A Parts Based Registration Loss for Detecting Knee Joint Areas
This work addresses the challenge of precise knee joint registration in medical imaging, which is incremental as it builds on existing registration methods with a novel loss formulation.
The paper tackles the problem of registering knee joint areas in medical images by introducing a parts-based loss that encourages similar spatial configurations of automatically selected abstract feature parts between a test image and a reference image, resulting in improved registration accuracy.
In this paper, a parts based loss is considered for finetune registering knee joint areas. Here the parts are defined as abstract feature vectors with location and they are automatically selected from a reference image. For a test image the detected parts are encouraged to have a similar spatial configuration than the corresponding parts in the reference image.