OBELISK-Net: Fewer Layers to Solve 3D Multi-Organ Segmentation with Sparse Deformable Convolutions

A series of recent research papers have demonstrated that convolutional encoder-decoder networks excel at object delineation tasks. This is very important for medical image understanding and analysis, since segmenting different organs, anatomies and pathologies lies at the core of computer-assistance for diagnosis and interventions (Litjens et  al., 2017). While the performance of automated algorithms has rapidly and steadily increased since the advent of deep learning, there is limited knowledge of why certain architectural choices lead to empirically observed improvements.

Source: Medical Image Analysis - February 12, 2019 Category: Radiology Authors: Mattias P. Heinrich, Ozan Oktay, Nassim Bouteldja Source Type: research