Brain functional connectivity correlation value adjustment method, brain functional connectivity correlation value adjustment system, brain activity classifier harmonization method, brain activity classifier harmonization system, and brain activity biomarker system

A system for harmonizing a brain activity classifier to execute a process of classification of at least one attribute based on a result of brain activity measurement of objects, comprising:

• a plurality of brain activity measurement devices provided respectively at a plurality of measurement sites, for time-sequentially measuring brain functions of a plurality of subjects;
• a computing system for correcting a brain functional connectivity correlation value as prescribed elements of a brain functional connectivity matrix representing temporal correlation of brain activities of the subjects as objects of training data;
• wherein
• said computing system includes
• a storage device for storing data, said storage device storing results of said measurements of said brain activities of a predetermined plurality of brain regions of each of a plurality of subjects including
• a plurality of traveling subjects who are common objects of measurements at each of said plurality of measurement sites,
• a plurality of healthy subjects who are objects of measurement at one or more of the plurality of measurement sites, and
• a plurality of patient subjects who are objects of measurement at one or more of the plurality of measurement sites;
• and
• said computing system further includes a processor and a memory;
• wherein
• said memory stores instructions which, when executed by said processor, performs a process comprising:
• calculating prescribed elements of said brain functional connectivity matrix for each of said plurality of subjects,
• defining a generalized linear mixed model to include a linear predictor having a plurality of factors at each of the plurality of measurement sites as fixed-effect coefficients, the generalized linear mixed model being defined so that each of the plurality of factors has a fixed effect average of zero across the plurality of measurement sites, wherein the plurality of factors are calculated separately and include
• measurement bias,
• sampling bias of healthy subjects exclusive of patient subjects of one or more diseases, and
• sampling bias of patient subjects of one or more diseases exclusive of the healthy subjects,
• and wherein the generalized linear mixed model is further defined to include a link function connecting the linear predictor to each element of the brain functional connectivity matrix
• applying said generalized linear mixed model to calculate the fixed effect of measurement bias for each element of said brain functional connectivity matrix at each of said plurality of measurement sites with respect to an average of the corresponding element across said plurality of measurement sites and across said plurality of subjects, and
• using said fixed effect of measurement bias at each of the plurality of measurement sites, correcting said brain functional connectivity correlation value as the prescribed elements of said brain functional connectivity matrix of each of the plurality of subjects measured at the corresponding measurement site and thereby obtaining an adjusted value;
• and
• training said brain activity classifier for said attribute through machine learning with feature selection, based on said adjusted value,
• wherein the generalized linear mixed model is applied using a regression model of a functional connectivity vector comprising a plurality of correlation values among brain areas defined as follows:
• Connectivity=xmTm+xShcTShc+xSdisTSdis+xdTd+xpTp+const+e, and
• ΣiNtspi=0,ΣkNsmk=0,ΣkNshShc k=0,ΣkNsdSdis k=0,d1(HC)=0,
• where Nts is the number of the plurality of traveling subjects, p is a participant factor of a particular traveling subject i, Ns is the number of the plurality of measurement sites, m is the measurement bias and mk being the measurement bias at a particular measurement site k of the Ns measurement sites, Shc is the sampling bias of healthy subjects with Shc k being the sampling bias of healthy subjects at the particular site k, Shc is the sampling bias of healthy subjects with Sdis k being the sampling bias of patient subjects at the particular site k, and d is a disease factor evaluated as a deviation from a healthy cohort HC, const is a constant, e represents noise, T represents a transpose of a vector, and xT represents a row vector,
• wherein, after a regression calculation is performed on the regression model, b-th connectivity of a subject is given as
• Connectivity=xmaTmb+xShcaTShcb+xSdisaTSdisb+xdaTdb+xpaTpb+const+e.

A method comprising:

• obtaining results of measurements of brain activities of a predetermined plurality of brain regions of each of a plurality of subjects including
• a plurality of traveling subjects for whom brain activities of a predetermined plurality of brain regions of each of said traveling subjects have been measured as common objects of measurements at each of a plurality of measurement sites,
• a plurality of healthy subjects who are objects of measurement at one or more of the plurality of measurement sites, and
• a plurality of patient subjects who are objects of measurement at one or more of the plurality of measurement sites;
• for each of said plurality of subjects, calculating prescribed elements of a brain functional connectivity matrix representing temporal correlation of said brain activities of said predetermined plurality of brain regions;
• defining a generalized linear mixed model to include a linear predictor having a plurality of factors at each of the plurality of measurement sites as fixed-effect coefficients, the generalized linear mixed model being defined so that each of the plurality of factors has a fixed effect average of zero across the plurality of measurement sites, wherein the plurality of factors are calculated separately and include
• measurement bias,
• sampling bias of healthy subjects exclusive of patient subjects of one or more diseases, and
• sampling bias of patient subjects of one or more diseases exclusive of the healthy subjects,
• and wherein the generalized linear mixed model is further defined to include a link function connecting the linear predictor to each element of the brain functional connectivity matrix;
• applying the generalized linear mixed model to calculate the fixed effect of measurement bias for each element of said brain functional connectivity matrix at each of the plurality of measurement sites with respect to an average of the corresponding element across said plurality of measurement sites and across said plurality of subjects;
• and
• using said fixed effect of measurement bias at each of the plurality of measurement sites, correcting a brain functional connectivity correlation value as the prescribed elements of said brain functional connectivity matrix of each of the plurality of subjects measured at the corresponding measurement site and thereby obtaining an adjusted value,
• wherein the generalized linear mixed model is applied using a regression model of a functional connectivity vector comprising a plurality of correlation values among brain areas defined as follows:
• Connectivity=xmTm+xShcTShc+xSdisTSdis+xdTd+xpTp+const+e, and
• ΣiNtspi=0,ΣkNsmk=0,ΣkNshShc k=0,ΣkNsdSdis k=0,d1(HC)=0,
• where Nts is the number of the plurality of traveling subjects, p is a participant factor of a particular traveling subject i, Ns is the number of the plurality of measurement sites, m is the measurement bias and mk being the measurement bias at a particular measurement site k of the Ns measurement sites, Shc is the sampling bias of healthy subjects with Shc k being the sampling bias of healthy subjects at the particular site k, Shc is the sampling bias of healthy subjects with Sdis k being the sampling bias of patient subjects at the particular site k, and dis a disease factor evaluated as a deviation from a healthy cohort HC, const is a constant, e represents noise, T represents a transpose of a vector, and xT represents a row vector,
• wherein, after a regression calculation is performed on the regression model, b-th connectivity of a subject is given as
• Connectivity=xmaTmb+xShcaTShcb+xSdisaTSdisb+xdaTdb+xpaTpb+const+e.