Holistic Decomposition Convolution for Effective Semantic Segmentation of 3D MR Images
This addresses the challenge of incorporating larger context in 3D volumetric segmentation for medical imaging, though it appears incremental as it builds on existing 3D CNN methods.
The paper tackled the problem of limited spatial context in 3D medical image segmentation due to GPU memory constraints by proposing Holistic Decomposition Convolution (HDC), which achieved effective segmentation on hip T1 and intervertebral disc T2 MR images.
Convolutional Neural Networks (CNNs) have achieved state-of-the-art performance in many different 2D medical image analysis tasks. In clinical practice, however, a large part of the medical imaging data available is in 3D. This has motivated the development of 3D CNNs for volumetric image segmentation in order to benefit from more spatial context. Due to GPU memory restrictions caused by moving to fully 3D, state-of-the-art methods depend on subvolume/patch processing and the size of the input patch is usually small, limiting the incorporation of larger context information for a better performance. In this paper, we propose a novel Holistic Decomposition Convolution (HDC), for an effective and efficient semantic segmentation of volumetric images. HDC consists of a periodic down-shuffling operation followed by a conventional 3D convolution. HDC has the advantage of significantly reducing the size of the data for sub-sequential processing while using all the information available in the input irrespective of the down-shuffling factors. Results obtained from comprehensive experiments conducted on hip T1 MR images and intervertebral disc T2 MR images demonstrate the efficacy of the present approach.