger, Florian; Ainouz, Samia (2020), Explainability for regression CNN in fetal head circumference estimation from ultrasound images, in Cardoso, Jaime; Van Nguyen, Hien; Heller, Nicholas, Interpretable and Annotation-Efficient Learning for Medical Image Computing: Third International Workshop, iMIMIC 2020, Second International Workshop, MIL3ID 2020, and 5th International Workshop, LABELS 2020, Held in Conjunction with MICCAI 2020, Proceedings, Springer, p. 73-82. 10.1007/978-3-030-61166-8_8
TypeCommunication / Conférence
Conference titleWorkshop on Interpretability of Machine Intelligence in Medical Image Computing at MICCAI 2020
Book titleInterpretable and Annotation-Efficient Learning for Medical Image Computing: Third International Workshop, iMIMIC 2020, Second International Workshop, MIL3ID 2020, and 5th International Workshop, LABELS 2020, Held in Conjunction with MICCAI 2020, Proceedings
Book authorCardoso, Jaime; Van Nguyen, Hien; Heller, Nicholas
Number of pages292
MetadataShow full item record
Laboratoire d'analyse et modélisation de systèmes pour l'aide à la décision [LAMSADE]
Abstract (EN)The measurement of fetal head circumference (HC) is performed throughout the pregnancy to monitor fetus growth using ultra-sound (US) images. Recently, methods that directly predict biometric from images, instead of resorting to segmentation, have emerged. In our previous work, we have proposed such method, based on a regression con-volutional neural network (CNN). If deep learning methods are the gold standard in most image processing tasks, they are often considered as black boxes and fails to provide interpretable decisions. In this paper, we investigate various saliency maps methods, to leverage their ability at explaining the predicted value of the regression CNN. Since saliency maps methods have been developed for classification CNN mostly, we provide an interpretation for regression saliency maps, as well as an adaptation of a perturbation-based quantitative evaluation of explanations methods. Results obtained on a public dataset of ultrasound images show that some saliency maps indeed exhibit the head contour as the most relevant features to assess the head circumference and also that the map quality depends on the backbone architecture and whether the prediction error is low or high.
Subjects / KeywordsSaliency maps; Explanation evaluation; Regression CNN; Biometric prediction; Medical imaging
Showing items related by title and author.
Ensemble learning based on functional connectivity and Riemannian geometry for robust workload estimation Corsi, Marie-Constance; Chevallier, Sylvain; Barthélemy, Quentin; Hoxha, Isabelle; Yger, Florian Communication / Conférence
A Vascular Endothelial Growth Factor-Dependent Sprouting Angiogenesis Assay Based on an In Vitro Human Blood Vessel Model for the Study of Anti-Angiogenic Drugs Pauty, Joris; Usuba, Ryo; Cheng, Irene Gayi; Hespel, Louise; Takahashi, Haruko; Kato, Keisuke; Kobayashi, Masayoshi; Nakajima, Hiroyuki; Lee, Eujin; Yger, Florian; Soncin, Fabrice; Matsunaga, Yukiko (2018) Article accepté pour publication ou publié
State estimation with nonlinear reduced models. Application to the reconstruction of blood flows with Doppler ultrasound images Galarce, Felipe; Gerbeau, Jean-Frédéric; Lombardi, Damiano; Mula, Olga (2019) Document de travail / Working paper
Galarce, Felipe; Gerbeau, Jean-Frédéric; Lombardi, Damiano; Mula, Olga (2021) Article accepté pour publication ou publié