2016/11/27 by Sayan Ghosal, Ghosal, Sayan, Nilanjan Ray +2
Computer Science · Medicine · #Advanced MRI Techniques and Applications #Advanced Neural Network Applications #Advanced Neuroimaging Techniques and Applications #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Medical Image Segmentation Techniques
paper · pdf · doi:10.48550/arxiv.1611.08796
openalex publication_date 2016/11/27 · openalex created_date 2022/10/05 · openalex updated_date 2026/07/28
Deformable registration is ubiquitous in medical image analysis. Many\ndeformable registration methods minimize sum of squared difference (SSD) as the\nregistration cost with respect to deformable model parameters. In this work, we\nconstruct a tight upper bound of the SSD registration cost by using a fully\nconvolutional neural network (FCNN) in the registration pipeline. The upper\nbound SSD (UB-SSD) enhances the original deformable model parameter space by\nadding a heatmap output from FCNN. Next, we minimize this UB-SSD by adjusting\nboth the parameters of the FCNN and the parameters of the deformable model in\ncoordinate descent. Our coordinate descent framework is end-to-end and can work\nwith any deformable registration method that uses SSD. We demonstrate\nexperimentally that our method enhances the accuracy of deformable registration\nalgorithms significantly on two publicly available 3D brain MRI data sets.\n