Method of investigating brain aging trajectory deviations in different brain regions of individuals with schizophrenia and brain-age prediction using same

Abstract

A method of predicting brain age for a subject having a mental health condition. The method comprises obtaining at least one medical image of a brain of a subject; preprocessing the medical image to produce a brain map; segmenting the brain map into more than one brain regions; and calculating a brain age of the subject based on a predetermined set of key features for each of the brain regions.

Description

METHOD OF INVESTIGATING BRAIN AGING TRAJECTORY DEVIATIONS IN DIFFERENT BRAIN REGIONS OF INDIVIDUALS WITH SCHIZOPHRENIA AND BRAIN-AGE PREDICTION USING SAMEFIELD OF THE INVENTION

[0001] The present disclosure relates to the technical field of neuroimaging, in particular, relates to construct brain-age prediction models for different brain regions and quantify the structural and functional deterioration of different brain regions in schizophrenia patients.BACKGROUND OF THE INVENTION

[0002] The background description provided herein is for the purpose of generally presenting the context of the present invention. The subject matter discussed in the background of the invention section should not be assumed to be prior art merely as a result of its mention in the background of the invention section. Similarly, a problem mentioned in the background of the invention section or associated with the subject matter of the background of the invention section should not be assumed to have been previously recognized in the prior art. The subject matter in the background of the invention section merely represents different approaches, which in and of themselves may also be inventions.

[0003] Schizophrenia is a chronic brain disorder with both positive and negative symptoms and a global prevalence of approximately 0.7%. The etiology of schizophrenia remains unclear thus far, and no effective treatment for completed prevention or alleviation of schizophrenia is available. Therefore, to further obtain new insights into the pathogenesis of schizophrenia, many neuroimaging-based studies have investigated schizophrenia-associated abnormalities in brain structure and function. Individuals with schizophrenia have many brain regions with a smaller-than-average volume. Similarly, abnormalities including cortical thinning in different brain regions were also observed in these patients. Diffusion tensor imaging (DTI) studies have reported significant decreases in fractional anisotropy (FA) values in specific brain regions, including the genu of the corpus callosum, right forceps minor, left inferior longitudinal fasciculus, left frontal lobe, and left temporal lobe. These findings suggest that individuals with schizophrenia have abnormalities in white matter integrity. Moreover, resting-state functional magnetic resonance imaging (fMRI) studies have reported that compared with healthy controls (HCs), individuals with schizophrenia have abnormal functional connectivity (FC) in many brainregions. It is indicated increased FC in the left insula and bilateral dorsolateral prefrontal cortex of individuals with schizophrenia. Moreover, another study indicated that individuals with schizophrenia have decreased FC within the language network. These findings jointly suggest that individuals with schizophrenia have deteriorated structure and function in various brain regions, which may manifest as deviations in brain aging trajectories in these brain regions.

[0004] With the advancement of artificial intelligence, a neuroimaging-based brain-age prediction approach has been applied in studies on neurological and psychiatric disorders, which investigated whether these diseases cause deviations in brain aging trajectory. The difference (termed the brain age gap) between chronological and brain ages was calculated. This index is mainly used to examine whether the neurological and psychiatric disorders of participants are associated with accelerated brain aging. Previous studies on brain-age prediction have also demonstrated that individuals with schizophrenia exhibit deviations in brain aging trajectories, as indicated in both T1 -weighted magnetic resonance imaging (MRI) and DTI findings. The recent studies used multimodal MRI to evaluate the brain age gap in patients with schizophrenia and obtained results consistent with those of studies using a single neuroimaging modality.

[0005] Compared with HCs, individuals with schizophrenia tend to have a brain age that is older than their chronological age. However, this observation has been based on brain-age prediction models constructed using the whole-brain images and data of each participant. Few studies on schizophrenia have focused on constructing brain-age prediction models based on different brain regions. Kaufmann et al. constructed the regional brain age to assess the different spatial brain age gap patterns across several brain disorders. They found that individuals with schizophrenia had the most pronounced acceleration of brain aging based on the model for frontal features. Man et al. found that the brain age gap had negative relationships with brain volume, including subcortical regions and the prefrontal cortex. Neuroimaging studies have indicated that schizophrenia affects different brain regions differently. Moreover, structural and functional brain abnormalities potentially cause deviations in brain aging trajectory.

[0006] Therefore, the development of brain-age prediction models for different brain regions of schizophrenia patients using different neuroimaging modalities, which can facilitate an examination of how schizophrenia affects different brain regions, are imperatively needed for yielding more comprehensive insights into the neuropathology of schizophrenia.SUMMARY OF THE INVENTION

[0007] These and other aspects of the present invention will become apparent from the following description of the preferred embodiment taken in conjunction with the following drawings, although variations and modifications therein may be effected without departing from the spirit and scope of the novel concepts of the disclosure.

[0008] In one aspect of the invention, a method of predicting brain age for a subject having a mental health condition. The method comprisesobtaining at least one medical image of a brain of a subject; preprocessing the medical image to produce a brain map; segmenting the brain map into more than one brain regions; and calculating a brain age of the subject based on a predetermined set of key features for each of the brain regions.

[0009] In one embodiment, the subject has a mental health condition.

[0010] In one embodiment, the predetermined set of key features are established using following steps identifying a set of key features for each of the brain regions; and generating a brain age prediction model for each of the brain regions based on the set of key features of the corresponding brain region; wherein the predetermined set of key features are the set of key features adopted for generating the brain age prediction model.

[0011] In one embodiment, the at least one medical image comprises a Tl-weighted MRI image.

[0012] In one embodiment, the step of preprocessing the medical image comprises reorienting the Tl-weighted MRI image; normalizing the reoriented Tl-weighted MRI image.

[0013] In one embodiment, the step of preprocessing the medical image further comprises pre-segmenting the reoriented Tl-weighted MRI image into gray matter (GM), white matter(WM), and cerebrospinal fluid regions.

[0014] In one embodiment, the brain map comprises a GM map.

[0015] In one embodiment, the at least one medical image comprises a resting-state fMRI image.

[0016] In one embodiment, the step of preprocessing the medical image comprises correcting the resting-state fMRI image; coregistering the corrected resting-state fMRI image with the Tl-weighted image; normalizing the corrected resting-state fMRI image; and resampling the normalized resting-state fMRI image.

[0017] In one embodiment, the brain map comprises a functional connectivity (FC) map.

[0018] In one embodiment, the at least one medical image comprises a DTI image.

[0019] In one embodiment, the step of preprocessing the medical image comprises correcting the DTI image; extracting data of at least one brain tissue from the corrected DTI image; fitting the extracted data of the at least one brain tissue into a tensor model; creating a FA image based on the fitted data of the at least one brain tissue; normalizing the FA image; and segmenting the FA image to produce at least one FA map.

[0020] In one embodiment, the brain map comprises a fractional anisotropy (FA) map of WM tracts.

[0021] In one embodiment, each of the brain regions comprises a plurality of voxels.

[0022] In one embodiment, the set of key features comprises a selected set of key voxels.

[0023] In one embodiment, the step of selecting the set of key features comprises calculating a Pearson’s correlation coefficient (r) between each voxel of different brain regions; and selecting at least 20% of the voxels having highest r values as the set of key voxels.

[0024] In one embodiment, the step of selecting the set of key features further comprises repeating the step of calculating the r value and the step of selecting at least 20% of the voxels for at least three times, and identifying at least three sets of key voxels; selected the set of key voxels which has been identified for at least two times as the selected set of key features.

[0025] In one embodiment, the mental health condition is schizophrenia.

[0026] In another aspect of the invention, a non-transitory computer readable medium storing a program causing a computer to execute a process for determining and predicting a brain age of a subject having a mental health condition. The process comprises obtaining at least one medical image of a brain of a subject; preprocessing the medical image to produce a brain map; segmenting the brain map into more than one brain regions; and calculating a brain age based on a predetermined set of key features for each of the brain regions.BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The accompanying drawings illustrate one or more embodiments of the invention and together with the written description, serve to explain the principles of the invention.Wherever possible, the same reference numbers are used throughout the drawings to refer to the same or like elements of an embodiment.

[0028] Fig. 1A shows illustration of neuroimaging preprocessing steps for Tl-weighted MRI, fMRI, and DTI.

[0029] Fig. IB shows a data training process for the images of schizophrenia patients.

[0030] Fig. 2A illustrates the brain regions with significantly accelerated aging and the effect sizes in participants with schizophrenia in the TAMI cohort.

[0031] Fig. 2B illustrates the brain regions with significantly accelerated aging and effect sizes in participants with schizophrenia in the BT cohort.

[0032] Fig. 3A illustrates the white matter tracts with significantly accelerated aging and the effect sizes in participants with schizophrenia in the TAMI cohort.

[0033] Fig. 3B illustrates the white matter tracts with significantly accelerated aging and effect sizes in participants with schizophrenia in the BT cohort.

[0034] Fig. 4 shows association of illness duration with brain age gaps across different brain regions in participants with schizophrenia.

[0035] Fig. 5 shows group differences in BrainAGE between participants with BD and healthy controls in 90 models for gray matter map.

[0036] Fig. 6 shows group differences in BrainAGE between participants with MDD and healthy controls in 90 models for gray matter map.

[0037] Fig. 7 shows group differences in BrainAGE between participants with BD and healthy controls in 90 models for standard deviation map.

[0038] Fig. 8 shows group differences in BrainAGE between participants with BD and healthy controls in 48 models for fractional anisotropy map.

[0039] Fig. 9 shows group differences in BrainAGE between participants with MDD and healthy controls in 48 models for fractional anisotropy map.DETAILED DESCRIPTION OF THE INVENTION

[0040] The invention will now be described more fully hereinafter with reference to the accompanying drawings, in which exemplary embodiments of the invention are shown. This invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention tothose skilled in the art. Like reference numerals refer to like elements throughout.

[0041] The terms used in this specification generally have their ordinary meanings in the art, within the context of the invention, and in the specific context where each term is used.Certain terms that are used to describe the invention are discussed below, or elsewhere in the specification, to provide additional guidance to the practitioner regarding the description of the invention. For convenience, certain terms may be highlighted, for example using italics and / or quotation marks. The use of highlighting has no influence on the scope and meaning of a term; the scope and meaning of a term is the same, in the same context, whether or not it is highlighted. It will be appreciated that same thing can be said in more than one way.Consequently, alternative language and synonyms may be used for any one or more of the terms discussed herein, nor is any special significance to be placed upon whether or not a term is elaborated or discussed herein. Synonyms for certain terms are provided. A recital of one or more synonyms does not exclude the use of other synonyms. The use of examples anywhere in this specification including examples of any terms discussed herein is illustrative only, and in no way limits the scope and meaning of the invention or of any exemplified term. Likewise, the invention is not limited to various embodiments given in this specification.

[0042] It will be understood that, as used in the description herein and throughout the claims that follow, the meaning of “a”, “an”, and “the” includes plural reference unless the context clearly dictates otherwise. Also, it will be understood that when an element is referred to as being “on” another element, it can be directly on the other element or intervening elements may be present therebetween. In contrast, when an element is referred to as being “directly on” another element, there are no intervening elements present. As used herein, the term “and / or” includes any and all combinations of one or more of the associated listed items.

[0043] It will be understood that, although the terms first, second, third etc. may be used herein to describe various elements, components, regions, layers and / or sections, these elements, components, regions, layers and / or sections should not be limited by these terms. These terms are only used to distinguish one element, component, region, layer or section from another element, component, region, layer or section. Thus, a first element, component, region, layer or section discussed below could be termed a second element, component, region, layer or section without departing from the teachings of the invention.

[0044] Furthermore, relative terms, such as “lower” or “bottom” and “upper” or “top,”may be used herein to describe one element’s relationship to another element as illustrated in the Figures. It will be understood that relative terms are intended to encompass different orientations of the device in addition to the orientation depicted in the Figures. For example, if the device in one of the figures is turned over, elements described as being on the “lower” side of other elements would then be oriented on “upper” sides of the other elements. The exemplary term “lower”, can therefore, encompasses both an orientation of “lower” and “upper,” depending of the particular orientation of the figure. Similarly, if the device in one of the figures is turned over, elements described as “below” or “beneath” other elements would then be oriented “above” the other elements. The exemplary terms “below” or “beneath” can, therefore, encompass both an orientation of above and below.

[0045] It will be further understood that the terms “comprises” and / or “comprising,” or “includes” and / or “including” or “has” and / or “having”, or “carry” and / or “carrying,” or “contain” and / or “containing,” or “involve” and / or “involving, and the like are to be open-ended, i.e., to mean including but not limited to. When used in this disclosure, they specify the presence of stated features, regions, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, regions, integers, steps, operations, elements, components, and / or groups thereof.

[0046] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and the present disclosure, and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.

[0047] As used in this disclosure, “around”, “about”, “approximately” or “substantially” shall generally mean within 20 percent, preferably within 10 percent, and more preferably within 5 percent of a given value or range. Numerical quantities given herein are approximate, meaning that the term “around”, “about”, “approximately” or “substantially” can be inferred if not expressly stated.

[0048] As used in this disclosure, the phrase “at least one of A, B, and C” should be construed to mean a logical (A or B or C), using a non-exclusive logical OR. As used herein, the term “and / or” includes any and all combinations of one or more of the associated listeditems.

[0049] Embodiments of the invention are illustrated in detail hereinafter with reference to accompanying drawings. The description below is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses. The broad teachings of the invention can be implemented in a variety of forms. Therefore, while this invention includes particular examples, the true scope of the invention should not be so limited since other modifications will become apparent upon a study of the drawings, the specification, and the following claims. For purposes of clarity, the same reference numbers will be used in the drawings to identify similar elements. It should be understood that one or more steps within a method may be executed in different order (or concurrently) without altering the principles of the invention.

[0050] The description will be made as to the embodiments of the present invention in conjunction with the accompanying drawings in Figs. 1-9.

[0051] The present invention is directed to brain-age prediction models for different brain regions of schizophrenia patients using different neuroimaging modalities, which can facilitate an examination of how schizophrenia affects different brain regions, and thus yield more comprehensive insights into the neuropathology of schizophrenia.

[0052] In one embodiment, the present invention is directed to brain-age prediction models for different brain regions of mental health patients using different neuroimaging modalities. The mental health condition of the present invention includes psychiatric disorders, neurodegenerative disorders. In one embodiment, the mental health condition of the present invention includes schizophrenia, major depression, bipolar disorder, anxiety disorder, demented disorder, Parkinson's disease and traumatic brain injury.

[0053] In one embodiment, the present invention discloses brain-age prediction models for different brain regions that were each based on Tl-weighted MRI, resting-state fMRI, or DTI. In one embodiment, the present invention has examined the impacts of schizophrenia on aging trajectory deviations in different brain regions. In one embodiment, the present invention has investigated relationships between clinicodemographic characteristics (e.g., illness duration, symptom severity, age of onset, history of nicotine use, body mass index, and antipsychotic equivalent dosage) and brain age gaps for brain regions that aged faster than usual.Example 1 - Brain-Age Prediction Models for Individuals With Schizophrenia MATERIALS AND METHODSParticipants

[0054] To construct the brain-age prediction models, a total of 230 HCs (mean age: 43.05 ± 15.59 years [range: 20-84 years]; sex distribution: 93 men and 137 women; mean Mini-Mental State Examination [MMSE] score: 29.00 ± 0,98; mean duration of education: 15.89 ± 3.67 years) from the discovery cohort “Taiwan Aging and Mental Illness” (TAMI) were recruited in the training dataset. The present invention also obtained neuroimaging data (i.e., Tl-weighted MRI, resting-state MRI, and DTI data) of 194 participants with schizophrenia (SCZ dataset) and 100 HCs (HC dataset) from the TAMI cohort to investigate the differences in brain-age gaps between the two groups. The HC dataset was also used to examine the reproducibility of the brain-age prediction models of the present invention.

[0055] In one embodiment, the present invention also incorporated an additional independent cohort (labeled as the BT cohort), including 50 HCs (BT-HC dataset) and 50 individuals with schizophrenia (BT-SCZ dataset) from the Tri-Service General Hospital Beitou Branch, for a final test of the model’s performance and comparison results. The participants with schizophrenia were diagnosed by two psychiatrists according to the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (DSM-IV-TR). The participants with mood components (i.e., schizoaffective disorder) and substance use disorders were excluded. The psychiatrists documented the participants’ antipsychotic information from their medical records. Anti-psychotic information was available for only 161 participants with schizophrenia in the both cohorts. The exclusion criterion for HCs was receiving a diagnosis of any psychiatric or neurological disease. All participants were asked to complete the MMSE, which was used to evaluate their general cognitive abilities. In addition, participants with schizophrenia received the Positive and Negative Syndrome Scale (PANSS) to assess the severity of symptoms.Image acquisition

[0056] The MRI data of all participants were obtained on a 3T MRI scanner (Siemens Magnetom Tim Trio, Erlangen, Germany) equipped with a 12-channel head coil. The scanning protocols were consistent with those commonly known in the field. Details of Tl-weighted MRI, resting-state MRI, and DTI scanning protocols are provided in the supplementary material.Image preprocessing

[0057] Gray matter map construction. In one embodiment, the present invention used Statistical Parametric Mapping (SPM) 12 and the DPABI toolbox running in MATLAB H2 022a(MathWorks, Natick, MA, USA) to preprocess the raw Tl-weighted MRI and raw resting-state fMRI data of each participant. For the Tl-weighted MRI data, preprocessing proceeded as follows: (1) the present invention reoriented images manually based on the anterior commissureposterior commissure (AC-PC) line; (2) the present invention normalized all images to the MNI152 standard space and segmented them into gray matter (GM), white matter, and cerebrospinal fluid regions; and (3) the present invention used the automated anatomical labeling (AAL) atlas to further segment the GM images and obtain 90 GM maps for each participant, as shown in Fig. 1A, which reflects illustration of neuroimaging preprocessing steps for Tl- weighted MRI, fMRI, and DTI. After data preprocessing, 90 GM, 90 FC, and 48 FA maps were obtained for subsequent analysis.FC map construction

[0058] In one embodiment, the present invention preprocessed the raw resting-state fMRI data per the following steps. First, the first five data points were removed. Second, all images are corrected by slice-timing, realigning, and manually reorienting images. Third, the reoriented images were coregistered with Tl-weighted images, normalized to the MNI152 standard space, and resampled to a 3 x 3 x 3 mm3 voxel. Fourth, covariates were regressed out, including those pertaining to the time courses of 6 head motions, white matter, and cerebrospinal fluid. Finally, temporal lowpass filtering (0.01-0.1 Hz) were also performed.

[0059] Subsequently, the present invention calculated the average voxel-wised FC using the Pearson’s correlation coefficient (r) between the blood-oxygen-level-dependent time series of each voxel. In one embodiment, the present invention also applied Fisher’s z transformation to improve the normality of the data. Finally, the present invention acquired 90 FC maps for each participant by applying the AAL atlas to segment the participants’ images, according to Fig. 1 A. FA map construction

[0060] In one embodiment, the present invention used the FMRIB Software Library v6.0 (FSL) to construct the FA maps. The preprocessing steps were as follows. (1) the eddy currents and movements in the raw DTI data were corrected by the eddy tool; (2) data on the brain tissues were extracted, and data on the nonbrain tissue were removed using the brain extraction tool; (3) FA images were created by fitting the eddy-corrected data into a tensor model at each voxel; (4) the FMRIB58 FA standard-space image was selected as the target to register and align all FA images; and (5) all FA images were normalized to the MNI152 standard space and segmentedinto 48 FA maps for each participant on the basis of the JHU-ICBM-Labels-1 mm atlas, according to Fig. 1A.Features selection

[0061] In order to reduce the irrelevant or partially relevant features, which might negatively impact model performance (e.g., overfitting to the training data), the present invention, in one embodiment, performed a feature selection procedure for the training dataset of multimodal MRI data. To identify the set of key voxels with the strongest correlations with chronological age in each brain region in all maps (i.e., the 90 GM, 90 FC, and 48 FA maps) for use as the key features, the present invention first randomly selected half of the participants in the training dataset to calculate the Pearson’s correlation coefficient (r) between each voxel of different brain regions (i.e., 90 GM and 90 FC maps) and white matter tracts (i.e., 48 FA maps) and chronological age; this step was repeated 100 times. Next, in one embodiment, the present invention refined the intersection of 50% of the voxels with the highest r value with chronological age in the 100 trials as the key voxels. In one embodiment, the present invention refined the intersection of at least 20% of the voxels with the highest r value with chronological age in the 100 trials as the key voxels. In one embodiment, the present invention refined the intersection of at least 30% of the voxels with the highest r value with chronological age in the 100 trials as the key voxels. In one embodiment, the present invention refined the intersection of at least 40% of the voxels with the highest r value with chronological age in the 100 trials as the key voxels. In one embodiment, the present invention refined the intersection of at least 50% of the voxels with the highest r value with chronological age in the 100 trials as the key voxels. In one embodiment, the present invention refined the intersection of at least 60% of the voxels with the highest r value with chronological age in the 100 trials as the key voxels. In one embodiment, the present invention refined the intersection of at least 70% of the voxels with the highest r value with chronological age in the 100 trials as the key voxels. In one embodiment, the present invention refined the intersection of at least 80% of the voxels with the highest r value with chronological age in the 100 trials as the key voxels. In one embodiment, the present invention refined the intersection of at least 90% of the voxels with the highest r value with chronological age in the 100 trials as the key voxels.

[0062] In one embodiment, the present invention then repeated the preceding two steps 100 times to obtain 100 sets of key voxels. Finally, the present invention selected the key voxelsthat were selected >10 times as the key features of each brain region. Consequently, the present invention established a set of features in 90 GM, 90 FC, and 48 FA maps to construct a predictive model of brain age in different brain regions, as shown in Fig. IB.Brain age prediction and brain age gap calculation

[0063] After comparing the brain age gaps between the two groups, we performed a multiple regression analysis for brain regions that aged faster than usual. In each regression model, the dependent variable was brain age gaps for a given brain region; the independent variables were clinicodemographic characteristics, including PANSS subscale scores (for positive symptoms, negative symptoms, and general psychopathology symptoms), illness duration, age of onset, history of nicotine use, and body mass index; and the control variables were chronological age and sex. In addition, after excluding participants without any antipsychotic information and controlling for chronological age and sex, we used a regression analysis to investigate the association between brain age gaps and chlorpromazine (CPZ) equivalent dosage. Finally, the FDR method was used to control for differences in the comparison procedures.

[0064] According to Fig. IB upper panel, after extracting voxels highly correlated with chronological age for each map as features, the present invention used the Gaussian process regression algorithm to train 228 brain-age prediction models and calculated the MAEs and Pearson’s correlation coefficient between corrected brain ages and chronological ages to evaluate the models’ performance. The HC and BT-HC datasets were also used to examine the reproducibility of the brain-age prediction models.

[0065] As shown in Fig, IB, lower panel, the trained models were applied to the HC, SCZ, BT-HC, and BT-SCZ datasets to estimate brain ages and calculate brain-age gaps. Finally, ANCOVA was used to examine the differences in brain age gaps between participants with schizophrenia and HCs for different brain regions in the TAMI and BT cohorts, respectively. DTI diffusion tensor imaging, GM gray matter, WM white matter, FC functional connectivity, FA fractional anisotropy, AAL automated anatomical labeling, MAE mean absolute error, SCZ individuals with schizophrenia, HCs healthy controls, BAG brain age gap, TAMI Taiwan Aging and Mental Illness, BT Tri-Service General Hospital Beitou Branch.

[0066] In one embodiment, the Gaussian process regression algorithm with fivefold cross-validation was used to train and assess the 90, 90, and 48 models for GM, FC, and FAmaps in the training dataset, respectively. In one embodiment, the present invention also applied the trained models to the HC and BT-HC datasets to evaluate reproducibility and predict brain ages in all participants with schizophrenia. To eliminate bias, in one embodiment, the present invention corrected all predicted brain ages using the following formulas.Brain age = a x chronological age + 0

[0067] The coefficient a represents the slope and P represents the intercept. This brain age was then corrected as follows:Correct brain age = brain age + [chronological age - (a x chronological age + P)]

[0068] The present invention calculated the mean absolute error (MAE) and Pearson’s correlation coefficient (r) between corrected brain age and chronological age to assess the performance of all models , as shown in Fig. IB upper panel. The brain age gaps in different brain regions for the TAMI cohort (194 participants with schizophrenia and 100 HCs) and BT cohort (50 participants with schizophrenia and 50 HCs) were calculated as follows, according to Fig. IB lower panel:Brain age gap = corrected brain age - chronological ageStatistical analysis

[0069] Independent t test and Chi-square test were used for the statistical analyses of continuous and categorical clinicodemographic variables, respectively. The P value was set at 0.05.

[0070] In one embodiment, the present invention used analysis of covariance (ANCOVA) to examine the differences in the brain age gaps between participants with schizophrenia and HCs in the TAMI and BT cohorts, with chronological age, sex, MMSE score, and duration of education as the covariates. Moreover, the false discovery rate (FDR) method was used to correct P values for multiple comparisons. After FDR correction, the significance level was set at 0.05. The partial eta-squared (partial T|2) values were calculated as effect size measures. In one embodiment, the BrainNet Viewer (www.nitrc.org / projects / bnv / ) was used for result visualizations.RESULTSParticipants’ clinicodemographic characteristics

[0071] In the TAMI cohort, the differences between the SCZ dataset (n = 194) and the HC dataset (n = 100) were nonsignificant in terms of age (P = 0.48) and sex (P = 0.24).Compared with the HC dataset, the SCZ dataset had worse MMSE scores and shorter durations of education (both P < 0.001). Similar to the TAMI cohort, the BT cohort had nonsignificant age (P = 0.11) and sex (P = 0.69) differences between the two groups. Moreover, the BT-HC dataset had a longer duration of education and MMSE score than did the BT-SCZ dataset (both P < 0.001).Brain-age prediction model performance

[0072] In total, 228 brain-age prediction models (i.e., 90 models for GM map, 90 models for FC map, and 48 models for FA map) were trained using the Gaussian process regression algorithm with fivefold cross-validation. In the 90 models for GM map, the results showed consistent MAEs (mean MAE: 6.00±0.38 years, range: 3.90-6.65 years) and strong correlations (mean r=0.90 ± 0.01, range: 0.88-0.95) between the corrected brain and chronological ages, as shown in Table 2. In the 90 models for FC map, the results revealed that corrected brain ages had strong correlations (mean r=0.97 ± 0.01, range: 0.94-0.99) with chronological ages and low MAEs (mean MAE: 3.28±0.78 years, range: 0.85-4.64 years, as shown in Table 3). In the 48 models for FA map, it is noted consistent MAEs (mean MAE: 5.38 ± 0.70 years, range: 2.55— 6.24 years) and strong linear correlations (mean r= 0.92 ± 0.02, range: 0.89-0.98) between corrected brain and chronological ages, as shown in Table 4.

[0073] The present invention then applied 228 trained models to the HC and BT-HC datasets to verify the reproducibility of the predictions; the results were similar to those of the training dataset as shown in Tables 2-4: In the 90 GM, 90 FC, and 48 FA map models, the mean (range) MAEs were, respectively, 6.49 ± 0.64 (4,52-7.86), 3.35 ± 0.81 (1.09-5.52), and 5.63 ± 0.93 (2.98-7.55) years for the HC dataset and 6.15 ± 0.70 (3.78-7.59), 3.91 ± 0.79 (1.89-5.84), and 5.19±0.98 (2.61-7.25) years for the BT-HC dataset; moreover, the mean (range) r was, respectively, 0.88 ± 0.02 (0.83-0.94), 0.96±0.02 (0.91-0.99), and 0.92±0.03 (0.86-0.98) for the HC dataset and 0.89 ± 0.03 (0.83-0.95), 0.95 ± 0.02 (0.92-0.99), and 0.91 ± 0.03 (0.83-0.98) for the BT-HC dataset. These results indicated the reliability and consistency of the brain-age prediction models across the different cohorts.Comparison of brain age gaps across different brain regions between groups

[0074] In the 90 models for GM map, results of the present invention revealed that the participants with schizophrenia had significantly larger brain age gaps than did the HCs in most brain regions after FDR correction in the two cohorts. Of the 90 brain regions, 71 and 66 hadsignificantly larger brain age gaps in the TAMI and BT cohorts, respectively, as shown in Figs. 2A and 2B and Table 5. Of the top 20 brain regions with the largest brain age gap in the TAMI and BT cohorts, 10 of the following brain regions were identified in both the cohorts: the left insula (TAMI cohort: adjusted P < 0.001, partial T|2 = 0.08; BT cohort: adjusted P= 0.002, partial r|2 = 0.14), right insula (TAMI cohort: adjusted P < 0.001, partial n2=0.06; BT cohort: adjusted P <0.001, partial r|2 = 0.20), opercular part of right inferior frontal gyrus (TAMI cohort: adjusted P < 0.001, partial r|2 = 0.07; BT cohort: adjusted P < 0.001, partial T|2 = 0.21), orbital part of left inferior frontal gyrus (TAMI cohort: adjusted P < 0.001, partial r|2 = 0.07; BT cohort: adjusted P = 0.001, partial r|2 = 0.16), left rolandic operculum (TAMI cohort: adjusted P < 0.001, partial r|2 = 0.06; BT cohort: adjusted P = 0.001, partial r|2 = 0.15), medial part of left superior frontal gyrus (TAMI cohort: adjusted P < 0.001, partial T|2 = 0.06; BT cohort: adjusted P = 0.002, partial r|2 = 0.13), medial orbital part of right superior frontal gyrus (TAMI cohort: adjusted P < 0.001, partial r|2 = 0.09; BT cohort: adjusted P = 0.002, partial r|2 = 0.13), right superior temporal gyrus (TAMI cohort: adjusted P = 0.001, partial r|2 = 0.05; BT cohort: adjusted P = 0.002, partial r|2 = 0.13), temporal poles of left superior temporal gyrus (TAMI cohort: adjusted P < 0.001, partial r|2 = 0.08; BT cohort: adjusted P = 0.002, partial r|2 = 0.13), and temporal poles of right superior temporal gyrus (TAMI cohort: adjusted P < 0.001, partial r|2 = 0.07; BT cohort: adjusted P = 0.002, partial r|2 = 0.13).

[0075] The results for 90 models for FC map showed a trend toward greater brain age gaps in the right middle frontal gyrus, the triangular part of the left inferior frontal gyrus, the left supplementary motor area, bilateral parahippocampal gyri, left pallidum, left superior temporal gyrus, and temporal pole of bilateral middle temporal gyri in the participants with schizophrenia compared with HCs in the TAMI cohort, according to Fig. 2A. The results for the BT cohort revealed a tendency to increase brain age in the dorsolateral part of the left superior frontal gyrus in patients with schizophrenia relative to the HCs, according to Fig. 2B. However, the differences between the two groups in the TAMI and BT cohorts after the FDR correction were nonsignificant, as detailed in Table 6.

[0076] As presented above, Figs. 2A and 2B illustrate group differences in brain age gaps between participants with schizophrenia and HCs in 90 models for GM map in the two cohorts. Fig. 2A illustrates the brain regions with significantly accelerated aging and the effect sizes in participants with schizophrenia in the TAMI cohort. Fig. 2B illustrates the brain regions withsignificantly accelerated aging and effect sizes in participants with schizophrenia in the BT cohort. The brain regions in the left panel display significant differences in brain age gaps after FDR correction in the TAMI and BT cohorts. The color bar represents effect size (partial r|2). The right panel presents bar charts of effect sizes (partial T|2 values) for brain regions with significant differences. The brain regions corresponding to the AAL number are presented in Table 5. Participants with schizophrenia had significant brain aging trajectory deviations in 71 of the 90 models for GM map in the TAMI cohort. Moreover, 66 brain regions had significantly larger brain age gaps in participants with schizophrenia than in HCs in the BT cohort. The present invention found 10 brain regions with the most pronounced deterioration in both cohorts, occurring primarily in the frontal lobe, temporal lobe, and insula, as detailed in Table 5. Abbreviations are as follow: TAMI Taiwan Aging and Mental Illness, BT Tri-Service General Hospital Beitou Branch, FDR false discovery rate, AAL automated anatomical labeling.

[0077] For the TAMI and BT cohorts in the 48 models for FA map, the participants with schizophrenia and HCs significantly differed in the aging trajectory of 15 and 33 white matter tracts, respectively, as shown in Figs. 3A and 3B, and Table 7. The present invention identified the largest brain age gap in the TAMI and BT cohorts in the 10 following white matter tracts: the middle cerebellar peduncle (TAMI cohort: adjusted P < 0.001, partial r|2 = 0.09; BT cohort: adjusted P = 0.002, partial2 = 0.12), body of corpus callosum (TAMI cohort: adjusted P = 0.05, partial r|2 = 0.02; BT cohort: adjusted P < 0.001, partial r|2 = 0.21), fornix column and body of fornix (TAMI cohort: adjusted P = 0.006, partial r|2 = 0.04; BT cohort: adjusted P = 0.002, partial T|2 = 0.12), left cerebral peduncle (TAMI cohort: adjusted P = 0.05, partial T|2 = 0.02; BT cohort: adjusted P = 0.001, partial r|2 = 0.14), bilateral anterior limb of internal capsule (left: TAMI cohort: adjusted P = 0.05, partial r|2 = 0.02; BT cohort: adjusted P = 0.02, partial r|2 = 0.06. right: TAMI cohort: adjusted P = 0.02, partial r|2 = 0.03; BT cohort: adjusted P = 0.02, partial r|2 = 0.06), left posterior thalamic radiation (TAMI cohort: adjusted P = 0.05, partial r|2 = 0.02; BT cohort: adjusted P = 0.03, partial |2 = 0.06), right sagittal stratum (TAMI cohort: adjusted P = 0.03, partial r|2 = 0.03; BT cohort: adjusted P = 0.006, partial r|2 = 0.09), right cingulum (hippocampus) (TAMI cohort: adjusted P = 0.02, partial r|2 = 0.03; BT cohort: adjusted P = 0.04, partial |2 = 0.05), and right fornix cres / stria terminalis (TAMI cohort: adjusted P = 0.01, partial r|2 = 0.04; BT cohort: adjusted P = 0.004, partial T|2 = 0.11).

[0078] Figs. 3A and 3B reflects group differences in brain-age gaps between participantswith schizophrenia and HCs in 48 models for FA map in the two cohorts. Fig. 3A illustrates the white matter tracts with significantly accelerated aging and the effect sizes in participants with schizophrenia in the TAMI cohort. Fig. 3B illustrates the white matter tracts with significantly accelerated aging and effect sizes in participants with schizophrenia in the BT cohort. The white matter tracts in the left panel indicate significant differences in brain age gaps after FDR correction in the TAMI and BT cohorts. The color bar represents effect size (partial T|2). The right panel illustrates bar charts of effect sizes (partial r|2 values) for white matter tracts with significant differences. The white matter tracts corresponding to the JHU-ICBM-Label number are presented in Table 7. Participants with schizophrenia had significantly larger brain age gaps in 15 of the 48 models for FA map in the TAMI cohort. Moreover, 33 white matter tracts had significantly larger brain age gaps in participants with schizophrenia than in HCs in the BT cohort, ad detailed in Table 7. Abbreviations are as follow: FA: fractional anisotropy, TAMLTaiwan Aging and Mental Illness, BT: Tri-Service General Hospital Beitou Branch, FDR: false discovery rate.

[0079] Association of deviated brain aging trajectories with clinicodemographic characteristics across different brain regions in participants with schizophrenia

[0080] In one embodiment, the present invention performed multiple regression analysis to further investigate the relationships between brain regions demonstrating accelerated aging and clinicodemographic characteristics of the participants with schizophrenia. The results suggested that illness duration was positively correlated with brain age gaps in 22 GM regions and 10 white matter tracts: bilateral insula (left: beta = 0.23, T = 3.16, adjusted P = 0.009; right: beta = 0.28, T = 3.56, adjusted P = 0.004), bilateral posterior cingulate gyri (left: beta = 0.25, T = 3.56, adjusted P = 0.005; right: beta = 0.21, T = 3.20, adjusted P = 0.008), temporal poles of bilateral superior temporal gyri (left: beta = 0.31, T = 4.11, adjusted P = 0.004; right: beta = 0.28, T = 3.97, adjusted P = 0.003), orbital part of bilateral inferior frontal gyri (left: beta = 0.26, T = 3.43, adjusted P = 0.006; right: beta = 0.28, T = 3.91, adjusted P = 0.002), bilateral lingual gyrus (left: beta = 0.28, T = 3.74, adjusted P = 0.004; right: beta = 0.20, T = 2.74, adjusted P = 0.024), orbital part of left superior frontal gyrus (beta = 0.19, T = 3.07, adjusted P = 0.010), left rolandic operculum (beta = 0.30, T = 3.39, adjusted P = 0.006), right hippocampus (beta = 0.22, T = 2.97, adjusted P = 0.013), left parahippocampal gyri (beta = 0.25, T = 3.27, adjusted P = 0.007), left amygdala (beta = 0.30, T = 4.11, adjusted P = 0.002), left calcarine fissure and surroundingcortex (beta = 0.22, T = 3.38, adjusted P = 0.006), right cuneus (beta = 0.20, T = 3.21, adjusted P = 0.008), left middle occipital gyrus (beta = 0.25, T = 3.10, adjusted P = 0.009), right postcentral gyrus (beta = 0.25, T = 3.57, adjusted P = 0.005), left superior temporal gyrus (beta = 0.27, T = 3.61, adjusted ? = 0.005), left middle temporal gyrus (beta = 0.25, T = 3.11, adjusted ? = 0.009), temporal pole of left middle temporal gyrus (beta = 0.21, T = 3.31, adjusted P = 0.007), middle cerebellar peduncle (beta = 0.16, T = 2.58, adjusted P = 0.020), body of corpus callosum (beta = 0.28, T = 4.94, adjusted P < 0.001), right inferior cerebellar peduncle (beta = 0.12, T = 3.20, adjusted P = 0.005), left cerebral peduncle (beta = 0.26, T = 3.92, adjusted P = 0.001), bilateral posterior corona radiata (left: beta = 0.17, T = 3.30, adjusted P = 0.007; right: beta = 0.17, T = 3.59, adjusted P = 0.003), left sagittal stratum (beta = 0.18, T = 3.80, adjusted P = 0.002), bilateral fornix cres / stria terminalis (left: beta = 0.22, T = 4.41, adjusted P < 0.001; right: beta = 0.18, T = 3.23, adjusted P = 0.005), and right tapetum (beta = 0.14, T = 3.25, adjusted ? = 0.006) , as shown in Fig. 4, in which panel A shows there were significantly positive correlations between illness duration and brain age gaps in 22 brain regions; and panel B shows ten white matter tracts were positively associated with illness duration.

[0081] For the result of age of onset, the present invention found that the age of onset of individuals with schizophrenia had a negative association with the brain age gaps in four white matter tracts, including middle cerebellar peduncle (beta = -0.18, T = -3.41, adjusted P = 0.01), splenium of corpus callosum (beta = -0.18, T = -2.88, adjusted P = 0.05), right cerebral peduncle (beta = -0.18, T = -2.76, adjusted P = 0.05), and left cerebral peduncle (beta= -0.21, T = -3.76, adjusted P = 0.01).

[0082] With regard to symptom severity, nonsignificant associations between PANSS subscale scores and brain age gaps were noted. In addition, the present invention revealed that CPZ dosages, history of nicotine use, and body mass index had no significant correlations with brain age gaps.DISCUSSION

[0083] The present invention constructs brain-age prediction models based on multimodal MRI data for different brain regions of individuals with schizophrenia and quantify their brain aging trajectory deviations. The data were also derived from cohorts from multiple centers. Individuals with schizophrenia exhibited accelerated aging in brain structures based on GM and FA maps in both the cohorts. Moreover, the brain aging trajectory deviations variedamong the brain regions. However, the brain-age prediction models based on FC maps indicated that the brain aging trajectories in the different brain regions of these individuals were similar to those of HCs. In addition, the large brain age gaps in 22 GM regions and 10 white matter tracts in individuals with schizophrenia increased with the progression of the illness.

[0084] In one embodiment, the present invention showed that the individuals with schizophrenia had accelerated aging in most brain regions in models for GM map, with the most considerable deterioration occurring primarily in the frontal lobe, temporal lobe, and insula. It is suggested that the most severe brain structural abnormalities related to schizophrenia mainly occurred in the frontotemporal regions. It has found that cortical thickness significantly contributes to normalized predicted age differences in schizophrenia in the frontal lobe, bilateral precunei, middle temporal gyri, temporal poles, lateral orbitofrontal gyri, and superior parietal gyri. It is also found that the subcortical regions and medial and lateral prefrontal cortices had the most significant negative correlations between GM volume and the brain age gap. Moreover, among the different brain regions, the frontal lobe of individuals with schizophrenia had the most pronounced acceleration of brain aging. It has been revealed that individuals with schizophrenia had significantly larger brain age gaps than those of healthy controls across different durations of illness in the brain volume and cortical thickness models. Neuroimaging-based studies have indicated that GM volume abnormalities mainly occur in the insular cortex, temporal poles, middle cingulum, thalamus, and orbital part of inferior and middle frontal gyri in individuals with schizophrenia. The results of a meta-analysis demonstrated that patients with schizophrenia have medium-size volume reductions in bilateral insula, particularly in the anterior insular subregion. These findings jointly suggest that the brain age gap can effectively reflect schizophrenia-related deterioration in brain structure.

[0085] Moreover, the disease-related structural alteration in the frontal lobe, temporal lobe, and insula may play a critical role in schizophrenia.

[0086] The present invention has constructed brain-age prediction models based on FC maps for individuals with schizophrenia, and is the first study to train a machine learning model to predict brain age in individuals with schizophrenia based on FC maps. Although the individuals with schizophrenia exhibited no significantly accelerated brain aging in the 90 models for FC map, the present invention found similar trends in the frontal and temporal lobes; this was similar to the results of models for GM map. A multimodal MRI study observedacceleration of brain age in young patients with schizophrenia and reported that several brain regions, including the temporal lobe, insula, and parietal lobe, are crucial features in a brain-age prediction model. In one embodiment, amplitude of low frequency fluctuation (ALFF), regional homogeneity (ReHo), degree centrality (DC) values, and other parameters, in addition to FC, are extracted to train and predict brain age in schizophrenia.

[0087] The 48 models for FA map detected 15 white matter tracts that exhibited deviating aging trajectories in the TAMI cohort and 33 white matter tracts that exhibited larger brain age gaps than HCs in the BT cohort. In one embodiment, the FA values of the fornix column and body of fornix, splenium of corpus callosum, left superior longitudinal fasciculus, and left superior corona radiata were extracted as the major features in the prediction model and found that young patients with schizophrenia had aberrant brain aging trajectories. Brain-age studies have indicated that individuals with schizophrenia had a brain age that was older than those of HCs in the FA-based model. However, recent study suggested that there were nonsignificant differences in the global brain age gap between participants with schizophrenia and healthy controls across different illness durations in the FA model. The possible reason is that computing the global brain age gap might reduce the sensitivity for detecting the deviation of aging trajectories of individual white matter tracts. The present invention also demonstrated that individuals with schizophrenia have lower FA values in the genu and body of corpus callosum, internal capsule, fornix, anterior and superior corona radiata, and cingulum. The fornix and cingulum (hippocampus), connected to the hippocampus, play essential roles in memory, and patients with neurodegenerative and psychiatric disorders exhibit abnormalities in these regions. In the present invention, the middle cerebellar peduncle of the individuals with schizophrenia in both the TAMI and BT cohorts had a relatively older brain age — suggesting that schizophrenia also results in white matter deterioration in the cerebellum. Studies have revealed that the disconnection between the cerebrum and cerebellum might result in FA reduction in the middle cerebellar peduncle along with cognitive impairments in individuals with schizophrenia. Thus, the present invention suggested that cerebrocerebellar connectivity disruptions might be involved in the neuropathology of schizophrenia.

[0088] In one embodiment, the present invention showed that in individuals with schizophrenia, brain aging accelerated with disease progression in 22 brain regions, including the frontal lobe, temporal lobe, parietal lobe, occipital lobe, insula, and subcortical regions; 10 whitematter tracts, including the cerebrum and cerebellum, also demonstrated brain aging exacerbation with disease progression. In the TAMI cohort, schizophrenia duration was 15.56 (range, 0-38) years, whereas it was 24.20 (range, 10-45) years in the BT cohort; this further explains the larger effect sizes in the BT cohort in the GM and FA map models compared with the TAMI cohort. In addition, compared with the TAMI cohort, the BT cohort had more white matter tracts that exhibited aging trajectory deviations. A meta-analysis found that compared with those with first-episode schizophrenia, patients with chronic schizophrenia exhibited cortical thinning in the right insula, orbital part of the right inferior frontal gyrus, left lateral middle temporal cortex, and right temporal pole. DTI studies have demonstrated that compared with HCs, individuals with chronic schizophrenia, but not those with first-episode schizophrenia, have lower FA values relative to healthy individuals. In the present invention, most brain regions exhibited brain aging acceleration in individuals with schizophrenia, but only 22 brain regions and 10 white matter tracts had positive associations with illness duration. In previous studies, brain age gap was non-significantly correlated with illness duration because the studies did not construct brain-region-differentiated prediction models of brain age.

[0089] The present invention suggested that individuals with early-onset schizophrenia might have greater deterioration of specific white matter tracts than those of individuals with late-onset schizophrenia. The onset of individuals with schizophrenia often occurs before the full maturation of white matter and is considered a neurodevelop mental disorder. Previous studies found that participants at ultra-high risk of psychosis who later developed psychosis had greater abnormalities of white matter integrity than those who did not transit to psychosis. These findings of an onset-related deterioration in FA indicated that psychosis might result from a stall in white matter maturation.

[0090] The main strength of the present invention is the development of brain-age prediction models for different brain regions based on data from Tl-weighted MRI, resting-state fMRI, and DTI. In one embodiment, the present invention also included two different cohorts to validate and test the models and results. The methodology of the present invention quantifies the structural and functional decline of different brain regions in individuals with schizophrenia and provide personalized quantification for clinical explainability. For clinical applications of brainage prediction in neuropsychiatric disorders, the brain age gap could also serve as an indicator for psychiatrists to assess treatment effects (e.g., whether the brain age gap decreases after thepatients receive the treatment). In addition, as opposed to the more abstract concepts of psychiatric disorders and symptoms, the brain-age prediction approach provides patients with a more straightforward understanding of their disease and treatment progression, which may further improve their treatment compliance and insight. Therefore, brain-age prediction is a promising and innovative approach for diagnosing and assessing the course and treatment responses of psychiatric disorders, which could be effectively used in clinical practice.

[0091] The present invention establishes brain-age prediction models for different brain regions in individuals with schizophrenia through multimodal MRI. In one embodiment, the present invention discloses that most GM regions in individuals with schizophrenia exhibit accelerated aging, particularly in the frontal lobe, temporal lobe, and insula and that parts of the white matter tracts, including the cerebrum and cerebellum, demonstrate aging trajectory deviations in individuals with schizophrenia. Notably, the accelerated aging of specific brain regions and white matter tracts worsens with the advancement of illness duration. Moreover, four white matter tracts had negative associations with the age of onset of individuals with schizophrenia. The different brain regions of individuals with schizophrenia differ in their deviations of aging trajectories. By constructing brain-age prediction models for different brain regions, the present invention quantifies the structural and functional deterioration of different brain regions and white matter tracts in schizophrenia. In addition, the methodology of the present invention examines the effects of schizophrenia on the dynamics in different brain regions and provide personalized quantification for clinical explainability.Example 2 & 3 - Brain-Age Prediction Models for Individuals With Bipolar Disorder & Major Depressive DisorderMethodsParticipants

[0092] Participants of the present invention were sourced from the discovery cohort “Taiwan Aging and Mental Illness” (TAMI). In one embodiment, the present invention established brain age prediction models using data from 230 healthy individuals aged 20 to 85. These models encompass 90 gray matter maps, 90 standard deviation maps, and 48 fractional anisotropy maps, totaling 228 brain age prediction models. Additionally, in one embodiment, the present invention included 68 MDD patients, 110 BD patients, and 110 healthy controls. The patient groups consisted of individuals diagnosed with BD or MDD, per the Diagnostic andStatistical Manual of Mental Disorders (DSM-IV-TR). Exclusion criteria for the study included any DSM-IV-TR diagnosis of schizophrenia or other psychoses, intellectual disability, organic mental disorders, autoimmune or immunological diseases, recent substance abuse, current pregnancy or breastfeeding, and unstable physical illness. The healthy control group was selected based on a lack of any history of neurological or psychiatric disorders. All participants provided written informed consent, and the Institutional Review Board of Taipei Veterans General Hospital approved this study.Image acquisition

[0093] In one embodiment, the MRI data acquisition for all participants of the present invention was conducted using a 3T MRI scanner (Siemens Magnetom Tim Trio, Erlangen, Germany) equipped with a 12-channel head coil at National Yang Ming Chiao Tung University. The scanning protocols consisted with those established in existing technologies.

[0094] In one embodiment, for T1 -weighted MRI, data acquisition was performed using a sagittal 3D magnetization-prepared rapid gradient echo (MPRAGE) sequence. The specific parameters set for this sequence were as follows: repetition time (TR) = 2530 ms, echo time (TE) = 3.5 ms, inversion time = 1100 ms, matrix size = 256 x 256, the number of slices = 192, slice thickness = 1 mm, voxel size = 1.0 x 1.0 x 1.0 mm3, and a flip angle of 7°.

[0095] In one embodiment, for the acquisition of resting-state fMRI images, a T2*- weighted gradient-echo-planar imaging (EPI) sequence was used. The settings for this sequence included: TR = 2500 ms, TE = 27 ms, matrix size = 64 x 64, voxel size = 3.4 x 3.4 x 3.4 mm3, total time points = 200, field of view (FOV) = 200 mm, and a flip angle of 77°.

[0096] In one embodiment, DTI images were obtained using a single-shot spin-echo EPI sequence in the axial plane. The parameters for this sequence were: TR = 11,000 ms, TE = 104 ms, number of excitations = 3, matrix size = 128 x 128, FOV = 26 cm, the number of slices = 70, slice thickness = 2.0 mm, a b-value of 1000 s / mm2, thirty isotropic diffusion directions, and three nondiffusion weighted T2 images.Image preprocessingGray matter map construction

[0097] In one embodiment, the construction of gray matter maps involved the preprocessing of raw T1 -weighted MRI and raw resting-state fMRI data for each participant using Statistical Parametric Mapping (SPM) 12 and the DPABI toolbox, operating withinMATLAB R2022a (MathWorks, Natick, MA, USA). In one embodiment, the specific steps included: (1) manual reorientation of the T1 -weighted MRI images based on the anterior commissure-posterior commissure line; (2) normalization of all images to the MNI152 standard space, followed by segmentation into gray matter, white matter, and cerebrospinal fluid regions; and (3) utilization of the automated anatomical labeling (AAL) atlas for further segmentation of the gray matter images, resulting in 90 gray matter maps for each participant.Standard deviation map construction

[0098] In one embodiment, the processing of raw resting-state fMRI data encompassed several key steps: (1) removal of the first five data points; (2) application of slice-timing correction, realignment, and manual reorientation of all images; (3) coregistration of the reoriented images with Tl-weighted images. (4) normalization to the MNI152 standard space and resampling to a voxel size of 3 x 3 x 3 mm3; (5) regression of covariates, including those related to 6 head motions, white matter, and cerebrospinal fluid; (6) execution of temporal lowpass filtering within the range of 0.01-0.1 Hz; (7) the voxel-wise standard deviation of the blood-oxygen-level-dependent signal of each voxel was calculated; and (8) ninety standard deviation map for each participant were obtained using the AAL atlas.Fractional anisotropy(FA) map construction

[0099] For fractional anisotropy map construction, the FMRIB Software Library v6.0 (FSL) was employed. The preprocessing involved: (1) correction of eddy currents and movements in the raw DTI data using the eddy tool; (2) extraction of brain tissue data and removal of nonbrain tissue from the DTI dataset using the brain extraction tool; (3) creation of fractional anisotropy images by fitting the eddy-corrected data into a tensor model at each voxel; (4) registration and alignment of all fractional anisotropy images using the FMRIB58 FA standard-space image as the target; (5) normalization of all fractional anisotropy images to the MNI152 standard space; and (6) segmentation into 48 fractional anisotropy maps for each participant using the JHU-ICBM-Labels-lmm atlas.Feature identification

[0100] In one embodiment, the present invention performed a systematic approach to identify key features (i.e., voxels) that exhibit the strongest correlations with chronological age across different brain regions. This process was applied to all 228 maps, including 90 gray matter, 90 standard deviation, and 48 fractional anisotropy maps.

[0101] In one embodiment, to begin with, the present invention randomly choses 70% of the participants from the training dataset to calculate Pearson’s correlation coefficient (r) for each participant and assess the relationship between each voxel in different brain regions and chronological age.

[0102] In one embodiment, this step was iteratively conducted 1000 times to ensure robustness. In one embodiment, this step was iteratively conducted 100-500 times to ensure robustness. In one embodiment, this step was iteratively conducted 500-1000 times to ensure robustness. In one embodiment, this step was iteratively conducted 1000-1500 times to ensure robustness. In one embodiment, this step was iteratively conducted 1500-2000 times to ensure robustness.

[0103] Subsequently, in one embodiment, 50% of voxels exhibiting the highest r values with chronological age were identified in these 100-2000 trials. This process aimed to refine and pinpoint the most relevant key voxels within each brain region. In one embodiment, at least 20% of voxels exhibiting the highest r values with chronological age were identified in these 100- 2000 trials. In one embodiment, at least 30% of voxels exhibiting the highest r values with chronological age were identified in these 100-2000 trials. In one embodiment, at least 40% of voxels exhibiting the highest r values with chronological age were identified in these 100-2000 trials. In one embodiment, at least 60% of voxels exhibiting the highest r values with chronological age were identified in these 100-2000 trials. In one embodiment, at least 70% of voxels exhibiting the highest r values with chronological age were identified in these 100-2000 trials. In one embodiment, at least 80% of voxels exhibiting the highest r values with chronological age were identified in these 100-2000 trials.

[0104] In one embodiment, to further reinforce the selection process, the initial two steps were replicated 100 times, leading to the generation of 100 different sets of key voxels. In one embodiment, the initial two steps were replicated at least 20 times, leading to the generation of at least 20 different sets of key voxels. In one embodiment, the initial two steps were replicated at least 50 times, leading to the generation of at least 50 different sets of key voxels. In one embodiment, the initial two steps were replicated at least 200 times, leading to the generation of at least 200 different sets of key voxels. In one embodiment, the initial two steps were replicated at least 500 times, leading to the generation of at least 500 different sets of key voxels. In one embodiment, the initial two steps were replicated at least 1000 times, leading to the generation ofat least 1000 different sets of key voxels.

[0105] This repeated iteration was essential to ascertain the consistency and reliability of the selected voxels. As a result of this rigorous and comprehensive feature selection process, a set of key features was established for each of the 90 gray matter, 90 f standard deviation, and 48 fractional anisotropy maps. These features were then utilized to construct a brain age prediction model for different brain regions.Brain age prediction and BrainAGE calculation

[0106] Gaussian process regression (GPR) algorithm, in one embodiment, was employed to train 228 brain age prediction models. GPR, recognized for its efficacy in handling complex datasets, was chosen due to its demonstrated success in various previous studies. The models were constructed for 90 gray matter, 90 standard deviation, and 48 fractional anisotropy maps within the training dataset. To ensure the models' generalizability, a five-fold cross-validation method was implemented.

[0107] In one embodiment, the models were also tested on an independent dataset of 110 healthy controls to evaluate their reproducibility and stability and to predict the brain age of all participants. A correction formula was utilized to address potential biases in brain age prediction, such as underestimation in older participants and overestimation in younger participants. This formula involved a linear equation with a representing the slope and 0 the intercept:Corrected brain age = Brain age + [chronological age - (otxchronological age + 0)]

[0108] This equation was used to adjust the brain age predictions, reducing bias. The performance of the models was then assessed by calculating the mean absolute error (MAE) and the Pearson correlation coefficient (r) between the corrected brain ages and the actual chronological ages. Further, the BrainAGE for different brain regions of the study participants was calculated as the difference between the corrected brain age and the chronological age: BrainAGE = Corrected brain age - chronological ageStatistical analysis

[0109] ANCOVA was utilized to test the differences in BrainAGE between BD and healthy controls as well as MDD group and healthy controls, respectively. Demographic and cognitive assessments such as chronological age, sex, Mini-Mental State Examination (MMSE) scores, and years of education were covariates. To address the issue of multiple comparisons in the statistical analysis, the False Discovery Rate (FDR) method was employed. This method ispivotal in adjusting p-values to diminish Type I errors. The p-values were corrected to maintain a significance level of 0.05, which was the benchmark for determining statistical significance. Partial r|2values were calculated as effect size.ResultsGray matter map

[0110] Fig. 5 shows group differences in BrainAGE between participants with BD and healthy controls in 90 models for gray matter map. In the BD group analysis of 90 gray matter regions, 66 demonstrated significant accelerated aging. The top 20 regions exhibiting the most abnormal acceleration in aging included the bilateral superior frontal gyrus (dorsolateral), right inferior frontal gyrus (opercular part), right rolandic operculum, left supplementary motor area, left olfactory cortex, left superior frontal gyrus (medial), bilateral superior frontal gyrus (medial orbital), right gyrus rectus, left insula, left hippocampus, left amygdala, right superior parietal gyrus, bilateral caudate nucleus, bilateral thalamus, left Heschl gyrus, and the left temporal pole (superior temporal gyrus).

[0111] Fig. 6 shows group differences in BrainAGE between participants with MDD and healthy controls in 90 models for gray matter map. For the MDD group, significant accelerated aging was observed in 67 gray matter regions. The top 20 regions showing the most pronounced acceleration in aging within the MDD group included the left superior frontal gyrus (dorsolateral), bilateral superior frontal gyrus (orbital part), right rolandic operculum, left supplementary motor area, left olfactory cortex, left superior frontal gyrus (medial), bilateral superior frontal gyrus (medial orbital), bilateral gyrus rectus, bilateral insula, left anterior cingulate and paracingulate gyri, right caudate nucleus, right thalamus, bilateral Heschl gyrus, left superior temporal gyrus, and the left temporal pole (superior temporal gyrus).Standard deviation map

[0112] Fig. 7 shows group differences in BrainAGE between participants with BD and healthy controls in 90 models for standard deviation map. In the BD group, the results of 90 brain regions showed accelerated aging in 17 different brain regions. These regions included the left inferior frontal gyrus (orbital part), the right insula, the right posterior cingulate gyrus, the left hippocampus, the left parahippocampal gyrus, the bilateral cuneus, the right superior occipital gyrus, the right superior parietal gyrus, the left precuneus, the right caudate nucleus, the bilateral Heschl's gyrus, the right superior temporal gyrus, the bilateral temporal pole (middletemporal gyrus), and the left inferior temporal gyrus.

[0113] Conversely, the analysis did not reveal statistically significant accelerated brain aging in the MDD group after applying the FDR correction. Nonetheless, a trend toward significantly larger BrainAGE was observed, including the left cuneus, bilateral superior occipital gyrus, left middle occipital gyrus, right postcentral gyrus, left superior parietal gyrus, left precuneus, left caudate nucleus, and right thalamus.Fractional anisotropy map

[0114] Fig. 8 shows group differences in BrainAGE between participants with BD and healthy controls in 48 models for fractional anisotropy map. Out of 48 white matter tracts analyzed, 43 showed accelerated aging in BD patients. The top 20 most abnormal white matter tracts included the middle cerebellar peduncle, corpus callosum, fornix (column and body), bilateral corticospinal tract, right inferior cerebellar peduncle, right anterior corona radiata, left superior corona radiata, bilateral posterior corona radiata, bilateral posterior thalamic radiation, left sagittal stratum, left external capsule, bilateral fornix (cres) / stria terminalis, and bilateral tapetum.

[0115] Fig. 9 shows group differences in BrainAGE between participants with MDD and healthy controls in 48 models for fractional anisotropy map. In MDD patients, 39 out of 48 white matter tracts displayed accelerated aging. The top 20 most notably aged structures included the middle cerebellar peduncle, corpus callosum, fornix (column and body of fornix), bilateral corticospinal tract, right inferior cerebellar peduncle, bilateral retrolenticular part of internal capsule, left superior corona radiata, bilateral posterior corona radiata, bilateral posterior thalamic radiation, right sagittal stratum, right external capsule, right fornix (cres) / Stria terminalis, and bilateral tapetum.

[0116] Specific patterns of white matter tract aging were observed in different affective disorders. In the BD group, specific aging trajectory deviations were observed in the left external capsule and left fornix (cres)Zstria terminalis. In contrast, the MDD group showed excessive aging in the bilateral retrolenticular part and right sagittal stratum. These findings underscore the importance of identifying disease-specific deteriorated regions in structural brain imaging, which is crucial for a deeper understanding of the pathophysiological mechanisms of various affective disorders and their connection to brain structural changes.Discussion

[0117] The present invention found that the BD group exhibited accelerated aging in 66out of 90 gray matter regions. A similar trend was observed in the MDD group, with 67 regions affected. Regarding standard deviation mapnectivity, the BD group showed accelerated aging in 17 specific brain regions. In contrast, the MDD group demonstrated no significant deviation in brain aging trajectories across 90 brain regions after FDR correction. These findings suggest distinct neural mechanisms in MDD compared to BD. In terms of fractional anisotropy, individuals with BD exhibited significantly larger BrainAGE in 43 out of the 48 analyzed white matter tracts. Meanwhile, participants with MDD showed a significantly larger BrainAGE in 39 white matter tracts. These findings indicate disorder-specific deterioration patterns in BD and MDD, which are crucial for understanding the neuropathology of these affective disorders and for developing precise treatment strategies.

[0118] In the gray matter model, the results of the present invention indicated that BrainAGE was elevated in the majority of brain regions in the BD group. This suggests that patients with BD exhibit widespread signs of degeneration in brain structure, potentially attributable to the disease. One BD study found that participants who had not been treated with lithium, the brain age exceeded their chronological age by an average of 4.28 ± 6.33 years. Conversely, participants receiving lithium treatment exhibited similar trends between brain age and chronological age, with a mean difference of 0.48 ± 7.60 years. A significantly larger BrainAGE was observed in participants with BD compared to healthy controls. In one study, 459 individuals with BD were employed to predict brain age. The findings indicated that, on average, BrainAGE of BD patients was approximately two years older compared to that of healthy controls. In contrast, another study demonstrated individual with BD in initial phase had no significantly larger BrainAGE. Moreover, it had been revealed that no significant differences were observed between brain age and chronological age of individuals with BD. In a sperate study, it has been demonstrated that BrainAGE were significantly observed in individuals with schizophrenia, but not in individuals with BD. The results of previous research regarding the brain age of bipolar disorder (BD) patients demonstrate inconsistent findings concerning whether they had larger brain ages. One explanation for this inconsistency is that these existing studies predominantly utilized data encompassing the whole brain for modeling. As such, existing methods may not accurately identify the degeneration in specific brain regions, contributing to the varied results. The present invention adopts a brain age prediction approach focusing on differentiated brain regions. This approach not only identified the brain regions exhibiting accelerated aging in BD patients but alsohad the potential to act as a biomarker, facilitating the development of future therapeutic strategies.

[0119] Approximately two-thirds of brain regions in the MDD group exhibited accelerated brain aging. A large-scale ENIGMA multisite replication study found that the estimated brain age of individuals with MDD was significantly higher than their chronological age, exceeding it by one year. These findings were consistent with results published by the authors in previous reports, which reported a BrainAGE of +1.08 years; and +2.78 years. Another study found that compared to healthy controls, the MDD group exhibited a significantly larger BrainAGE, with an average increase of 2.11 years. Notably, among male participants, a higher severity of depression was observed. These findings align closely with the results produced by the present invention. However, unlike prior research that solely focused on predicting the overall accelerated aging of the brain, the methodology of the present invention enables a detailed evaluation of degeneration in specific brain regions among individuals with MDD, particularly in the superior frontal gyrus and the temporal gyrus.

[0120] The present invention reveals that 13 out of the 20 brain regions exhibiting the most significant deterioration were common to both the BD and MDD groups. The affected brain regions included the left superior frontal gyrus (dorsolateral), right rolandic operculum, left supplementary motor area, left olfactory cortex, left superior frontal gyrus (medial), bilateral superior frontal gyrus (medial orbital), right gyrus rectus, left insula, right caudate nucleus, right thalamus, left Heschl gyrus, and the left temporal pole (superior temporal gyrus). These findings suggested that these regions may share similar neuropathological mechanisms underlying the observed degeneration. Conversely, the present invention also identified thatBD and MDD groups possessed different brain regions with accelerated aging. A meta-analysis disclosed that BrainAGE varied across mental disorders, identifying schizophrenia as exhibiting the largest BrainAGE, succeeded by BD, and then MDD. This observation suggests that different psychiatric disorders were characterized by differing degrees of brain degeneration.

[0121] Furthermore, the present invention extended this understanding by quantitatively assessing the disparities in degeneration across different brain regions in different affective disorders. Notably, individuals with BD exhibited pronounced aging in several key regions, including the right inferior frontal gyrus (operculum), the left hippocampus, the left amygdala, and the right superior parietal gyrus. These brain regions have also demonstrated consistent abnormalities in individuals with BD in previous brain imaging studies. In the MDD group, thepresent invention also identified specific structural abnormalities in various brain regions. These anomalies were notably presented in the bilateral superior frontal gyrus (orbital) and the left anterior cingulate and paracingulate gyri. These specific brain regions have also been previously identified as exhibiting significant abnormalities in brain imaging studies of MDD participants. These findings of the present invention not only highlighted the varied and complex nature of brain aging in affective disorders but also underscored the presence of both common and disorderspecific patterns of brain region aging in BD and MDD. This understanding is critical for precising treatments and interventions tailored to the unique neuropathological profiles of these disorders.

[0122] In the brain age models for the standard deviation map, the present invention revealed that there were 18 brain regions exhibiting signs of accelerated brain aging in the BD group; however, comparable results were not observed in the MDD group. In a related study accomplished, the present invention employed functional connectivity maps as features to develop and train brain age prediction models for individuals with schizophrenia. However, a previous study suggested that blood-oxygen-level-dependent variability demonstrated a spatially coherent pattern that is markedly distinct from the average of blood-oxygen-level-dependent and showed a robust relationship with age. Consequently, the current study utilized standard deviation maps to train brain age prediction models, successfully identifying brain regions that accelerated brain aging in the BD group.

[0123] However, blood-oxygen-level-dependent signals can be utilized to calculate various functional characteristics. In one embodiment, these diverse features are integrated to develop brain age prediction models, thereby enriching the present invention regarding how affective disorders influence the degeneration of different brain functions. This absence of significant findings suggested that MDD may involve different neural mechanisms than BD. The distinct nature of these findings in MDD emphasizes the complexity and heterogeneity of affective disorders, highlighting the necessity for personalized and disease-specific approaches in research and clinical treatment.

[0124] In the 48 models for fractional anisotropy maps, the present invention indicated that BD and MDD groups demonstrated deviation in brain aging trajectories across most white matter tracts. Additionally, among the top 20 white matter tracts identified for showing the most significant accelerated aging in both groups, 13 white matter tracts were shared between the BD and MDD groups. These shared tracts included the middle cerebellar peduncle, the genu of thecorpus callosum, the body of the corpus callosum, the splenium of the corpus callosum, fornix (column and body), bilateral corticospinal tract, the right inferior cerebellar peduncle, the left superior corona radiata, bilateral posterior corona radiata, bilateral posterior thalamic radiation, the right fornix (cres) / stria terminalis, and bilateral tapetum. The observed overlaps in white matter tracts between BD and MDD suggested a potential commonality in the microstructural alterations of white matter affected by these affective disorders. These findings indicated that BD and MDD might share similar underlying pathological processes. Further investigation is necessary to clarify the specific biological or neuropathological changes driving these alterations and to examine the therapeutic potential of targeting these tracts. Such research could open avenues for potentially mitigating the deterioration process in individuals with BD and MDD, offering new directions for treatment strategies. In addition, the present invention identified that the BD group exhibited a disease-specific accelerated aging in the right anterior corona radiata. In individuals with MDD, the present invention revealed that the bilateral retrolenticular part of internal capsule exhibited significantly larger BrainAGEs, a finding that was uniquely characteristic of MDD. These results further suggested that distinct pathological processes underlie these affective disorders. The specificity observed in white matter alterations between BD and MDD underscored the subtle differences in their neuropathologies. These results of the present invention enhanced understanding of white matter microstructural changes associated with each disorder. The present invention offered the potential for more personalized and effective treatments tailored to the specific neuroanatomical alterations observed in BD and MDD.

[0125] The strengths of the present invention lie in its utilization of multimodal MRI to construct brain age models for affective disorders. This approach enabled the present invention to delineate the degenerative trajectories of BD and MDD across various brain structures and functions by providing a quantitative index (i.e., BrainAGE). In one embodiment, the present invention offers a comprehensive understanding of the deterioration processes within the brain that are associated with these disorders, thereby contributing valuable insights into underlying neuropathologies of BD and MDD. In addition, the present invention developed brain age prediction models for different brain regions, which improved the understanding of the degenerative impacts of affective disorders on different brain regions to develop targeted brain regions and personalize the treatment to enhance the effectiveness.

[0126] In one embodiment, the present invention had also been applied to schizophreniapatients, constructing brain age prediction models for various brain regions and validating them across two cohorts. The consistency of the results across these cohorts underscored the reliability of this approach, highlighting its potential as a promising method in the assessment and treatment of psychiatric disorders.Conclusion

[0127] The present invention suggested specific commonalities and unique characteristics after conducting machine learning predictions and statistical analyses on the brain structure and function in individuals with affective disorders. In particular brain regions, accelerated aging was identified across two types of affective disorders, suggesting the potential for shared neurobi ologi cal mechanisms. However, each affective disorder also exhibited distinct patterns of neural degeneration in certain regions, which might potentially correlate with their respective clinical features. These findings are pivotal in advancing our understanding of the pathophysiology of affective disorders and developing more tailored treatment approaches, thereby enabling more effective, personalized treatments for individuals with affective disorders. Our results highlight the complexity of neuroanatomical and functional changes in affective disorders. Future research should further elucidate these findings' implications, particularly in the clinical management and treatment strategies for BD and MDD, while also considering the limitations of our current study and exploring broader applications.

[0128] The foregoing description of the exemplary embodiments of the invention has been presented only for the purposes of illustration and description and is not intended to be exhaustive or to limit the invention to the precise forms disclosed. Many modifications and variations are possible in light of the above teaching.

[0129] While there has been shown several and alternate embodiments of the present invention, it is to be understood that certain changes can be made as would be known to one skilled in the art without departing from the underlying scope of the invention as is discussed and set forth above and below including claims and drawings. Furthermore, the embodiments described above and claims set forth below are only intended to illustrate the principles of the present invention and are not intended to limit the scope of the invention to the disclosed elements.

[0130] References cited in the instant application, which may include patents, patent applications and various publications, are cited and discussed in the description of this invention.The citation and / or discussion of such references is provided merely to clarify the description of the present invention and is not an admission that any such reference is “prior art” to the invention described herein. All references cited and discussed in the description of this invention, are incorporated herein by reference in their entireties and to the same extent as if each reference was individually incorporated by reference.Table 1 details additional clinicodemographic information in the two datasets1 TAM1 Taiwan Aging and Mental Illness, BT Tri-Service General Hospital Beitou Branch, SCZ individuals with schizophrenia, HC healthy controls, MMSE Mini-Mental State Examination, PANSS Positive and Negative Syndrome Scale, CPZ chlorpromazine.2 independent t test, significance level = 0.05. 3bx2test, significance level = 0 05.4.cOnly 161 participants with schizophrenia had verified medication records in the TAMI cohort.Table 2. Model performances of 90 brain-age prediction models for GM map in training, HC, and BT-HC datasetsAbbreviations: GM, gray matter; HC, healthy control; MAE, mean absolute error; r: Pearson's correlation coefficient; TAMI, Taiwan Aging and Mental Illness; BT, Tri-Service General Hospital Beitou Branch. Table 3. Model performances of 90 brain-age prediction models for FC map in training, HC, andBT-HC datasetsAbbreviations: FC, functional connectivity; HC, healthy control; MAE, mean absolute error; r:Pearson's correlation coefficient; TAMI, Taiwan Aging and Mental Illness; BT, Tri-Service General Hospital Beitou Branch.Table 4. Model performances of 48 brain-age prediction models for FA map in training, HC, andBT-HC datasetsAbbreviations: FA, fractional anisotropy; HC, healthy control; MAE, mean absolute error; r:Pearson's correlation coefficient; TAMI, Taiwan Aging and Mental Illness; BT, Tri-Service General Hospital Beitou Branch. Table 5. Group differences in brain age gaps between participants with schizophrenia and HCs in90 models for GM map in the two cohorts.Abbreviations: GM, gray matter; SCZ, individuals with schizophrenia; HC, healthy control; SD, standard deviation; TAMI, Taiwan Aging and Mental Illness; BT, Tri-Service General Hospital Beitou Branch; FDR, false discovery rate.Significant differences after FDR correction are shown in red (adjusted P < 0.05).Table 6. Group differences in brain age gaps between participants with schizophrenia and HCs in90 models for FC map in the two cohorts.Abbreviations: FC, functional connectivity; SCZ, individuals with schizophrenia; HC, healthy control; SD, standard deviation; TAMI, Taiwan Aging and Mental Illness; BT, Tri-Service General Hospital Beitou Branch; FDR, false discovery rate.Significant differences after FDR correction are shown in red (adjusted P < 0.05).Table 7. Group differences in brain age gaps between participants with schizophrenia and HCs in48 models for FA map in the two cohorts.control; SD, standard deviation; TAMI, Taiwan Aging and Mental Illness; BT, Tri-ServiceGeneral Hospital Beitou Branch; FDR, false discovery rate.Significant differences after FDR correction are shown in red (adjusted P < 0.05).Table 8. Associations of brain age gaps with MMSE score in participants with schizophrenia in 90 models for GM map.schizophrenia; TAMI, Taiwan Aging and Mental Illness; BT, Tri-Service General Hospital Beitou Branch; FDR, false discovery rate.Significant level after FDR correction is shown in red (adjusted P < 0.05).Table 9. Associations of brain age gaps with MMSE score in participants with schizophrenia in90 models for FC map.Abbreviations: FC, functional connectivity; MMSE, Mini-mental state examination; SCZ, individuals with schizophrenia; TAMI, Taiwan Aging and Mental Illness; BT, Tri-Service General Hospital Beitou Branch; FDR, false discovery rate.Significant level after FDR correction is shown in red (adjusted P < 0.05).Table 10. Associations of brain age gaps with MMSE score in participants with schizophrenia in48 models for FA map.FA, fractional anisotropy; MMSE, Mini-mental state examination; SCZ, individuals with schizophrenia; TAMI, Taiwan Aging and Mental Illness; BT, Tri-Service General Hospital Beitou Branch; FDR, false discovery rate.Significant level after FDR correction is shown in red (adjusted P < 0.05)

Claims

CLAIMSWhat is claimed is:

1. A method of predicting brain age for a subj ect, the method comprising:5 obtaining at least one medical image of a brain of a subject; preprocessing the medical image to produce a brain map; segmenting the brain map into more than one brain regions; and calculating a brain age prediction of the subject based on a predetermined set of key features for each of the brain regions.

102. The method according to claim 1, wherein the subject has a mental health condition.

3. The method according to claim 1, wherein the predetermined set of key features are established using following steps:15 identifying a set of key features for each of the brain regions; and generating a brain age prediction model for each of the brain regions based on the set of key features of the corresponding brain region; wherein the predetermined set of key features are the set of key features adopted for generating the brain age prediction model.20 4. The method according to claim 1, wherein the at least one medical image comprises a T1 -weighted MRI image.

5. The method of claim 4, wherein the step of preprocessing the medical image comprises25 reorienting the T1 -weighted MRI image; normalizing the reoriented T1 -weighted MRI image.

6. The method of claim 5, wherein the step of preprocessing the medical image further comprises pre-segmenting the reoriented T1 -weighted MRI image into gray matter (GM),30 white matter(WM), and cerebrospinal fluid regions.

7. The method of claim 6, wherein the brain map comprises a GM map.

8. The method of claim 4, wherein the at least one medical image further comprises a resting-state fMRI image.

9. The method of claim 8, wherein the step of preprocessing the medical image comprises correcting the resting-state fMRI image; coregistering the corrected resting-state fMRI image with the T1 -weighted image; normalizing the corrected resting-state fMRI image; and resampling the normalized resting-state fMRI image.

10. The method of claim 8, wherein the brain map comprises a functional connectivity (FC) map.

11. The method of claim 8, wherein the at least one medical image further comprises a DTI image.

12. The method of claim 8, wherein the step of preprocessing the medical image comprises correcting the DTI image; extracting data of at least one brain tissue from the corrected DTI image; fitting the extracted data of the at least one brain tissue into a tensor model; creating a FA image based on the fitted data of the at least one brain tissue; normalizing the FA image; and segmenting the FA image to produce at least one FA map.

13. The method of claim 12, wherein the brain map comprises a fractional anisotropy (FA) map of WM tracts.

14. The method of claim 3, wherein each of the brain regions comprises a plurality of voxels.

15. The method of claim 14, wherein the set of key features comprises a selected set of key voxels.

16. The method of claim 14, wherein the step of identifying the set of key features comprises calculating a Pearson’s correlation coefficient (r) between each voxel of different brain regions and a chronological age of the subject; and selecting at least 20% of the voxels having highest r values as the set of key voxels.

17. The method of claim 16, wherein the step of identifying the set of key features further comprises repeating the step of calculating the r value and the step of selecting at least 20% of the voxels for at least three times, and identifying at least three sets of key voxels; selecting the set of key voxels which has been identified for at least two times as the predetermined set of key features for generating the brain age prediction model.

18. The method of claim 1, wherein the mental health condition is schizophrenia.

19. A non-transitory computer readable medium storing a program causing a computer to execute a process for determining and predicting a brain age of a subject having a mental health condition, the process comprising: obtaining at least one medical image of a brain of a subject; preprocessing the medical image to produce a brain map; segmenting the brain map into more than one brain regions; and calculating a brain age based on a predetermined set of key features for each of the brain regions.

20. The non-transitory computer readable medium according to claim 19, wherein the subject has a mental health condition.

21. The non-transitory computer readable medium according to claim 19, wherein the predetermined set of key features are established using following steps: identifying a set of key features for each of the brain regions; and generating a brain age prediction model for each of the brain regions based on the set of key features of the corresponding brain region; wherein the predetermined set of key features are the set of key features adopted for generating the brain age prediction model.

22. The non-transitory computer readable medium according to claim 19, wherein the at least one medical image comprises a Tl-weighted MRI image.

23. The non-transitory computer readable medium of claim 22, wherein the step of preprocessing the medical image comprises reorienting the Tl-weighted MRI image; normalizing the reoriented Tl-weighted MRI image.

24. The non-transitory computer readable medium of claim 23, wherein the step of preprocessing the medical image further comprises pre-segmenting the reoriented Tl- weighted MRI image into gray matter (GM), white matter(WM), and cerebrospinal fluid regions.

25. The non-transitory computer readable medium of claim 24, wherein the brain map comprises a GM map.

26. The non-transitory computer readable medium of claim 22, wherein the at least one medical image comprises a resting-state fMRI image.

27. The non-transitory computer readable medium of claim 26, wherein the step of preprocessing the medical image comprisescorrecting the resting-state fMRI image; coregistering the corrected resting-state fMRI image with the T1 -weighted image; normalizing the corrected resting-state fMRI image; and resampling the normalized resting-state fMRI image,28. The non-transitory computer readable medium of claim 26, wherein the brain map comprises a functional connectivity (FC) map.

29. The non-transitory computer readable medium of claim 26, wherein the at least one medical image comprises a DTI image.

30. The non-transitory computer readable medium of claim 26, wherein the step of preprocessing the medical image comprises correcting the DTI image; extracting data of at least one brain tissue from the corrected DTI image; fitting the extracted data of the at least one brain tissue into a tensor model; creating a FA image based on the fitted data of the at least one brain tissue; normalizing the FA image; and segmenting the FA image to produce at least one FA map.

31. The non-transitory computer readable medium of claim 30, wherein the brain map comprises a fractional anisotropy (FA) map of WM tracts.

32. The non-transitory computer readable medium of claim 19, wherein each of the brain regions comprises a plurality of voxels.

33. The non-transitory computer readable medium of claim 32, wherein the set of key features comprises a selected set of key voxels.

34. The non-transitory computer readable medium of claim 32, wherein the step of selecting the set of key features comprisescalculating a Pearson’s correlation coefficient (r) between each voxel of different brain regions; and selecting at least 20% of the voxels having highest r values as the set of key voxels.

35. The non-transitory computer readable medium of claim 34, wherein the step of selecting the set of key features further comprises repeating the step of calculating the r value and the step of selecting at least 20% of the voxels for at least three times, and identifying at least three sets of key voxels; selected the set of key voxels which has been identified for at least two times as the selected set of key features.

36. The non-transitory computer readable medium of claim 20, wherein the mental health condition is schizophrenia.

Images