Diagnostic system for major depression

Abstract

The invention belongs to the technical field of intelligent diagnosis of mental diseases, and particularly relates to a major depression diagnosis system. The system comprises an input module used for inputting indexes of a patient; wherein the indexes comprise at least one of the following information: a neurocognitive test, child trauma information, multi-modal neuroiconomics information and transcriptomics information; the prediction module is integrated with a machine learning model and is used for obtaining a prediction result about whether the patient suffers from the major depression or not according to the indexes; and the output module is used for outputting the prediction result. The system provided by the invention can accurately diagnose the major depression, can effectively assist doctors in diagnosing and screening the major depression, reduces the workload of workers, and has a very good application prospect.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent diagnosis of mental illness, and in particular relates to a major depression diagnosis system. Background Art

[0002] Major Depressive Disorder (MDD), also known as clinical depression, major depressive disorder, and unipolar depression, is a mental illness that is characterized by a depressed emotional state, low self-esteem, and loss of interest in activities that were previously enjoyed.

[0003] The most widely used diagnostic criteria for major depressive disorder are the fourth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV-TR) of the United States and the International Statistical Classification of Diseases and Related Health Problems (ICD-10) of the World Health Organization (ICD-10 refers to the disorder described in this article as recurrent depressive disorder). The latter is commonly used in European countries, while the former is commonly used in the United States and many other non-European countries. These diagnostic criteria all rely on doctors with professional experience to make a diagnosis, which is not conducive to the screening and diagnosis of MDD.

[0004] Artificial intelligence has many uses in the medical field and has been applied to the diagnosis of various diseases. For example, "CN202010170779.7 A knowledge graph-based intelligent diagnosis device and system for depression" has attempted to diagnose depression through machine learning methods. However, major depressive disorder is a complex clinical syndrome caused by the combined action of multiple etiological mechanisms. In order to accurately diagnose major depressive disorder, what kind of model algorithm and input features should be used is still an urgent problem to be solved in this field. Summary of the invention

[0005] In view of the problems of the prior art, the present invention provides a major depression diagnosis system, which aims to optimize model algorithms and input features to achieve artificial intelligence diagnosis of major depression.

[0006] A major depression diagnosis system, comprising:

[0007] An input module, for inputting patient indicators; wherein the indicators include at least one of the following information: neurocognitive testing, childhood trauma information, multimodal neuroimaging information, and transcriptomics information;

[0008] A prediction module, integrated with a machine learning model, for obtaining a prediction result of whether the patient suffers from major depression based on the indicator;

[0009] An output module is used to output the prediction result.

[0010] Preferably, the algorithm of the machine learning model is a random forest algorithm.

[0011] Preferably, the neurocognitive test includes at least one of the following tests: RVP test, PRM test, DMS test, SWM test, IED test, SOC test of Cambridge Automated Cognitive Test Battery.

[0012] Preferably, the childhood trauma information includes at least one of the following information: childhood emotional abuse, emotional neglect, physical abuse, physical neglect, and sexual abuse.

[0013] Preferably, the multimodal neuroimaging information includes structural brain imaging information and / or resting-state functional brain network information;

[0014] The structural brain image information includes at least one of the following information: gray matter volume of the left superior frontal gyrus, gray matter volume of the left lingual gyrus, gray matter volume of the left superior temporal gyrus, gray matter volume of the right inferior frontal gyrus of the right operculum, and gray matter volume of the right middle frontal gyrus;

[0015] The resting state functional brain network information includes at least one of the following information: edge analysis results, graph theory analysis results, and dynamic brain network analysis results;

[0016] Preferably, the edge analysis results include:

[0017]

[0018]

[0019]

[0020] And / or, the graph analysis results include at least one of the following information: local efficiency, clustering coefficient, assortativity, left superior frontal gyrus node efficiency, left triangular inferior frontal gyrus node efficiency, bilateral parahippocampal gyrus node efficiency, bilateral PoCG node efficiency, left IOG node efficiency, left superior parietal gyrus node efficiency, left superior temporal gyrus node efficiency, right cerebellum IV-V area node efficiency, left MFG node efficiency, right tLP node efficiency, left tVL node efficiency, right PoCG node betweenness center, right superior parietal gyrus node betweenness center, cerebellum vermis I-II area node betweenness center;

[0021] And / or, the dynamic brain network analysis results include at least one of the following information: the state proportion of state 1, the state proportion of state 5, the state proportion of state 8, the state proportion of state 9, the state proportion of state 10, the average residence time of state 4, the average residence time of state 10, the average residence time of state 5, and the average residence time of state 8.

[0022] Preferably, the transcriptomic information includes coefficients of a difference module obtained after WGCNA analysis of mRNA, and the functions of the difference module include:

[0023]

[0024]

[0025] Preferably, the indicators include: childhood trauma, visual memory, the edge between the left inferior occipital gyrus and the right parahippocampal gyrus, the degree center of the left inferior occipital gyrus, the gray matter volume of the left inferior frontal gyrus, the gray matter volume of the right orbital inferior frontal gyrus, the edge between the left inferior occipital gyrus and the right rectus muscle, visual delayed memory, the edge between the left inferior occipital gyrus and the middle of the right orbital frontal lobe, and the edge between the left cuneus and the left central operculum.

[0026] The present invention also provides a computer-readable storage medium, on which is stored: a computer program for implementing the above-mentioned major depression diagnosis system.

[0027] The present invention provides an artificial intelligence diagnosis system for major depression, and provides a preferred model algorithm and input features, which can achieve good diagnostic performance. The system of the present invention can effectively assist doctors in diagnosing and screening major depression, and reduce manual workload. Therefore, the present invention has a good application prospect.

[0028] Obviously, according to the above contents of the present invention, in accordance with common technical knowledge and customary means in the art, without departing from the above basic technical ideas of the present invention, other various forms of modification, replacement or change may be made.

[0029] The following is a further detailed description of the above contents of the present invention through specific implementation methods in the form of embodiments. However, this should not be understood as the scope of the above subject matter of the present invention being limited to the following examples. All technologies implemented based on the above contents of the present invention belong to the scope of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 This is a schematic diagram of the process of Example 1. DETAILED DESCRIPTION

[0031] It should be noted that the algorithms of data collection, transmission, storage and processing steps not specifically described in the embodiments, as well as the hardware structure, circuit connection, etc. not specifically described can all be implemented through the contents disclosed in the prior art.

[0032] Example 1 Major Depression Diagnosis System

[0033] The major depression diagnosis system of this embodiment includes:

[0034] An input module, used to input the patient's indicators;

[0035] A prediction module, integrated with a random forest model, is used to obtain a prediction result of whether the patient suffers from major depression according to the indicator;

[0036] An output module is used to output the prediction result.

[0037] Among them, the indicators include: neurocognitive testing, childhood trauma information, multimodal neuroimaging information, and transcriptomic information;

[0038] The neurocognitive tests include: RVP test, PRM test, DMS test, SWM test, IED test and SOC test of the Cambridge Neuro Psychological Automated Test Battery (CANTAB).

[0039] The childhood trauma information includes: childhood emotional abuse, emotional neglect, physical abuse, physical neglect, and sexual abuse.

[0040] The multimodal neuroimaging information includes structural brain imaging information and resting-state functional brain network information;

[0041] The structural brain imaging information includes: the gray matter volume of the left superior frontal gyrus, the gray matter volume of the left lingual gyrus, the gray matter volume of the left superior temporal gyrus, the gray matter volume of the right inferior frontal gyrus of the right operculum, and the gray matter volume of the right middle frontal gyrus;

[0042] The resting state functional brain network information includes: edge analysis results, graph theory analysis results and dynamic brain network analysis results;

[0043] The edge analysis results include:

[0044]

[0045]

[0046]

[0047] The graph analysis results include at least one of the following information: local efficiency, clustering coefficient, assortativity, left superior frontal gyrus node efficiency, left triangular inferior frontal gyrus node efficiency, bilateral parahippocampal gyrus node efficiency, bilateral PoCG node efficiency, left IOG node efficiency, left superior parietal gyrus node efficiency, left superior temporal gyrus node efficiency, right cerebellum IV-V area node efficiency, left MFG node efficiency, right tLP node efficiency, left tVL node efficiency, right PoCG node betweenness center, right superior parietal gyrus node betweenness center, cerebellum vermis I-II area node betweenness center;

[0048] The dynamic brain network analysis results include at least one of the following information: the state proportion of state 1, the state proportion of state 5, the state proportion of state 8, the state proportion of state 9, the state proportion of state 10, the average residence time of state 4, the average residence time of state 10, the average residence time of state 5, and the average residence time of state 8.

[0049] The transcriptomic information includes coefficients of difference modules obtained after WGCNA analysis of mRNA, and the difference modules include:

[0050]

[0051]

[0052] The technical solution of the present invention is further illustrated by experiments below.

[0053] Experimental Example 1: Exploration of the pathogenesis of major depressive disorder based on multimodal neuroimaging and transcriptomics

[0054] 1. Experimental Methods

[0055] like Figure 1As shown in the figure, 195 MDD patients and 180 healthy controls (HC) were included. The 17-item Hamilton Depression Scale (HAMD) and Hamilton Anxiety Scale (HAMA) were used to assess the severity of the subjects' recent depression and anxiety symptoms, respectively. Six tasks (RVP test, PRM test, DMS test, SWM test, IED test and SOC test) of the Cambridge Neuro Psychological Automated Test Battery (CANTAB) were used to assess the subjects' neurocognitive function. Childhood Trauma Questionaries (CTQ) were used to assess the emotional neglect, emotional abuse, physical neglect, physical abuse and sexual abuse experienced by the subjects before the age of 18. All subjects completed head structural MRI and resting functional MRI scans, extracted the subjects' gray matter volume (GMV), and the resting-state brain network edges, topological properties and dynamic brain network indicators calculated by the hidden Markov model (HMM) based on the AAL3 brain atlas. In addition, peripheral venous blood was collected for the detection of transcriptomic indicators, and weighted gene co-expression network analysis (WGCNA) was used to find modules related to MDD disease. The Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) databases were used to enrich the functions of the differential modules, and protein-protein interaction (PPI) analysis was further used to find key (Hubgene, Hub) genes. Finally, the two groups were compared to find differences in neurocognition, childhood trauma, multimodal brain imaging, and transcriptomics between MDD patients and HCs, and a random forest classification model was used to integrate multidimensional indicators for diagnosis and prediction of MDD. All indicators of differences were corrected for FDR multiple comparisons, and q<0.05 indicated that the difference was statistically significant.

[0056] 2. Experimental Results

[0057] 1. Neurocognitive test: Compared with HC, MDD patients had higher reaction sensitivity in RVP test (F=5.160, q=0.035), lower correct hit rate (F=21.836, q<0.001), and longer average reaction time (F=15.199, q<0.001); lower delayed correct percentage in PRM test (F=15.199, q<0.001), and lower total correct response percentage in DMS test (F=3 2.369, q<0.001); in the SWM test, the total number of errors (F=14.532, q<0.001), the number of errors between groups (F=18.514, q<0.001) and the strategy score (F=15.273, q<0.001) were higher; in the IED test, the total number of errors (F=6.822, q=0.016) was higher; in the SOC test, the number of completed steps within the minimum number of steps (F=22.539, q<0.001) was lower. This suggests that the MDD patient group has poor sustained attention, delayed visual memory, spatial working memory, inhibition, transformation and planning abilities.

[0058] 2. Childhood trauma: The rates of childhood emotional abuse, emotional neglect, physical abuse, physical neglect, sexual abuse, and exposure to at least one type of trauma in MDD patients were 19.33%, 36.69%, 13.33%, 40.00%, 11.79%, and 55.33%, respectively, which were higher than those in HC (0.67%, 6.11%, 2.22%, 9.33%, 4.67%, and 22.67%). The q values ​​were all less than 0.05, and the differences were statistically significant.

[0059] 3. Multimodal neuroimaging:

[0060] (1) In terms of structural brain imaging, the GMV of the left superior frontal gyrus (t=5.049, P=0.017), left lingual gyrus (t=5.227, P=0.002), left superior temporal gyrus (t=5.750, P=0.001), right inferior frontal gyrus of the operculum (t=5.118, P=0.011), and right middle frontal gyrus (MFG) (t=5.159, P=0.015) in the MDD group were significantly lower than those in the HC group (FWE correction).

[0061] (2) Resting-state functional brain networks

[0062] ①The results of edge analysis showed that compared with HC, MDD patients had 41 weakened edges, including 33 nodes (P=0.0409, NBS corrected). The nodes with the largest number of abnormal edges were the left posterior central gyrus (PoCG), the left inferior occipital gyrus (IOG), and the right middle occipital gyrus (MOG). 26 enhanced edges included 24 nodes (P=0.0339, NBS corrected). The nodes with the largest number of abnormal edges were the left ventrolateral thalamic nucleus (tVL) and the right posterior thalamic nucleus (tLP).

[0063] ② The results of graph theory analysis showed that the local efficiency (t=-3.570, P<0.001), clustering coefficient (t=-2.796, P=0.005) and assortativity (t=-2.411, P=0.016) of MDD patients were significantly lower than those of HC. In terms of node attributes, after FDR correction, it was found that the left superior frontal gyrus, left triangular frontal gyrus, bilateral parahippocampal gyrus, bilateral PoCG, left IOG, left superior parietal gyrus, left superior temporal gyrus, right cerebellum IV of MDD patients were significantly lower than those of HC. -V area node efficiency was significantly decreased, while the node efficiency of the left MFG, right tLP, and left tVL increased; the node degree of the left triangular inferior frontal gyrus, bilateral PoCG, left temporal pole, and right cerebellar IV-V areas was decreased, while the node efficiency of the left MFG, right middle and lateral nucleus of the thalamus, left tVL, and left tLP increased; the node betweenness center of the right PoCG, right superior parietal gyrus, and cerebellar vermis I-II areas was decreased, all q<0.05, indicating that the global and local topological properties of the brain of MDD patients were impaired.

[0064] ③The results of dynamic brain network analysis showed that compared with HC, the proportion of states 1, 5, 8, and 9 in the MDD group increased significantly, and the proportion of state 10 decreased significantly; the average residence time in states 4 and 10 decreased significantly, and the average residence time in states 5 and 8 increased significantly, all q<0.05, indicating that the dynamic changes of the resting-state brain network in MDD patients were abnormal.

[0065] 4. Transcriptomics: After high-throughput sequencing, a total of 61,806 genes were measured, and after screening, the number of genes included in the analysis was 10,769. Using WGCNA, a total of 21 gene modules were identified, and 3 modules were significantly associated with MDD, of which the yellow module contained 1,183 genes (r=0.21, P=0.0006), the green module contained 858 genes (r=-0.17, P=0.005), and the light green module contained 51 genes (r=0.18, P=0.003).

[0066] Through GO and KEGG enrichment analysis, the yellow module was enriched with 88 BP entries, 23 CC entries, 11 MF entries and 9 KEGG pathways, mainly involving the stress-activated protein kinase signaling cascade (SAPK) signaling pathway (Enrichment = 0.150, q < 0.001), pattern recognition receptor signaling pathway (Pattern Recognition Receptor Signaling Pathway, PRR) signaling pathway (Enrichment=0.160, q<0.001), Toll-like receptor signaling pathway (Enrichment=0.187, q<0.001), IL-6 production process (Enrichment=0.142, q=0.003), Notch signaling pathway (Enrichment=0.220, q=0.024), NF-κB signaling pathway (Enrichment=0.054, q=0.041), Th17 cell differentiation (Enrichment=0.157, q=0.025), herpes virus infection (Enrichment=0.134, q=0.025) and other inflammatory immune process related pathways. The green module was enriched with 59 BP entries, 25 CC entries, 18 MF entries and 11 KEGG pathways, mainly involving biological regulation-related processes such as RNA generation (Enrichment=0.231, q<0.001), mitochondrial gene expression (Enrichment=0.209, q<0.001), and cellular respiration (Enrichment=0.117, q<0.001), as well as pathways related to neurodegenerative diseases and metabolic diseases such as Parkinson's disease (Enrichment=0.102, q=0.002), Huntington's disease (Enrichment=0.088, q=0.015), non-alcoholic fatty liver disease (Enrichment=0.103, q=0.034), and diabetic cardiomyopathy (Enrichment=0.108, q=0.003).

[0067] PPI analysis found three Hub genes in the yellow module, namely KPNB1, KPNA4, and PSEN1. These three genes are highly expressed in brain tissue; five Hub genes were found in the green module, namely RPS15A, RPS21, RPL27A, RPL5, and RPL7A. These five genes are also expressed in brain tissue.

[0068] The indicators of the transcriptomics input model are the coefficients of the differential modules obtained after WGCNA analysis of mRNA.

[0069] 5. Prediction of diagnostic results by random forest model: According to the system described in Example 1, multi-dimensional indicators including multimodal neuroimaging, childhood trauma, neurocognition and transcriptomics were integrated, and the random forest model was used to diagnose MDD, with a diagnostic accuracy of 77.99%, a sensitivity of 71.24% and a specificity of 84.78%. The top 10 indicators in terms of importance were childhood trauma, visual memory, the edge between the left inferior occipital gyrus and the right parahippocampal gyrus, the degree center of the left inferior occipital gyrus, the gray matter volume of the left inferior frontal gyrus, the gray matter volume of the right orbital inferior frontal gyrus, the edge between the left inferior occipital gyrus and the right rectus muscle, visual delayed memory, the edge between the left inferior occipital gyrus and the middle of the right orbital frontal lobe, and the edge between the left cuneus and the left central operculum.

[0070] It can be seen from the above embodiments and experimental examples that the system provided by the present invention can accurately diagnose major depression, can effectively assist doctors in diagnosing and screening major depression, reduce manual workload, and has good application prospects.

Claims

1. A major depression diagnosis system, characterized in that: include: An input module, for inputting patient indicators; wherein the indicators include at least one of the following information: neurocognitive testing, childhood trauma information, multimodal neuroimaging information, and transcriptomics information; A prediction module, integrated with a machine learning model, for obtaining a prediction result of whether the patient suffers from major depression based on the indicator; An output module is used to output the prediction result.

2. The major depression diagnosis system according to claim 1, characterized in that: The algorithm of the machine learning model is the random forest algorithm.

3. The major depression diagnosis system according to claim 1, characterized in that: The neurocognitive test includes at least one of the following tests: RVP test, PRM test, DMS test, SWM test, IED test, SOC test of Cambridge Automated Cognitive Function Test Set.

4. The major depression diagnosis system according to claim 1, characterized in that: The childhood trauma information includes at least one of the following information: childhood emotional abuse, emotional neglect, physical abuse, physical neglect, and sexual abuse.

5. The major depression diagnosis system according to claim 1, characterized in that: The multimodal neuroimaging information includes structural brain imaging information and / or resting-state functional brain network information; The structural brain image information includes at least one of the following information: gray matter volume of the left superior frontal gyrus, gray matter volume of the left lingual gyrus, gray matter volume of the left superior temporal gyrus, gray matter volume of the right inferior frontal gyrus of the right operculum, and gray matter volume of the right middle frontal gyrus; The resting-state functional brain network information includes at least one of the following information: edge analysis results, graph theory analysis results, and dynamic brain network analysis results.

6. The major depression diagnosis system according to claim 5, characterized in that: The edge analysis results include: And / or, the graph analysis results include at least one of the following information: local efficiency, clustering coefficient, assortativity, left superior frontal gyrus node efficiency, left triangular inferior frontal gyrus node efficiency, bilateral parahippocampal gyrus node efficiency, bilateral PoCG node efficiency, left IOG node efficiency, left superior parietal gyrus node efficiency, left superior temporal gyrus node efficiency, right cerebellum IV-V area node efficiency, left MFG node efficiency, right tLP node efficiency, left tVL node efficiency, right PoCG node betweenness center, right superior parietal gyrus node betweenness center, cerebellum vermis I-II area node betweenness center; And / or, the dynamic brain network analysis results include at least one of the following information: the state proportion of state 1, the state proportion of state 5, the state proportion of state 8, the state proportion of state 9, the state proportion of state 10, the average residence time of state 4, the average residence time of state 10, the average residence time of state 5, and the average residence time of state 8.

7. The major depression diagnosis system according to claim 1, characterized in that: The transcriptomic information includes the coefficients of the difference module obtained after WGCNA analysis of mRNA, and the functions of the difference module include:

8. The major depression diagnosis system according to any one of claims 1 to 7, characterized in that: The indicators include: childhood trauma, visual memory, the edge between the left inferior occipital gyrus and the right parahippocampal gyrus, the degree center of the left inferior occipital gyrus, the gray matter volume of the left inferior frontal gyrus, the gray matter volume of the right orbital inferior frontal gyrus, the edge between the left inferior occipital gyrus and the right rectus muscle, visual delayed memory, the edge between the left inferior occipital gyrus and the middle of the right orbital frontal lobe, and the edge between the left cuneus and the left central operculum.

9. A computer-readable storage medium, characterized in that: Stored thereon is: a computer program for implementing the major depression diagnosis system described in claims 1-8.

Images