CPC A61B 6/037 (2013.01) [A61B 6/032 (2013.01); A61B 6/5217 (2013.01); A61B 6/5235 (2013.01); G06F 18/21 (2023.01); G06F 18/251 (2023.01); G06N 3/08 (2013.01); G06T 7/0014 (2013.01); G06T 7/70 (2017.01); G06T 9/00 (2013.01); G06T 11/008 (2013.01); G06V 10/764 (2022.01); G06V 10/806 (2022.01); G16H 30/40 (2018.01); G16H 50/20 (2018.01); G06T 2207/10081 (2013.01); G06T 2207/10104 (2013.01); G06T 2207/20076 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/30096 (2013.01); G06V 2201/03 (2022.01)] | 20 Claims |
1. A method for performing a computer-aided diagnosis (CAD), comprising:
acquiring a medical image set;
generating a three-dimensional (3D) tumor distance map corresponding to the medical image set, each voxel of the tumor distance map representing a distance from the voxel to a nearest boundary of a primary tumor present in the medical image set; and
performing neural-network processing of the medical image set to generate a predicted probability map to predict presence and locations of oncology significant lymph nodes (OSLNs) in the medical image set, wherein voxels in the medical image set are stratified and processed according to the tumor distance map, wherein:
the medical image set includes a 3D non-contrast computer tomography (CT) image and a 3D positron emission tomography (PET) image registered to the CT image; and
performing neural-network processing of the medical image set includes:
dividing voxels in each of the CT image and the PET image into tumor-proximal voxels and tumor-distal voxels according to the tumor distance map and a distance threshold;
processing the CT image with a first sub-network trained on CT images with corresponding ground-truth maps to generate a first prediction map based on the tumor-proximal voxels and a second prediction map based on the tumor-distal voxels;
processing the CT image, the PET image, and the tumor distance map with a second sub-network jointly trained on CT images, PET images, tumor distance maps, and corresponding ground-truth maps to generate a third prediction map based on the tumor-proximal voxels and a fourth prediction map based on the tumor-distal voxels; and
performing a fusion operation on the first, second, third and fourth prediction maps to generate a fused prediction map.
|