Automatic Classification of Proximal Femur Fractures based on Attention Models
Abstract:
We target the automatic classication of fractures from clinical X-Ray images following the Arbeitsgemeinschaft Osteosynthese (AO) classication standard. We decompose the problem into the localisation of the region-of-interest (ROI) and the classication of the localized region. Our solution relies on current advances in multi-task end-to-end deep learning. More specially, we adapt an attention model known as Spatial Transformer to learn an image-dependent localization of the ROI trained only from image classication labels. As a case study, we focus here on the classication of proximal femur fractures. We provide a detailed quantitative and qualitative validation on a dataset of 1000 images and report high accuracy with regard to inter-expert correlation values reported in the literature. «
We target the automatic classication of fractures from clinical X-Ray images following the Arbeitsgemeinschaft Osteosynthese (AO) classication standard. We decompose the problem into the localisation of the region-of-interest (ROI) and the classication of the localized region. Our solution relies on current advances in multi-task end-to-end deep learning. More specially, we adapt an attention model known as Spatial Transformer to learn an image-dependent localization of the ROI trained only from... »
Keywords:
MICCAI,CAMP,MLMI,deeplearning
Book / Congress title:
International Workshop on Machine Learning in Medical Imaging