Spatially Covariant Lesion Segmentation
This addresses efficiency challenges for resource-limited clinical applications in medical imaging, though it is incremental as it builds on existing neural network methods.
The paper tackled the problem of improving computational efficiency while maintaining or increasing accuracy for lesion segmentation in medical images by proposing a spatially covariant pixel-aligned classifier (SCP). The result was a 23.8% reduction in GPU memory usage, 64.9% reduction in FLOPs, and 74.7% reduction in network size with similar or better accuracy.
Compared to natural images, medical images usually show stronger visual patterns and therefore this adds flexibility and elasticity to resource-limited clinical applications by injecting proper priors into neural networks. In this paper, we propose spatially covariant pixel-aligned classifier (SCP) to improve the computational efficiency and meantime maintain or increase accuracy for lesion segmentation. SCP relaxes the spatial invariance constraint imposed by convolutional operations and optimizes an underlying implicit function that maps image coordinates to network weights, the parameters of which are obtained along with the backbone network training and later used for generating network weights to capture spatially covariant contextual information. We demonstrate the effectiveness and efficiency of the proposed SCP using two lesion segmentation tasks from different imaging modalities: white matter hyperintensity segmentation in magnetic resonance imaging and liver tumor segmentation in contrast-enhanced abdominal computerized tomography. The network using SCP has achieved 23.8%, 64.9% and 74.7% reduction in GPU memory usage, FLOPs, and network size with similar or better accuracy for lesion segmentation.