US 12,437,876 B2
Disease prediction system and apparatus based on multi-relation functional connectivity matrix
Yu Zhang, Hangzhou (CN); Jun Li, Hangzhou (CN); Chaoliang Sun, Hangzhou (CN); Huan Zhang, Hangzhou (CN); Zhichao Wang, Hangzhou (CN); and Jingsong Li, Hangzhou (CN)
Assigned to ZHEJIANG LAB, Hangzhou (CN)
Filed by ZHEJIANG LAB, Hangzhou (CN)
Filed on Mar. 31, 2023, as Appl. No. 18/129,754.
Application 18/129,754 is a continuation of application No. PCT/CN2022/113907, filed on Aug. 22, 2022.
Claims priority of application No. 202210218603.3 (CN), filed on Mar. 8, 2022.
Prior Publication US 2023/0290514 A1, Sep. 14, 2023
Int. Cl. G16H 50/30 (2018.01); G06F 17/16 (2006.01); G06T 7/00 (2017.01); G06V 10/62 (2022.01)
CPC G16H 50/30 (2018.01) [G06F 17/16 (2013.01); G06T 7/0012 (2013.01); G06V 10/62 (2022.01); G06T 2207/10088 (2013.01); G06T 2207/30016 (2013.01)] 5 Claims
OG exemplary drawing
 
1. A disease prediction method based on a multi-relation functional connectivity matrix, comprising disease prediction process and a storing process, wherein the storing process comprises storing user's resting state brain functional magnetic resonance image data;
the disease prediction process comprises the following steps:
a subject acquisition and pre-processing step: acquiring a plurality of structural images, generating, with a processor, a plurality of pre-processed resting state brain functional magnetic resonance images of subjects by removing skulls from the plurality of structural images and performing head motion correction, time alignment, spatial smoothing, image registration and spatial normalization on the plurality of structural images;
a brain region time series extraction step: using a brain image atlas to extract a time series of each brain region from pre-processed data of each subject;
a Pearson correlation coefficient calculation step: calculating a Pearson correlation coefficient of the time series of every two brain regions for each subject, to obtain a Pearson correlation coefficient matrix COR between the time series of the brain regions;
a dynamic time warping step: calculating a dynamic time warping (DTW) distance of the time series of every two brain regions for each subject, to obtain an original DTW distance matrix;
a matrix conversion step: performing corresponding conversion on the DTW distance matrix obtained by the dynamic time warping step in combination with the Pearson correlation coefficient matrix calculated by the Pearson correlation coefficient calculation step for each subject, to obtain a converted DTW′ matrix which includes correlation degree and correlation direction information and whose numerical range is equivalent to a value range of the Pearson coefficient;
wherein the matrix conversion step comprises the following substeps:
a first conversion substep: inverting the DTW distance matrix to obtain a DTW inverse matrix, that is, a DTW_OP matrix;
a normalization substep: performing normal distribution fitting on the DTW_OP matrix to obtain means and standard deviations of normal distribution, and performing a normalization operation on each value in the DTW_OP matrix to obtain a DTW normalization matrix, that is, a DTW_NORM matrix;
a second conversion substep: converting all negative numbers in the DTW_NORM matrix to 0 to obtain a DTW matrix without negative numbers, that is, a DTW_POS matrix;
a third conversion substep: multiplying the values of corresponding positions in the DTW_POS matrix by 1 or −1 according to the positive or negative values in the Pearson correlation coefficient matrix to obtain the DTW′ matrix; and
a matrix combination step: performing weighted combination on the Pearson correlation coefficient matrix COR and the DTW′ matrix for each subject to obtain a functional connectivity matrix FC, wherein a combination mode of the weighted combination is FC=COR+δDTW′, where δ is a weight coefficient; and a disease prediction result is obtained through the functional connectivity matrix FC.