Teenager mental symptom development early warning method and system based on cognitive evaluation

Abstract

The invention discloses a teenager mental symptom development early warning method and system based on cognitive evaluation, relates to the technical field of biomedical engineering, and solves the problems that the prior art is based on specific crowd or small sample data, lacks large-scale queue verification and is poor in reliability. The technical problems that the universality and the stability of the model are insufficient, and the model is difficult to be widely applied to people with different regions, races and cultural backgrounds are solved. The method comprises the following steps: acquiring cognitive function evaluation data, brain nuclear magnetic image data and a mental symptom scale of a user; taking the cognitive function evaluation data, the brain nuclear magnetic image data and the mental symptom scale as input data, and performing key feature screening on the input data; obtaining a target model and training the target model through the input data after key feature screening; obtaining the mental symptom type of the user through the target model; the method avoids invasive operation and is suitable for large-scale screening.

Description

Technical Field

[0001] The present invention belongs to the field of biomedical engineering, relates to sensor technology, and specifically is a method and system for warning of the development of adolescent mental symptoms based on cognitive assessment. Background Art

[0002] The NIH Toolbox includes tests in multiple fields such as cognition, sensation, movement, and emotion, such as language, memory, executive function, working memory, processing speed, audition, vision, taste, smell, and pain; it also includes questionnaires for assessing psychological well-being, social relationships, stress, self-efficacy, and negative emotions. The scores obtained from seven core tasks evaluated by the NIH Toolbox are integrated into seven valid and reliable psychometric indexes at the ABCD study baseline, which include: Picture Vocabulary Test score, Oral Reading Recognition Test score, Flanker Inhibition Control and Attention Test score, List Sorting Working Memory Test score, Dimensional Change Card Sorting Test score, Pattern Comparison Processing Speed Test score, and Picture Sequence Memory Test score. Among them, the scores of the Picture Vocabulary Test and the Oral Reading Recognition Test are added together to calculate the Crystallized Intelligence Composite Score for measuring the user's crystallized intelligence level. And the scores of the Flanker Inhibition Control and Attention Test, List Sorting Working Memory Test, Dimensional Change Card Sorting Test, Pattern Comparison Processing Speed Test, and Picture Sequence Memory Test are added together to calculate the Fluid Intelligence Composite Score for measuring the user's fluid intelligence level, which mainly reflects the user's executive function;

[0003] Existing technologies mostly rely on single-dimensional data, such as behavior scales, and cannot comprehensively reflect the complexity of adolescent mental health, resulting in an average low prediction accuracy and being difficult to meet the actual needs; traditional statistical methods are difficult to capture the non-linear relationship between cognitive function and symptom trajectories, such as the inverted U-shaped relationship, which limits the understanding of the development mechanism of mental health problems; existing technologies are based on specific populations or small-sample data and lack large-scale cohort verification, resulting in technical problems such as insufficient universality and stability of the model and being difficult to be widely applied to populations with different geographical, ethnic, and cultural backgrounds.

[0004] The present invention provides a method and system for warning of the development of adolescent mental symptoms based on cognitive assessment to solve the above technical problems. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art; for this purpose, the present invention proposes a method and system for warning of the development of adolescent mental symptoms based on cognitive assessment, which is used to solve the technical problems that existing technologies are based on specific populations or small-sample data and lack large-scale cohort verification, resulting in insufficient universality and stability of the model and being difficult to be widely applied to populations with different geographical, ethnic, and cultural backgrounds.

[0006] To achieve the above object, a first aspect of the present invention provides a warning method for the development of adolescent mental symptoms based on cognitive assessment, including:

[0007] Obtain the cognitive function indicators of the user through a standardized tool, and obtain the cognitive function assessment data of the user based on the cognitive function indicators; obtain the structural magnetic resonance image of the user's brain through a scanner, preprocess the structural magnetic resonance image, and obtain the brain nuclear magnetic resonance image data based on the preprocessed structural magnetic resonance image; obtain the mental symptom scale through the CBCL technology; wherein, the cognitive function assessment data includes: the comprehensive score of crystallized intelligence and the comprehensive score of fluid intelligence;

[0008] Use the cognitive function assessment data, the brain nuclear magnetic resonance image data, and the mental symptom scale as input data, and perform key feature screening on the input data; obtain the target model and train the target model with the input data after key feature screening;

[0009] Obtain the type of mental symptoms of the user through the target model; send a corresponding warning signal according to the type of mental symptoms of the user; wherein, the type of mental symptoms of the user includes: users with symptom remission and users with late-onset symptoms.

[0010] Preferably, the obtaining of the cognitive function assessment data of the user based on the cognitive function indicators includes:

[0011] Retrieve a third-party standardized tool to obtain the cognitive function indicators of the user; wherein, the cognitive function indicators include: the score of the picture vocabulary test, the score of the oral reading recognition test, the score of the flanker inhibition and attention test, the score of the list sorting working memory test, the score of the dimensional change card sorting test, the score of the pattern comparison processing speed test, and the score of the picture sequence memory test;

[0012] Add the score of the picture vocabulary test and the score of the oral reading recognition test to obtain the comprehensive score of crystallized intelligence;

[0013] Sum the scores of the flanker inhibition control and attention test, the list sorting working memory test, the dimensional change card sorting test, the pattern comparison processing speed test, and the picture sequence memory test to obtain the comprehensive score of fluid intelligence.

[0014] The present invention constructs a comprehensive score of crystallized intelligence to measure the level of crystallized intelligence of the user, and a comprehensive score of fluid intelligence to measure the level of fluid intelligence of the user, which mainly reflects the executive function of the user.

[0015] Preferably, the obtaining of the structural magnetic resonance image of the user's brain through a scanner includes:

[0016] Retrieve high-resolution and sequence-weighted imaging data of the user's brain using a 3T scanner; wherein, the imaging data includes: images T1w of the T1-weighted sequence and images T2w of the T2-weighted sequence, and the T1-weighted sequence and the T2-weighted sequence are isotropic within a set distance.

[0017] Use fast gradient echo technology and field map scanning to correct the inhomogeneity of the B0 field in the imaging data to obtain image data in DICOM format.

[0018] Automatically unpack the DICOM format image data and convert it into a compressed volume file to obtain a structural magnetic resonance image.

[0019] The present invention can obtain high-resolution imaging data of the user's brain by using a 3T scanner. Compared with low-field scanners, the 3T scanner has a higher signal-to-noise ratio and image contrast, can more clearly display the fine structures of the brain, and provides a key basis for clinical scenarios such as the diagnosis of neuropsychiatric diseases and the detection of brain structural abnormalities; by simultaneously acquiring images of the T1-weighted sequence and the T2-weighted sequence, the characteristics of brain tissue can be reflected from different angles.

[0020] Preferably, the preprocessing of the structural magnetic resonance image includes:

[0021] Perform gradient non-linear distortion correction on the structural magnetic resonance image through a specific non-linear transformation by the scanner.

[0022] Register the image T2w to the image T1w by the mutual information method, and perform rigid registration and resampling to match the reference brain in the standard space; estimate the smoothly varying B1 bias field and perform normalization processing by adjusting the structural magnetic resonance image to a preset brain white matter intensity standard value.

[0023] The present invention corrects the magnetic resonance image, estimates the smoothly varying B1 bias field, which is beneficial to quantifying and eliminating the differences in the intensity of the B1 field in different regions of the head, making the brain parenchyma signal more uniform, and correcting the image distortion caused by magnetic field inhomogeneity.

[0024] Preferably, the acquisition of brain nuclear magnetic resonance imaging data based on the preprocessed structural magnetic resonance image includes:

[0025] Reconstruct the automatic labeling of brain segmentation, cortical surface, and subcortical structures and surface-based non-linear registration through third-party software; calculate morphological indexes such as cortical thickness, area, volume, and sulcus depth, as well as image intensity indexes of the structural magnetic resonance image and cortical contrast, use an unsmoothed surface-based atlas, and calculate the average value for each cortical region.

[0026] Evaluate the severity of image artifacts and reconstruction inaccuracies, assign an overall QC score to each image reconstruction according to the severity, and determine whether the overall QC score is lower than a preset QC threshold; if so, mark the corresponding image as out of use; if not, mark the corresponding image as suitable for use; use the images marked as qualified as brain MRI image data.

[0027] It should be noted that QC represents Quality Control; the overall QC score is assigned by those skilled in the art according to the severity.

[0028] Through the qualified brain MRI image data at baseline and combined with longitudinal follow-up measurements, the present invention can monitor disease progression or treatment response.

[0029] Preferably, obtaining the mental symptom scale through the CBCL technology includes:

[0030] Obtain the t-scores of three CBCL evaluation indicators through the CBCL technology; among them, the three CBCL evaluation indicators include: internalizing problem evaluation indicator, externalizing problem evaluation indicator, and total problem evaluation indicator;

[0031] Mark users with a clinical score value of CBCL externalizing symptoms exceeding 60 points as the high externalizing symptoms group; mark users with a clinical score value of CBCL internalizing symptoms exceeding 60 points as the high internalizing symptoms group;

[0032] Mark users whose clinical thresholds of CBCL externalizing and internalizing symptom scores do not exceed 60 points as belonging to the control group;

[0033] Mark the high externalizing symptoms group, high internalizing symptoms group, and control group as mental symptom scale evaluation data.

[0034] It should be noted that CBCL (Child Behavior Checklist) represents a standardized assessment tool: Child Behavior Checklist; the higher the CBCL score, the more mental problems the corresponding user has.

[0035] The present invention can quantify the similarity of CBCL scores of the same subject in two independent evaluations by using test-retest reliability, identify and eliminate random errors caused by changes in the test environment and fluctuations in the subject's state, and improve data reliability; in multi-center studies or different cultural backgrounds, test-retest reliability verification can ensure the cross-environment applicability of the CBCL tool and avoid measurement biases caused by cultural differences or operational differences.

[0036] Preferably, the screening of key features from the input data includes:

[0037] S110: Clean, standardize, and align the collected cognitive function assessment data, brain MRI image data, and psychiatric symptom scales; remove the brain MRI image data with head movement exceeding 3 mm and the outliers with a combined fluid intelligence score lower than 70 points;

[0038] S120: Retrieve different types of models and train the models with the cognitive function assessment data, brain MRI image data, and psychiatric symptom scales; rank the importance of the features according to the set criteria and remove the features with the lowest ranking in the models;

[0039] S130: Repeat step S120 until the remaining number of features meets the preset feature quantity requirements.

[0040] It should be noted that users can, according to their own needs, select different models, set the number of features to be retained, and specify the evaluation criteria for feature importance.

[0041] The present invention eliminates the interference of motion artifacts on brain region feature extraction by removing the image data with head movement exceeding the threshold; the filtering criterion of a combined fluid intelligence score greater than or equal to 70 can exclude samples with severe cognitive impairment, which is beneficial to ensuring the stability of the training set data distribution.

[0042] Preferably, the obtaining of the target model and training the target model with the input data after key feature screening includes:

[0043] Call three algorithms: regularized logistic regression, linear support vector machine, and random forest, and train the model in the way of bagging ensemble learning to construct several identical base learners, and obtain the hyperparameters of the model by the majority voting method combining the base learners;

[0044] Optimize the hyperparameters of the model by using the 5-fold cross-validation method and the grid search algorithm; obtain the average AUC of the model through repeated cross-validation based on the hyperparameters; select the model with the highest average AUC as the target model, where the target models include: externalizing symptom remission model, externalizing symptom late-onset model, internalizing symptom remission model, and internalizing symptom late-onset model.

[0045] It should be noted that AUC (Area Under the Curve) is a model evaluation index.

[0046] The present invention combines them in the way of the majority voting method to improve the generalization ability of the model; create balanced training samples by randomly undersampling the majority class in the training samples and repeating cross-validation, which is beneficial to dealing with the problem of sample imbalance; by setting the range and interval of the number of base learners, it is beneficial to achieving the best balance between improving prediction accuracy and reducing computational cost.

[0047] Preferably, obtaining the user's mental symptom type through the target model includes:

[0048] Retrieve the target model, and input the obtained cognitive function assessment data, brain MRI image data, and mental symptom scale assessment into the trained corresponding target model; preset several data collection time points, and obtain target subgroups with different symptom trajectories through the target model; mark users in the target subgroup with symptom scores greater than or equal to the baseline symptoms as late-onset symptom users; mark users in the target subgroup with symptom scores less than the baseline symptoms as symptom remission users; wherein, the baseline symptom is: the symptom score obtained at the first time point of data collection.

[0049] To achieve the above object, the second aspect of the present invention provides an early warning system for the development of adolescent mental symptoms based on cognitive assessment, including: a data collection module, a model training module, and a prediction and early warning model;

[0050] Data collection module: Obtain the user's cognitive function indicators through a standardized tool, and obtain the user's cognitive function assessment data based on the cognitive function indicators; obtain the structural magnetic resonance image of the user's brain through a scanner, preprocess the structural magnetic resonance image, and obtain the brain MRI image data based on the preprocessed structural magnetic resonance image; obtain the mental symptom scale through the CBCL technology; wherein, the cognitive function assessment data includes: the comprehensive score of crystallized intelligence and the comprehensive score of fluid intelligence.

[0051] Model training module: Use the cognitive function assessment data, the brain MRI image data, and the mental symptom scale as input data, and perform key feature screening on the input data; obtain the target model and train the target model with the input data after key feature screening.

[0052] Prediction and early warning model: Obtain the user's mental symptom type through the target model; issue corresponding early warning signals according to the user's mental symptom type; wherein, the user's mental symptom type includes: symptom remission users and late-onset symptom users.

[0053] Compared with the prior art, the beneficial effects of the present invention are:

[0054] 1. The present invention incorporates multi-modal data such as clinical, cognitive, brain imaging, and genetic data into a unified framework for the first time, comprehensively reflecting the multi-dimensional characteristics of adolescents' mental health. By adopting an integrated learning algorithm, it can effectively capture the complex non-linear relationship between cognitive function and symptom trajectories, improving the fitting ability and prediction accuracy of the model for practical problems. Based on a cross-cohort weak classifier library and a similarity-weighted method, it integrates the advantages of multiple models, enhances the generalization ability and adaptability of the model, and ensures stability and reliability in different populations and scenarios. The present invention only requires non-invasive detection means such as baseline cognitive tests and resting-state MRI, avoiding invasive operations, being suitable for large-scale screening, and having good clinical application prospects and generalizability.

[0055] 2. The present invention combines cognitive function assessment data, brain MRI image data, and mental symptom scales. This method can comprehensively evaluate the mental state of users from multiple dimensions. The fusion of such multi-modal data helps to capture information that may be missed by a single data source, thereby improving the accuracy of mental symptom type recognition. By identifying the unique mental symptom types of users, it is beneficial to conduct personalized analysis according to the specific situation of each user, and this personalized analysis helps to provide more accurate and effective intervention measures or treatment plans for different users. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following described drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0057] Figure 1 It is a schematic flow chart for the early warning development of the present invention;

[0058] Figure 2 It is a schematic diagram of the specific steps for the preprocessing of structural magnetic resonance images in the present invention;

[0059] Figure 3 It is a schematic diagram of the specific steps for the model prediction of the present invention;

[0060] Figure 4 It is a schematic diagram of the module relationship included in the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0061] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0062] Please refer to Figure 1 , an embodiment of the first aspect of the present invention provides a method for predicting the development of adolescent mental symptoms based on cognitive assessment, including:

[0063] Obtain the cognitive function indicators of the user through a standardized tool, and obtain the cognitive function assessment data of the user based on the cognitive function indicators; obtain the structural magnetic resonance image of the user's brain through a scanner, preprocess the structural magnetic resonance image, and obtain brain nuclear magnetic resonance image data based on the preprocessed structural magnetic resonance image; obtain a mental symptom scale through the CBCL technology; wherein, the cognitive function assessment data includes: the comprehensive score of crystallized intelligence and the comprehensive score of fluid intelligence.

[0064] Use the cognitive function assessment data, brain nuclear magnetic resonance image data, and mental symptom scale as input data, and perform key feature screening on the input data; obtain a target model and train the target model with the input data after key feature screening.

[0065] Obtain the user's mental symptom type through the target model; send a corresponding warning signal according to the user's mental symptom type; wherein, the user's mental symptom type includes: symptom remission users and symptom late-onset users.

[0066] Please refer to Figure 2 , the specific steps of preprocessing the structural magnetic resonance image: retrieve the high-resolution and sequence-weighted imaging data of the user's brain by a 3T scanner; wherein, the imaging data includes: the image T1w of the T1-weighted sequence and the image T2w of the T2-weighted sequence, and the T1-weighted sequence and the T2-weighted sequence are isotropic within a set distance.

[0067] Use the fast gradient echo technique and field map scanning to correct the inhomogeneity of the B0 field in the imaging data to obtain DICOM format image data.

[0068] Automatically unpack the DICOM format image data and convert it into a compressed volume file to obtain the structural magnetic resonance image.

[0069] Perform gradient non-linear distortion correction on the structural magnetic resonance image and perform specific non-linear transformation through the scanner.

[0070] The image T2w is registered to the image T1w by the mutual information method, and rigid registration and resampling are performed to match the reference brain in the standard space; the smoothly varying B1 bias field is estimated and normalization is performed by adjusting the structural magnetic resonance image to a preset brain white matter intensity standard value.

[0071] For example, imaging data of high-resolution T1-weighted (distance: 1.0 mm, direction: isotropic) structural magnetic resonance imaging was obtained on a 3T scanner; a 3T MRI system was used to image the brains of children aged 9 to 10 years, and images of T1-weighted (T1w) and T2-weighted (T2w) sequences were acquired. Fast gradient echo technology and field map scanning were used to correct B0 field inhomogeneity; the DICOM format image data was transmitted to the data analysis and information center via the network, automatically unpacked and converted into a compressed volume file.

[0072] In the preprocessing stage, the T1w and T2w images were corrected for gradient non-linear distortion, and a specific non-linear transformation was performed using the scanner.

[0073] Steps for registering the T2w image to the T1w image by the mutual information method: Read the T1w image (as the fixed image) and the T2w image (as the moving image). The existing T1w image and T2w image are both grayscale images of 256×256 pixels. By statistical analysis of the histogram of the T1w image, its entropy [H(T1w)] was calculated to be 5.2, the entropy of the T2w image [H(T2w)] was 4.8, and the joint entropy [H(T1w, T2w)] was 7.5; the initial mutual information [I(T1w, T2w)] between the gray value distributions of the T1w and T2w images was calculated by the formula [I(T1w, T2w)] = H(T1w) + H(T2w) - H(T1w, T2w) = 5.2 + 4.8 - 7.5 = 2.5.

[0074] The position and orientation of the T2w image were adjusted by an optimization algorithm. When the translation parameters were (-3, 8) pixels and the rotation angle was 5 degrees, the new mutual information was calculated to be 3.2.

[0075] Steps for resampling to match the reference brain in the standard space: Determine the reference brain coordinate system: According to the target standard space used (such as the MNI space), determine its corresponding coordinate system and orientation; the voxel size of the reference brain in the existing standard space is 1×1×1 mm 3 , while the voxel size of the registered T2w image is 2×2×2 mm 3 ; The new resampling grid is marked as consistent with the reference brain. Then, for each voxel in the T2w image, the gray value corresponding to its position in the reference brain grid is calculated by cubic spline interpolation to complete the resampling.

[0076] Steps for estimating the smoothly varying B1 bias field and correcting image distortion: Establish a bias field model. Assume that the B1 bias field is a smoothly varying function. Denote the coordinates of each voxel in the T1w and T2w images as (x, y), and construct a two-dimensional quadratic polynomial model [f(x, y) = a0 + a1x + a2y + a3x 2 + a4xy + a5y2] to estimate the B1 bias field;

[0077] Through statistical analysis of the gray values of the T1w and T2w images, use the expectation maximization algorithm to iteratively solve the parameters [a0, a1, a2, a3, a4, a5]; when the parameters converge, obtain the bias field function; for each voxel in the T2w image, its corrected gray value is [I'(T1w, T2w)] = I(T1w, T2w) / f(x, y), thus correcting the image distortion caused by magnetic field inhomogeneity;

[0078] Steps for normalizing the image to a preset brain white matter intensity standard value: Determine the brain white matter intensity standard value: According to the research purpose and the characteristics of the dataset, set a target brain white matter intensity standard value of 110;

[0079] Within the brain white matter region of the corrected T2w image (which can be determined by automatic segmentation or manual delineation of the brain white matter mask), calculate the average gray value of the brain white matter;

[0080] Calculate the scaling factor scale through the formula scale = preset standard value / current brain white matter intensity mean;

[0081] Multiply the gray value of each voxel in the corrected T2w image by the scaling factor scale to adjust the brain white matter intensity to the preset standard value of 110, completing the normalization process.

[0082] Please refer to Figure 3 , for the specific steps of model prediction: Retrieve the target model, and input the obtained cognitive function assessment data, brain MRI data, and mental symptom scale assessment into the trained corresponding target model; preset several data collection time points, and obtain target subgroups with different symptom trajectories through the target model; mark users in the target subgroup with symptom scores greater than or equal to the baseline symptoms as late-onset symptom users; mark users in the target subgroup with symptom scores less than the baseline symptoms as symptom remission users; where the baseline symptom is: the symptom score obtained at the first time point of data collection.

[0083] For example, retrieve the target model, and input the obtained cognitive function assessment data, brain MRI image data, and mental symptom scale assessment into the corresponding trained target model; the preset time points for data collection are the 14-year-old, 16-year-old, 19-year-old, and 23-year-old of the to-be-tested users A, B, C, D, E, and F respectively; the symptom scores obtained at the first time point of data collection are: an externalizing symptom score of 7 points and an internalizing symptom score of 5 points;

[0084] Among them, for the to-be-tested users A, B, and C, through the mental symptom scale assessment data, the externalizing and internalizing symptom scores of users A, B, and C exceed 7 and 5 points respectively for the first time at the second, third, and fourth time points, and in subsequent time points, the externalizing symptom score is greater than or equal to 7 points or the internalizing symptom score is greater than or equal to 5 points. Then, users A, B, and C are marked as late-onset symptom users, and through the late-onset symptom target model, it is obtained that user A is a late-onset externalizing symptom user, and users B and C are late-onset internalizing symptom users, and corresponding warning signals are sent;

[0085] For the to-be-tested users D, E, and F, through the mental symptom scale assessment data, the initial externalizing and internalizing symptom scores exceed 7 and 5 points respectively, but drop below 7 and 5 points at the second, third, or fourth time point respectively, and in subsequent time points, the externalizing symptom score is lower than 7 points or the internalizing symptom score is lower than 5 points. Then, users D, E, and F are marked as symptom remission users, and through the symptom remission target model, it is obtained that users D and E are externalizing symptom remission users, and user F is an internalizing symptom remission user.

[0086] Please refer to Figure 4 , the second aspect embodiment of the present invention provides an early warning system for the development of adolescent mental symptoms based on cognitive assessment, including: a data collection module, a model training module, and a prediction and early warning model;

[0087] Data collection module: Obtain the cognitive function indicators of the user through a standardized tool, and obtain the cognitive function assessment data of the user based on the cognitive function indicators; obtain the structural magnetic resonance image of the user's brain through a scanner, preprocess the structural magnetic resonance image, and obtain the brain MRI image data based on the preprocessed structural magnetic resonance image; obtain the mental symptom scale through the CBCL technology; among them, the cognitive function assessment data includes: the comprehensive score of crystallized intelligence and the comprehensive score of fluid intelligence;

[0088] Model training module: Use the cognitive function assessment data, brain MRI image data, and mental symptom scale as input data, and perform key feature screening on the input data; obtain the target model and train the target model with the input data after key feature screening;

[0089] Prediction and early warning model: Obtain the types of users' mental symptoms through the target model; Send corresponding early warning signals according to the types of users' mental symptoms; Among them, the types of users' mental symptoms include: users with symptom remission and users with late-onset symptoms.

[0090] Some of the data in the above formula are calculated by removing the dimension and taking their numerical values. The formula is obtained by software simulation of a large amount of collected data to get a formula closest to the actual situation; The preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

[0091] The working principle of the present invention: The present invention obtains the cognitive function indicators of users through a standardization tool, and obtains the cognitive function evaluation data of users based on the cognitive function indicators; Obtain the structural magnetic resonance image of the user's brain through a scanner, preprocess the structural magnetic resonance image, and obtain brain nuclear magnetic resonance imaging data based on the preprocessed structural magnetic resonance image; Obtain a mental symptom scale through the CBCL technology; Use the cognitive function evaluation data, brain nuclear magnetic resonance imaging data, and mental symptom scale as input data, and perform key feature screening on the input data; Obtain a target model and train the target model with the input data after key feature screening; Obtain the types of users' mental symptoms through the target model; Send corresponding early warning signals according to the types of users' mental symptoms.

[0092] The above embodiments are only used to illustrate the technical method of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A warning method for the development of adolescent mental symptoms based on cognitive assessment, characterized in that, Including: Obtaining the cognitive function index of the user through a standardization tool, and obtaining the cognitive function evaluation data of the user based on the cognitive function index; Obtaining the structural magnetic resonance image of the user's brain through a scanner, preprocessing the structural magnetic resonance image, and obtaining brain nuclear magnetic resonance imaging data based on the preprocessed structural magnetic resonance image; Obtaining a mental symptom scale through the CBCL technology; wherein, the cognitive function evaluation data includes: the comprehensive score of crystallized intelligence and the comprehensive score of fluid intelligence; Taking the cognitive function evaluation data, the brain nuclear magnetic resonance imaging data, and the mental symptom scale as input data, and performing key feature screening on the input data; obtaining a target model and training the target model with the input data after key feature screening; Obtaining the type of mental symptoms of the user through the target model; sending a corresponding warning signal according to the type of mental symptoms of the user; wherein, the type of mental symptoms of the user includes: users with symptom remission and users with late-onset symptoms.

2. The early warning method for the development of adolescent mental symptoms based on cognitive assessment according to claim 1, wherein The obtaining of the cognitive function evaluation data of the user based on the cognitive function index includes: Obtaining the cognitive function index of the user; wherein, the cognitive function index includes: the score of the picture vocabulary test, the score of the oral reading recognition test, the score of the flanker inhibition and attention test, the score of the list sorting working memory test, the score of the dimensional change card sorting test, the score of the pattern comparison processing speed test, and the score of the picture sequence memory test; Adding the score of the picture vocabulary test and the score of the oral reading recognition test to obtain the comprehensive score of crystallized intelligence; Summing up the scores of the flanker inhibition control and attention test, the list sorting working memory test, the dimensional change card sorting test, the pattern comparison processing speed test, and the picture sequence memory test to obtain the comprehensive score of fluid intelligence.

3. The early warning method for the development of adolescent mental symptoms based on cognitive assessment according to claim 1, characterized in that, The obtaining of the structural magnetic resonance image of the user's brain through the scanner includes: Obtaining high-resolution and sequence-weighted imaging data of the user's brain through a scanner; wherein, the imaging data includes: the image T1w of the T1-weighted sequence and the image T2w of the T2-weighted sequence, and the T1-weighted sequence and the T2-weighted sequence are isotropic within a set distance; Using the fast gradient echo technology and field map scanning to correct the inhomogeneity of the B0 field in the imaging data to obtain image data in DICOM format; Automatically unpacking the DICOM format image data and converting it into a compressed volume file to obtain the structural magnetic resonance image.

4. The early warning method for the development of adolescent mental symptoms based on cognitive assessment according to claim 1, characterized in that, The preprocessing of the structural magnetic resonance image includes: Performing gradient non-linear distortion correction on the structural magnetic resonance image and performing specific non-linear transformation through a scanner; Registering the image T2w to the image T1w through the mutual information method, and performing rigid body registration and resampling to match the reference brain in the standard space; estimating the smoothly varying B1 bias field and normalizing by adjusting the structural magnetic resonance image to a preset brain white matter intensity standard value.

5. A warning method for the development of adolescent mental symptoms based on cognitive assessment according to claim 1, characterized in that, The obtaining of the brain nuclear magnetic resonance imaging data based on the preprocessed structural magnetic resonance image includes: Brain parcellation, cortical surface, automatic labeling of subcortical structures, and surface-based nonlinear registration were reconstructed using third-party software. Morphological indices of cortical thickness, area, volume, and sulcus depth were calculated, as well as image intensity indices of structural magnetic resonance images and cortical contrast, using unsmoothed surface-based atlases that were averaged for each cortical region. Assess the severity of image artifacts and inaccurate reconstruction, assign an overall QC score to each image reconstruction based on the severity, and determine whether the overall QC score is lower than the preset QC threshold; if yes, mark the corresponding image as discontinued; if no, mark the corresponding image as suitable for use; images marked as qualified are used as brain MRI data.

6. The early warning method for the development of adolescent mental symptoms based on cognitive assessment according to claim 1, wherein, The psychiatric symptom scale obtained by the CBCL technique includes: The t scores of three CBCL evaluation indicators were obtained through CBCL technology; the three CBCL evaluation indicators included: internalizing problem evaluation indicator, externalizing problem evaluation indicator and total problem evaluation indicator; Users whose clinical scores of externalizing symptoms on the CBCL exceed 60 points are marked as the high externalizing symptom group; users whose clinical scores of internalizing symptoms on the CBCL exceed 60 points are marked as the high internalizing symptom group; Users whose CBCL externalizing and internalizing symptom scores did not exceed the clinical threshold of 60 points were marked as belonging to the control group; The externalizing high symptom group, internalizing high symptom group, and control group were labeled as the Psychiatric Symptom Scale assessment data.

7. A warning method for the development of adolescent mental symptoms based on cognitive assessment according to claim 1, characterized in that The input data is screened for key features, including: S110: Clean, standardize and align the collected cognitive function assessment data, brain MRI data and psychiatric symptom scales; remove abnormal values of brain MRI data with head movement exceeding 3mm and fluid intelligence comprehensive scores below 70 points; S120: Retrieve different types of models and train the models using cognitive function assessment data, brain MRI data, and psychiatric symptom scales; rank the features according to the set criteria and remove the lowest-ranked features in the model; S130: Repeat step S120 until the number of remaining features meets the preset feature number requirement.

8. The early warning method for the development of adolescent mental symptoms based on cognitive assessment according to claim 1, characterized in that, The step of obtaining the target model and training the target model using the input data filtered by key features includes: Call the regularized logistic regression, linear support vector machine and random forest algorithms, use bagging ensemble learning to train the model, build several identical base learners, and obtain the model's hyperparameters by combining the base learners through majority voting. The hyperparameters of the model were optimized by using the 5-fold cross-validation method and the grid search algorithm. The average AUC of the model was obtained by repeated cross-validation based on the hyperparameters. The model with the highest average AUC was selected as the target model, where the target models included: externalizing symptom remission model, externalizing symptom late onset model, internalizing symptom remission model, and internalizing symptom late onset model.

9. The early warning method for the development of adolescent mental symptoms based on cognitive assessment according to claim 1, wherein The method of obtaining the user's mental symptom type through the target model includes: Retrieve the target model, and input the obtained cognitive function assessment data, brain MRI image data, and mental symptom scale assessment into the trained corresponding target model; preset several data collection time points, and obtain target subgroups with different symptom trajectories through the target model; mark users in the target subgroup with symptom scores greater than or equal to the baseline symptoms as users with late-onset symptoms; mark users in the target subgroup with symptom scores less than the baseline symptoms as users with symptom remission; where the baseline symptom is: the symptom score obtained at the first time point of data collection.

10. A warning system for the development of adolescent mental symptoms based on cognitive assessment, adapted to a method for warning the development of adolescent mental symptoms based on cognitive assessment according to any one of claims 1-9, characterized in that, It includes: a data collection module, a model training module, and a prediction and early warning model; The data collection module: obtains the cognitive function indicators of the user through a standardized tool, and obtains the cognitive function assessment data of the user based on the cognitive function indicators; obtains the structural magnetic resonance image of the user's brain through a scanner, preprocesses the structural magnetic resonance image, and obtains the brain MRI image data based on the preprocessed structural magnetic resonance image; obtains the mental symptom scale through the CBCL technology; where the cognitive function assessment data includes: the comprehensive score of crystallized intelligence and the comprehensive score of fluid intelligence; The model training module: uses the cognitive function assessment data, the brain MRI image data, and the mental symptom scale as input data, and performs key feature screening on the input data; obtains the target model and trains the target model with the input data after key feature screening; The prediction and early warning model: obtains the user's mental symptom type through the target model; issues a corresponding early warning signal according to the user's mental symptom type; where the user's mental symptom type includes: users with symptom remission and users with late-onset symptoms.

Images