US 12,205,277 B2
Tubular structure segmentation
Yikang Liu, Cambridge, MA (US); Shanhui Sun, Lexington, MA (US); Terrence Chen, Lexington, MA (US); Zhang Chen, Brookline, MA (US); and Xiao Chen, Cambridge, MA (US)
Assigned to Shanghai United Imaging Intelligence Co., Ltd., Shanghai (CN)
Filed by Shanghai United Imaging Intelligence Co., Ltd., Shanghai (CN)
Filed on Dec. 29, 2021, as Appl. No. 17/564,304.
Prior Publication US 2023/0206428 A1, Jun. 29, 2023
Int. Cl. G06T 7/00 (2017.01); G06N 3/02 (2006.01); G06T 7/11 (2017.01); G06T 7/194 (2017.01)
CPC G06T 7/0012 (2013.01) [G06N 3/02 (2013.01); G06T 7/11 (2017.01); G06T 7/194 (2017.01); G06T 2207/10121 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/30101 (2013.01)] 20 Claims
OG exemplary drawing
 
1. An apparatus, comprising:
one or more processors configured to:
receive a medical image, wherein the medical image depicts a first type of tubular structures; and
segment the first type of tubular structures from the medical image using an artificial neural network (ANN), wherein the ANN is trained to segment the first type of tubular structures through a training process that comprises:
training the ANN during a first stage of the training process to learn a segmentation model for segmenting a second type of tubular structures based on annotated medical images of the second type of tubular structures, wherein the second type of tubular structures is a different type of tubular structures than the first type of tubular structures; and
further training the ANN during a second stage of the training process to segment the first type of tubular structures based on the segmentation model learned from the first stage of the training process, wherein the second stage of the training process comprises:
providing a first training image comprising the first type of tubular structures to the ANN;
using the ANN to generate a first segmentation of the first type of tubular structures based on the first training image and the segmentation model learned from the first stage of the training process;
correcting the first segmentation generated by the ANN based on one or more characteristics of the first type of tubular structures to derive a corrected segmentation; and
adjusting the segmentation model based on a difference between the first segmentation generated by the ANN and the corrected segmentation.