BME Frontiers / 2022 / Article / Alg 1

Research Article

Automated Segmentation and Connectivity Analysis for Normal Pressure Hydrocephalus

Algorithm 1

NPH prediction. CT scan segmentation.
Input: CT scan of the subject, MNI152 Atlas, and population average diffusion data for diffusion tractography computation in the MNI152 space
Output: 3D segmentation of CT scan into 3 regions, and NPH score
1 Compute the affine transform matrix that aligns the MNI152 atlas to the subject space
2 Compute probability maps using ground truth annotations and transform it into the subject space using
3 The CT scans are segmented using modified 3D UNet and affine transformed probability maps. The resulting segmentation is mapped back to the MNI152 space using for the tractography computations below
4 Compute ROA (Region of Avoidance) tractographs using the ventricles as the ROA
5 Compute patient-specific connectivity matrix and the derived network properties
6 Using the segmentation volumes and network properties, predict whether the subject has NPH