Projection-Based 2.5D U-net Architecture for Fast Volumetric Segmentation
This addresses the problem of high storage and training time for volumetric segmentation in medical imaging or similar domains, though it is incremental as it builds on existing U-net architectures.
The paper tackles the computational inefficiency of 3D convolutions for volumetric segmentation by introducing a network that uses 2D convolutions on maximum intensity projections, resulting in faster training and better performance than 3D U-net for a binary segmentation task.
Convolutional neural networks are state-of-the-art for various segmentation tasks. While for 2D images these networks are also computationally efficient, 3D convolutions have huge storage requirements and require long training time. To overcome this issue, we introduce a network structure for volumetric data without 3D convolutional layers. The main idea is to include maximum intensity projections from different directions to transform the volumetric data to a sequence of images, where each image contains information of the full data. We then apply 2D convolutions to these projection images and lift them again to volumetric data using a trainable reconstruction algorithm.The proposed network architecture has less storage requirements than network structures using 3D convolutions. For a tested binary segmentation task, it even shows better performance than the 3D U-net and can be trained much faster.