US 12,731,260 B2
Rapid, accurate and machine-agnostic segmentation and quantification method and device for coronavirus CT-based diagnosis
Xin Gao, Thuwal (SA); Longxi Zhou, Thuwal (SA); and Zhongxiao Li, Thuwal (SA)
Assigned to KING ABDULLAH UNIVERSITY OF SCIENCE AND TECHNOLOGY, Thuwal (SA)
Appl. No. 17/917,036
Filed by KING ABDULLAH UNIVERSITY OF SCIENCE AND TECHNOLOGY, Thuwal (SA)
PCT Filed Apr. 12, 2021, PCT No. PCT/IB2021/053012
§ 371(c)(1), (2) Date Oct. 5, 2022,
PCT Pub. No. WO2021/209887, PCT Pub. Date Oct. 21, 2021.
Claims priority of provisional application 63/009,026, filed on Apr. 13, 2020.
Prior Publication US 2023/0154006 A1, May 18, 2023
Int. Cl. G06T 7/11 (2017.01); G06T 7/00 (2017.01); G06N 3/02 (2006.01)
CPC G06T 7/11 (2017.01) [G06T 7/0012 (2013.01)] 16 Claims
OG exemplary drawing
 
1. A machine-agnostic segmentation and quantification method for coronavirus diagnostic, the method comprising:
receiving computer tomograph, CT, raw scans;
normalizing the CT raw scans to obtain normalized data custom character, wherein the normalized data custom character is normalized in terms of dimension, resolution, and signal intensity;
generating augmented data based on (1) the CT raw scans and (2) a dynamic simulation model that simulates a progression of infection regions over time, wherein the dynamic simulation model uses a state ψ and a transition function T, the state ψ includes the normalized data custom character and an infection mask custom character, and the transition function T includes a normalized data transition function Tcustom character that is applied to the normalized data custom character, and an infection mask transition function Tcustom character that is applied to the infection mask custom character, respectively, and the transition function Tis applied to the state ψ to generate the augmented data;
segmenting three different 2-dimensional, 2D, images from the normalized data custom character, which correspond to a same voxel, custom character, using three functions custom character and custom character respectively; and
quantizing each voxel custom character to have a value of 0 or 1, based on (1) summing up outputs of the three functions custom character and custom character applying a nixed threshold, and (2) an aggregation function g,
wherein the value 0 indicates that the voxel is not infected with the coronavirus, and the value 1 indicates that the voxel is infected with the coronavirus, and
wherein the three functions custom character and custom character are independent functions trained independently based on the augmented data.