Slicer Networks
This work addresses efficiency and accuracy challenges in medical imaging for tasks such as segmentation and registration, representing an incremental improvement with a novel architectural integration.
The paper tackles the problem of leveraging low-frequency approximations in medical image analysis by proposing the Slicer Network, which uses a learnable bilateral grid to refine feature maps, resulting in improved accuracy and efficiency across tasks like image registration and lesion segmentation.
In medical imaging, scans often reveal objects with varied contrasts but consistent internal intensities or textures. This characteristic enables the use of low-frequency approximations for tasks such as segmentation and deformation field estimation. Yet, integrating this concept into neural network architectures for medical image analysis remains underexplored. In this paper, we propose the Slicer Network, a novel architecture designed to leverage these traits. Comprising an encoder utilizing models like vision transformers for feature extraction and a slicer employing a learnable bilateral grid, the Slicer Network strategically refines and upsamples feature maps via a splatting-blurring-slicing process. This introduces an edge-preserving low-frequency approximation for the network outcome, effectively enlarging the effective receptive field. The enhancement not only reduces computational complexity but also boosts overall performance. Experiments across different medical imaging applications, including unsupervised and keypoints-based image registration and lesion segmentation, have verified the Slicer Network's improved accuracy and efficiency.