Method and system for predicting curative effect of electroshock treatment for juvenile patients with severe depression

Abstract

The invention belongs to the technical field of medical image analysis, and particularly discloses a method and system for predicting the electroshock treatment effect of a teenager patient with severe depression, and the method comprises the steps: collecting the structural magnetic resonance imaging data of the teenager patient with depressive disorder before and after electroshock treatment ECT, and carrying out the preprocessing; obtaining brain grey matter morphological characteristics of the teenager depressive disorder patient; the method comprises the following steps: constructing a brain structure connection network of a patient based on brain grey matter morphological characteristics of a teenager depressive disorder patient; based on a brain structure connection network and a graph theory, extracting topological features of a brain network; and inputting the extracted brain network topology features into a two-stage machine learning model for curative effect prediction. According to the technical scheme, the most representative key information is extracted, and the brain structure network features and the machine learning technology can be effectively fused, so that the prediction accuracy of the ECT curative effect of the teenager MDD patient is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical image analysis and relates to a method and system for predicting the efficacy of electroconvulsive therapy for adolescent patients with severe depression. Background Art

[0002] Major depressive disorder (MDD) is a severe and disabling condition affecting over 350 million people worldwide. MDD is more common during adolescence, a developmental period characterized by dynamic changes in brain structure and functional organization. Compared with adults with MDD, adolescents with MDD exhibit higher rates of irritability and lower remission rates.

[0003] Evidence suggests that approximately 60% of adolescents with MDD do not respond well to medication and psychotherapy, failing to achieve optimal clinical outcomes. Electroconvulsive therapy (ECT) has been shown to rapidly alleviate depressive symptoms in treatment-resistant adolescents with MDD, with response rates as high as 70%-80%. In adolescents with MDD, ECT not only reduces suicidal ideation and anxiety symptoms but also improves overall functional status. However, the proportion of adolescents with MDD receiving ECT is much lower than that of adults with MDD, and research on this topic is limited.

[0004] The brain structure and functional changes mediated by ECT in adolescent MDD patients are still not well understood. Specific network characteristics and prediction models for adolescent groups have not yet been perfected. Individual efficacy varies significantly, and there is a lack of reliable prediction methods. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for predicting the efficacy of electroconvulsive therapy for adolescent patients with major depression, aiming to improve the accuracy of ECT efficacy prediction.

[0006] To achieve the above objectives, the basic scheme of the present invention is: a method for predicting the efficacy of electroconvulsive therapy for adolescent patients with major depression, comprising the following steps:

[0007] Structural magnetic resonance imaging data of adolescent patients with depressive disorders before and after electroconvulsive therapy were collected and preprocessed to obtain the brain gray matter morphological characteristics of adolescent patients with depressive disorders;

[0008] Based on the gray matter morphological characteristics of adolescent depressive disorder patients, a brain structural connection network of patients was constructed;

[0009] Based on the brain structure connection network and graph theory, the topological features of the brain structure connection network are extracted;

[0010] The extracted brain network topology features are input into a two-stage machine learning model for efficacy prediction:

[0011] (1) The first stage: compress high-dimensional topological features through dimensionality reduction algorithms to select the most discriminative features;

[0012] (2) The second stage: The features after dimensionality reduction are input into the prediction model to obtain the ECT efficacy prediction results and the prediction results are analyzed for interpretability using the SHAP method.

[0013] The working principle and beneficial effects of this basic solution are: This technical solution effectively integrates structural brain magnetic resonance imaging features with machine learning technology, thereby significantly improving the prediction accuracy of ECT efficacy in adolescent MDD patients

[0014] Furthermore, the method for collecting structural magnetic resonance imaging data of adolescent MDD patients before and after ECT treatment is as follows:

[0015] Structural magnetic resonance imaging scans were performed one day before and one day after the MDD patients completed ECT.

[0016] Structural magnetic resonance imaging (MRI) scans were performed using a 3T system and a 32-channel head coil. During the scan, subjects were instructed to close their eyes but remain awake and try not to think. Foam pads and earplugs were used to reduce head movement and instrument noise. Structural imaging used a perturbed gradient echo sequence to acquire high-resolution 3D T1-weighted images. Scan parameters were as follows: repetition time (TR) = 2000 ms, inversion time (TI) = 900 ms, echo time (TE) = 2.56 ms, flip angle = 9°, matrix = 256 × 256, field of view (FOV) = 25.6 cm × 25.6 cm, slice thickness = 1 mm, 192 slices per block, and voxel size = 1.0 mm × 1.0 mm × 1.0 mm.

[0017] All images were visually inspected by two senior radiologists to confirm the absence of obvious lesions or scanning artifacts and to correct for distortion caused by gradient nonlinearity.

[0018] Collect relevant data to facilitate analysis and processing.

[0019] Furthermore, the method for preprocessing the acquired structural magnetic resonance imaging images is as follows:

[0020] The SPM8 toolbox in the MATLAB R2022a platform was used to preprocess the structural magnetic resonance imaging data to obtain the morphological characteristics of the patient's brain gray matter. The specific processing steps are as follows:

[0021] (1) Use the “Segment” function to segment the original structural image into gray matter, white matter, and cerebrospinal fluid;

[0022] (2) Use the “Normalise: Est&Write” tool to register / normalize the gray matter images of the healthy control group;

[0023] (3) In the “Smooth” function, a Gaussian kernel with a full width at half maximum of 6 mm was used to spatially smooth the registered gray matter image;

[0024] (4) Use the “Image Calculator” to average the smoothed gray matter image to generate a custom gray matter template;

[0025] (5) All participants were registered to the newly generated template using the “Normalise: Est&Write” tool again, and a second spatial smoothing was performed to improve the signal-to-noise ratio.

[0026] Preprocess the data and simplify it for subsequent use.

[0027] Furthermore, based on the gray matter morphological characteristics of adolescent depressive disorder patients, the steps for constructing the patient's brain structural connection network are as follows:

[0028] Based on the automatic anatomical annotation template, the brain gray matter was segmented into 90 regions of interest. Each region of interest was regarded as a node in the network, and the connections between different nodes were regarded as edges of the network.

[0029] Each node is quantified by the probability density function of the gray matter intensity values ​​of all voxels in the region, and each edge is quantified by the index of Kullback-Leibler divergence similarity KLS, which ranges from 0 to 1, with 1 indicating that the distributions are exactly the same, and the closer the value is to 1, the higher the similarity of the two distributions;

[0030] A 90×90 structural connectivity network was constructed for each patient before and after ECT treatment. The rows and columns of the structural connectivity network matrix represent different nodes, and each network element represents the similarity between any pair of nodes.

[0031] Construct a brain structural connectivity network and quantify the structural covariation relationship between brain regions.

[0032] Furthermore, based on the brain structure connection network and graph theory, the specific steps for extracting the topological features of the brain structure connection network are as follows:

[0033] Based on graph complex network theory, the GRETNA toolbox was used to comprehensively investigate the changes in the topological properties of the brain structure connection network in adolescent MDD patients before and after ECT.

[0034] Topological attributes include seven global network topology indicators: global efficiency Eglob, local efficiency Eloc, clustering coefficient Cp, characteristic path length Lp, standardized clustering coefficient γ, standardized characteristic path length λ, small-worldness index σ, and 270 node-level attributes: degree, efficiency, betweenness centrality × 90 brain regions;

[0035] The sparsity S of all correlation matrices was set in the range of 0.05 to 0.34, with a step size of 0.01, based on criteria such as average node degree > 2×log(90) and small-worldness index σ > 1.0. To sensitively capture changes in network topology, the area under the curve (AUC) of each indicator within this sparsity range was calculated as a comprehensive indicator of network topology properties.

[0036] Using brain structural connectivity networks and graph theory, we extract the topological features of the brain networks of adolescent MDD patients, which is beneficial for predicting therapeutic effects. Furthermore, the two-stage machine learning model is specifically as follows:

[0037] Phase 1: Based on the high-dimensional nature of neuroimaging data, feature dimensionality reduction helps retain key features and improve the accuracy of the prediction model. Three dimensionality reduction algorithms are used: kernel principal component analysis (KPCA), factor analysis (FA), and autoencoder (AE).

[0038] Phase II: To predict the efficacy of ECT, the topological features after dimensionality reduction were imported into the prediction model. The prediction model used one of the support vector machine (SVM), random forest (RF), and penalized logistic regression (PLR). The feature interpretability of the prediction results was analyzed using the SHAP method.

[0039] Select prediction models according to needs to improve the accuracy of predicting the efficacy of ECT in adolescent MDD patients.

[0040] Furthermore, adjust the position and parameters of the electrodes of the ECT treatment device. The specific steps are as follows:

[0041] During the first treatment, electrodes were placed on the patient's bilateral temporal lobes. The initial electrical stimulation energy was determined based on the patient's age and gender: initial energy percentage = age × 50%. Male patients were increased by 5%-10%, while female patients were decreased by 5%-10%.

[0042] Seizures were monitored by electroencephalography, and stimulation energy was gradually increased if there was no response. All patients received propofol (1.4 mg / kg) for anesthesia and succinylcholine (0.5 mg / kg) for muscle relaxation during ECT.

[0043] If abnormalities in node centrality in specific brain regions are identified as being associated with therapeutic efficacy, the position and parameters of the electrodes in the ECT treatment device will be readjusted: if the predicted results show abnormal prefrontal connectivity, prefrontal electrodes will be added to enhance regulation of the default mode network; if the abnormalities are concentrated in the right hemisphere, left brain stimulation will be reduced to reduce cognitive side effects.

[0044] Based on the prediction results and their feature interpretability analysis, the position and parameters of the electrodes of the ECT treatment device are adjusted to optimize the treatment effect.

[0045] The present invention also provides a system for predicting the efficacy of electroconvulsive therapy for adolescent patients with severe depression, comprising a bed, and an auxiliary mechanism, a protective mechanism, an ECT treatment mechanism, a human-computer interaction module, and a prediction module arranged on the bed;

[0046] The auxiliary mechanism includes a groove, a pillow and an elevator. The groove is set on the bed at a position corresponding to the patient's thoracic vertebra, the pillow is placed in the groove, and the telescopic end of the elevator is connected to the bottom of the pillow to control the up and down displacement of the pillow.

[0047] The protective mechanism includes an elastic band, a socket and a tongue depressor. One end of the elastic band and the socket are respectively installed on both sides of the bed corresponding to the position of the patient's head. The other end of the elastic band is connected to a plug that can be snapped into the socket. The tongue depressor is installed on the elastic band.

[0048] The human-computer interaction module is used to collect structural magnetic resonance imaging data of adolescent MDD patients before and after ECT treatment. The output end of the human-computer interaction module is connected to the input end of the prediction module. The prediction module executes the method described in the present invention, obtains the prediction results and outputs them to the human-computer interaction module for display.

[0049] This system is equipped with a bed, and the patient lies flat on the bed for ECT treatment and efficacy prediction. The ECT treatment institution provides corresponding treatment to the patient. The human-computer interaction module collects structural magnetic resonance imaging data of adolescent MDD patients before and after ECT treatment, and transmits it to the prediction module to obtain the prediction results, which are displayed on the human-computer interaction module for easy viewing.

[0050] Before the patient undergoes ECT, the auxiliary and protective mechanisms are activated. The auxiliary mechanism's elevator mechanism shifts the cushion in the recess upward, creating a bulge on the bed's upper surface. This acts between the patient's shoulder blades, corresponding to the mid-thoracic spine. This protrudes the patient's spine, maintains airway patency, and prevents suffocation. This eliminates the need for medical staff to place objects under the patient, reducing their burden.

[0051] Insert the plug on the elastic band of the protective mechanism into the socket, and place the tongue depressor on the elastic band into the corresponding position in the patient's mouth. The elastic force of the elastic band is used to press the tongue depressor against the patient's tongue. The structure is simple and can also prevent the individual tongue depressor from falling out of the patient's mouth. It is also easy to take and put, and easy to operate.

[0052] Furthermore, the protective mechanism also includes gauze and a clamping piece. The gauze can be wound around the tongue depressor. The clamping piece is installed at both ends of the tongue depressor and is used to clamp the two ends of the gauze.

[0053] The gauze wrapped around the tongue depressor makes the gauze softer and increases the patient's comfort. Clamps are set on both sides of the tongue depressor to clamp and position the two ends of the gauze, making it easy to position and remove the gauze.

[0054] Furthermore, the protection mechanism further comprises a first airbag, a second airbag and an inflation and deflation mechanism;

[0055] A clamping block is provided at one end of the first airbag, and a clamping slot is provided on the tongue depressor to engage with the clamping block. The first airbag is expanded to fill the patient's oral cavity;

[0056] The second airbag is a long strip airbag, and both ends of the second airbag can be clamped and connected with the clamping member;

[0057] The inflation and deflation mechanism is installed on the bed body and is connected to the first airbag and the second airbag in a one-to-one correspondence through the first air pipe and the second air pipe. The first air pipe and the second air pipe are both provided with electric valves.

[0058] A first airbag is provided, and when needed, the inflation and deflation mechanism is used to inflate the first airbag, causing it to expand and fill the patient's mouth, thereby optimizing the protective effect. The airbag can also be pressed against the tongue depressor to make its positioning on the patient's tongue more stable, thereby preventing the patient from biting his tongue during ECT treatment.

[0059] The second airbag is elongated and sized to support the patient's jaw when inflated, preventing dislocation. It can be made of an elastic material, such as rubber. Clamps also secure the second airbag at both ends, facilitating assembly and disassembly. Adjustable clamps can also be used to position the second airbag, allowing the size of the supporting portion to be adjusted to suit specific needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 1 is a schematic top view of the structure of a system for predicting the efficacy of electroconvulsive therapy for adolescent patients with severe depression according to the present invention;

[0061] Figure 2 It is a front view structural schematic diagram of the system for predicting the efficacy of electroconvulsive therapy for adolescent patients with severe depression according to the present invention.

[0062] The reference numerals in the drawings of the specification include: bed 1, auxiliary mechanism 2, protection mechanism 3, ECT treatment mechanism 4, human-computer interaction module 5, prediction module 6;

[0063] Groove 21, pillow 22, lift 23;

[0064] Elastic band 31 , socket 32 ​​, tongue depressor 33 , plug 34 , clamping part 35 , inflation and deflation mechanism 36 . DETAILED DESCRIPTION

[0065] The following describes embodiments of the present invention in detail. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended only to explain the present invention and are not to be construed as limiting the present invention.

[0066] In the description of the present invention, it should be understood that the terms "longitudinal", "transverse", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention.

[0067] In the description of the present invention, unless otherwise specified and limited, it should be noted that the terms "installed", "connected" and "connected" should be understood in a broad sense. For example, it can be a mechanical connection or an electrical connection, or it can be the internal communication between two components. It can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to the specific circumstances.

[0068] The present invention discloses a method for predicting the efficacy of electroconvulsive therapy (ECT) for adolescent patients with major depressive disorder (MDD). The method effectively integrates structural brain magnetic resonance imaging (MRI) features with machine learning technology to improve the accuracy of predicting the efficacy of ECT for adolescent patients with MDD. The method for predicting the efficacy of ECT for adolescent patients with MDD includes the following steps:

[0069] Structural magnetic resonance imaging data of adolescent patients with depressive disorders were collected before and after electroconvulsive therapy (ECT), and preprocessed to obtain the gray matter morphological characteristics of the brains of adolescent patients with depressive disorders.

[0070] Based on the gray matter morphological characteristics of adolescent depressive disorder patients, a brain structural connection network of patients was constructed;

[0071] Based on the brain structure connection network and graph theory, the topological features of the brain structure connection network are extracted;

[0072] The extracted brain network topology features are input into a two-stage machine learning model for efficacy prediction:

[0073] (1) The first stage: compress high-dimensional topological features through dimensionality reduction algorithms to select the most discriminative features;

[0074] (2) The second stage: The features after dimensionality reduction are input into the prediction model to obtain the ECT efficacy prediction results and the prediction results are analyzed for interpretability using the SHAP method.

[0075] The label of the prediction model is binary (1 / 0), where 1 indicates that ECT treatment is effective and 0 indicates that ECT treatment is ineffective (before and after ECT treatment, the Hamilton Depression Rating Scale (HAMD) is used to assess the severity of patients' depressive symptoms. If the HAMD score decrease rate ((HAMD score before ECT - HAMD score after ECT) / HAMD score before ECT) ≥ 50%, the individual label is 1, otherwise it is 0).

[0076] The performance of the prediction model was evaluated by average sensitivity, specificity, and accuracy. In addition, 1000 random permutation tests were performed to determine whether the prediction performance of the model was significantly higher than the random level: the training set labels were randomly shuffled in each permutation, the model was retrained, and the classification accuracy was reassessed; after 1000 iterations, the probability that the permuted accuracy distribution was greater than or equal to the true accuracy was calculated as the significance level. The SHAP method was applied for interpretability analysis, focusing on the top 10% features / brain regions that contributed most to the prediction accuracy to explore the interpretability of the model. In the present invention, machine learning analysis was implemented based on Python 3.8.

[0077] In a preferred embodiment of the present invention, the method for acquiring structural magnetic resonance imaging data of adolescent MDD patients before and after ECT treatment is as follows:

[0078] Structural magnetic resonance imaging scans were performed one day before and one day after the MDD patients completed ECT.

[0079] Structural magnetic resonance imaging (MRI) scans were performed using a 3T system and a 32-channel head coil. During the scan, subjects were instructed to close their eyes but remain awake and try not to think. Foam pads and earplugs were used to reduce head movement and instrument noise. Structural imaging used a perturbed gradient echo sequence to acquire high-resolution 3D T1-weighted images. Scan parameters were as follows: repetition time (TR) = 2000 ms, inversion time (TI) = 900 ms, echo time (TE) = 2.56 ms, flip angle = 9°, matrix = 256 × 256, field of view (FOV) = 25.6 cm × 25.6 cm, slice thickness = 1 mm, 192 slices per block, and voxel size = 1.0 mm × 1.0 mm × 1.0 mm.

[0080] All images were visually inspected by two senior radiologists to confirm the absence of obvious lesions or scanning artifacts and to correct for distortion caused by gradient nonlinearity.

[0081] In a preferred embodiment of the present invention, the method for preprocessing the acquired structural magnetic resonance imaging data is as follows:

[0082] The SPM8 toolbox in the MATLAB R2022a platform was used to preprocess the structural magnetic resonance imaging images to obtain the morphological characteristics of the patient's brain gray matter. The specific processing steps are as follows:

[0083] (1) Use the “Segment” function to segment the original structural image into gray matter, white matter, and cerebrospinal fluid;

[0084] (2) Use the “Normalise: Est&Write” tool to register / normalize the gray matter images of the healthy control group;

[0085] (3) In the “Smooth” function, a Gaussian kernel with a full width at half maximum of 6 mm was used to spatially smooth the registered gray matter image;

[0086] (4) Use the “Image Calculator” to average the smoothed gray matter image to generate a custom gray matter template;

[0087] (5) All participants were registered to the newly generated template using the “Normalise: Est&Write” tool again, and a second spatial smoothing was performed to improve the signal-to-noise ratio.

[0088] MRI segmented brain templates serve as standard references for brain image registration. Commonly used templates are often based on Western European or North American adults. If the reference templates do not match age or ethnicity, severe brain tissue deformation can occur. Studies have confirmed that these templates are not suitable for brain images of Chinese adolescents. Compared to tools like SPM12 and CAT12, SPM8 offers greater operational flexibility in generating custom brain templates, significantly improving preprocessing accuracy.

[0089] In a preferred embodiment of the present invention, the steps of constructing a brain structural connectivity network of adolescent patients with depressive disorders based on their gray matter morphological characteristics are as follows:

[0090] Based on the automatic anatomical annotation template, the brain gray matter was segmented into 90 regions of interest. Each region of interest was regarded as a node in the network, and the connections between different nodes were regarded as edges of the network.

[0091] Each node is quantified by the probability density function of the gray matter intensity values ​​of all voxels in the region, and each edge is quantified by the index of Kullback-Leibler divergence similarity KLS, which ranges from 0 to 1, with 1 indicating that the distributions are exactly the same, and the closer the value is to 1, the higher the similarity of the two distributions;

[0092] A 90×90 structural connectivity network was constructed for each patient before and after ECT treatment. The rows and columns of the structural connectivity network matrix represent different nodes, and each network element represents the similarity between any pair of nodes.

[0093] In a preferred embodiment of the present invention, the steps of extracting the brain network topology characteristics of adolescent MDD patients using brain structural connectivity network and graph theory are as follows:

[0094] Based on graph complex network theory, the GRETNA toolbox (https: / / www.nitrc.org / projects / gretna / ) was used to comprehensively investigate the changes in the topological properties of the brain structural connection network in adolescent MDD patients before and after ECT.

[0095] Topological properties include seven global network topology indicators: global efficiency (Eglob), local efficiency (Eloc), clustering coefficient (Cp), characteristic path length (Lp), normalized clustering coefficient (γ), normalized characteristic path length (λ), small-world property (σ), and 270 node-level indicators: node degree, node efficiency, and betweenness centrality × 90 brain regions.

[0096] To ensure the accuracy of topological feature calculations, the sparsity S of all correlation matrices was set between 0.05 and 0.34 with a step size of 0.01, based on criteria such as average node degree > 2×log(90) and small-worldness index σ > 1.0. To sensitively capture network topological changes, the area under the curve (AUC) of each metric within this sparsity range was calculated as a comprehensive indicator of network topological properties.

[0097] By utilizing brain structural connectivity networks and graph theory, we can extract the topological characteristics of brain networks in adolescent MDD patients, which is beneficial for predicting therapeutic effects.

[0098] In a preferred embodiment of the present invention, the two-stage machine learning model is specifically:

[0099] Phase 1: Considering the high-dimensional nature of neuroimaging data, feature dimensionality reduction helps preserve key features and improve the accuracy of the prediction model. For this study, existing dimensionality reduction algorithms such as kernel principal component analysis (KPCA), factor analysis (FA), and autoencoders (AE) can be used.

[0100] Phase II: To predict ECT efficacy, the reduced topological features are imported into a prediction model. The prediction classification model can be a classic and widely used machine learning model, such as support vector machine (SVM), random forest (RF), or penalized logistic regression (PLR). Alternatively, the prediction model uses FA+SVM and performs feature interpretability analysis to improve the accuracy of predicting ECT efficacy in adolescent MDD patients.

[0101] In a preferred embodiment of the present invention, the feature interpretability analysis of the model prediction results is performed based on the SHAP method to adjust the position and parameters of the electrodes of the ECT treatment device. The specific steps are as follows:

[0102] During the first treatment, electrodes were placed on the patient's bilateral temporal lobes. The initial electrical stimulation energy was determined based on the patient's age and gender: initial energy percentage = age × 50%. Male patients were increased by 5%-10%, while female patients were decreased by 5%-10%.

[0103] Epileptic seizures were monitored by electroencephalography, and if there was no response, the stimulation energy was gradually increased. All patients were anesthetized with propofol (1.4 mg / kg) and muscle relaxant with succinylcholine (0.5 mg / kg) during ECT.

[0104] If abnormalities in nodal centrality in specific brain regions are identified that are associated with efficacy, the electrode positions and parameters of the ECT device can be readjusted:

[0105] If the prediction results show abnormal prefrontal connectivity, then add prefrontal electrodes to enhance the regulation of the default mode network;

[0106] If the abnormality is concentrated in the right hemisphere, then stimulation of the left brain is reduced, minimizing cognitive side effects.

[0107] Based on the prediction results, the position and parameters of the electrodes of the ECT treatment device are adjusted to optimize the treatment effect.

[0108] The present invention also provides a system for predicting the efficacy of electroconvulsive therapy for adolescent patients with severe depression. Figure 1 and Figure 2 As shown, it includes a bed 1, and an auxiliary mechanism 2, a protection mechanism 3, an ECT treatment mechanism 4, a human-computer interaction module 5 (such as a computer connected to a camera) and a prediction module 6 arranged on the bed 1.

[0109] The auxiliary mechanism 2 includes a groove 21, a bolster 22, and an elevator 23. The groove 21 is set on the bed 1 at a position corresponding to the patient's thoracic vertebrae, and the bolster 22 is placed in the groove 21. The telescopic end of the elevator 23 is connected to the bottom of the bolster 22 (which can be a sand pillow, cotton pillow, etc.) to control the vertical displacement of the bolster 22. The elevator 23 can use a pneumatic telescopic rod, a hydraulic telescopic rod, an electric telescopic rod, etc. to achieve reciprocating linear lifting motion. The control end of the elevator 23 can be electrically connected to a push button installed on the bed 1. The operation of the elevator 23 is controlled by pressing the button.

[0110] The protective mechanism 3 includes an elastic band 31, a socket 32, and a tongue depressor 33. One end of the elastic band 31 and the socket 32 ​​are respectively mounted on the bed 1 at opposite sides of the bed corresponding to the position of the patient's head. The other end of the elastic band 31 is connected to a plug 34 that can be snap-fitted into the socket 32. The tongue depressor 33 is mounted on the elastic band 31. The plug 34 and the socket 32 ​​can adopt a conventional snap-fit ​​structure. One end of the elastic band 31 can be tied and connected to the plug 34 to adjust the length of the elastic band 31.

[0111] Human-computer interaction module 5 is used to collect structural magnetic resonance imaging data of adolescent MDD patients before and after ECT treatment. The output end of human-computer interaction module 5 is electrically connected to the input end of prediction module 6. Prediction module 6 executes the method described in the present invention, obtains prediction results, and outputs them to the human-computer interaction module for display.

[0112] This system is equipped with a bed 1, and the patient lies flat on the bed 1 for ECT treatment and efficacy prediction. The ECT treatment institution 4 performs corresponding treatment on the patient. The human-computer interaction module 5 collects structural magnetic resonance imaging data of adolescent MDD patients before and after ECT treatment and transmits it to the prediction module 6 to obtain the prediction results and display them on the human-computer interaction module 5 for easy viewing.

[0113] Before a patient undergoes ECT treatment, assist mechanism 2 and protective mechanism 3 are activated. Elevator 23 of assist mechanism 2 activates, driving bolster 22 in recess 21 upward. This bulges the upper surface of bed 1 and acts between the patient's shoulder blades, corresponding to the mid-thoracic spine. This protrudes the patient's spine, maintains airway patency, and prevents suffocation. This eliminates the need for medical staff to place objects under the patient, reducing their burden.

[0114] Insert the plug 34 on the elastic band 31 of the protective mechanism 3 into the socket 32, and place the tongue depressor 33 on the elastic band 31 into the corresponding position in the patient's mouth. The elastic force of the elastic band 31 is used to achieve tight compression between the tongue depressor 33 and the patient's tongue. The structure is simple and can also prevent the individual tongue depressor 33 from falling out of the patient's mouth. It is also easy to take and put, and convenient to operate.

[0115] In a preferred embodiment of the present invention, the protective mechanism 3 further includes gauze and a clamp 35. The gauze can be wrapped around the tongue depressor 33. The clamp 35 is mounted at both ends of the tongue depressor 33 and is used to clamp the ends of the gauze. The clamp 35 can be a miniature electric / pneumatic clamp, or a tightly clamping steel clamp.

[0116] Gauze is set to wrap around tongue depressor 33, and gauze is softer, increases patient's comfort. Clamping piece 35 is set on tongue depressor 33 both sides, realizes the clamping location to gauze two ends, is convenient to gauze location and disassembly and replacement.

[0117] In a preferred embodiment of the present invention, the protection mechanism 3 further includes a first airbag, a second airbag and an inflation and deflation mechanism 36 .

[0118] The first airbag is relatively small and does not affect the use of the tongue depressor 33 when not inflated. A latch is provided at one end of the first airbag, and a slot is provided on the tongue depressor 33 to engage with the latch. This allows the first airbag to be removed from the slot of the tongue depressor 33 when not in use, making it easier to use. When inflated, the first airbag fills the patient's oral cavity.

[0119] The second airbag is a long strip airbag, and both ends of the second airbag can be clamped and connected with the clamping member 35.

[0120] The inflation / deflation mechanism 36 is mounted on the bed 1 and is connected to the first and second airbags via a first air tube and a second air tube, with the second air tube preferably communicating with the center of the second airbag. Both the first and second air tubes are equipped with electric valves. The control terminals of the inflation / deflation mechanism 36 and the electric valves are electrically connected to corresponding control buttons mounted on the bed 1. Medical personnel can control the corresponding devices using these control buttons.

[0121] A first airbag is provided, and when needed, the inflation and deflation mechanism 36 is used to inflate the first airbag, and the electric valve on the first trachea is controlled to be in an open state, so that the first airbag expands and fills the patient's mouth, thereby optimizing the protective effect. The tongue depressor 33 can also be pressed to make the positioning of the tongue of the patient more stable, thereby preventing the patient from biting the tongue during ECT treatment.

[0122] The second airbag is elongated and sized to support the patient's jaw when inflated, preventing joint dislocation. The second airbag can be made of an elastic material, such as elastic rubber. Clamps 35 also hold the second airbag in place at both ends, facilitating assembly and disassembly. The clamps 35 can also be adjusted to different positions, allowing the size of the supporting portion of the second airbag to be adjusted to suit specific needs.

[0123] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0124] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

Claims

1. A method for predicting the efficacy of electroconvulsive therapy for adolescent patients with major depression, characterized in that: The steps include: Structural magnetic resonance imaging data of adolescent patients with depressive disorders before and after electroconvulsive therapy were collected and preprocessed to obtain the brain gray matter morphological characteristics of adolescent patients with depressive disorders; Based on the gray matter morphological characteristics of adolescent depressive disorder patients, a brain structural connection network of patients was constructed; Based on the brain structure connection network and graph theory, the topological features of the brain structure connection network are extracted; The extracted brain network topology features are input into a two-stage machine learning model for efficacy prediction: (1) The first stage: compress high-dimensional topological features through dimensionality reduction algorithms to select the most discriminative features; (2) The second stage: The features after dimensionality reduction are input into the prediction model to obtain the ECT efficacy prediction results and the prediction results are analyzed for interpretability using the SHAP method.

2. The method for predicting the efficacy of electroconvulsive therapy for adolescent patients with major depression according to claim 1, wherein: Structural MRI data were collected from adolescent patients with MDD before and after ECT treatment using the following method: Structural magnetic resonance imaging scans were performed one day before and one day after the MDD patients completed ECT. Structural MRI scans were performed using a 3T system and a 32-channel head coil. During the scan, participants were asked to close their eyes but remain awake and try not to think. Foam pads and earplugs were also used to reduce head movement and instrument noise. Structural imaging was performed using a perturbation gradient echo sequence to acquire high-resolution 3D T1-weighted images. The scanning parameters were as follows: repetition time (TR) = 2000 ms, inversion time (TI) = 900 ms, echo time (TE) = 2.56 ms, flip angle = 9°, matrix = 256 × 256, field of view (FOV) = 25.6 cm × 25.6 cm, slice thickness = 1 mm, 192 slices per block, and voxel size = 1.0 mm × 1.0 mm × 1.0 mm. All images were visually inspected by two senior radiologists to confirm the absence of obvious lesions or scanning artifacts and to correct for distortion caused by gradient nonlinearity.

3. The method for predicting the efficacy of electroconvulsive therapy for adolescents with major depression according to claim 1, wherein: The method for preprocessing the acquired structural magnetic resonance imaging images is as follows: The SPM8 toolbox in the MATLAB R2022a platform was used to preprocess the structural magnetic resonance imaging data to obtain the morphological characteristics of the patient's brain gray matter. The specific processing steps are as follows: (1) Use the "Segment" function to segment the original structural image into gray matter, white matter, and cerebrospinal fluid; (2) Use the "Normalise: Est&Write" tool to register / normalize the gray matter images of the healthy control group; (3) In the "Smooth" function, a Gaussian kernel with a full width at half maximum of 6 mm was used to spatially smooth the registered gray matter image; (4) Use "Image Calculator" to average the smoothed gray matter image to generate a custom gray matter template; (5) All participants were registered to the newly generated template using the "Normalise: Est&Write" tool again, and a second spatial smoothing was performed to improve the signal-to-noise ratio.

4. The method for predicting the efficacy of electroconvulsive therapy for adolescents with major depression according to claim 1, wherein: Based on the gray matter morphological characteristics of adolescent depressive disorder patients, the steps for constructing the patient's brain structural connection network are as follows: Based on the automatic anatomical annotation template, the brain gray matter was segmented into 90 regions of interest. Each region of interest was regarded as a node in the network, and the connections between different nodes were regarded as edges of the network. Each node is quantified by the probability density function of the gray matter intensity values ​​of all voxels in the region, and each edge is quantified by the index of Kullback-Leibler divergence similarity KLS, which ranges from 0 to 1, with 1 indicating that the distributions are exactly the same, and the closer the value is to 1, the higher the similarity of the two distributions; A 90×90 structural connectivity network was constructed for each patient before and after ECT treatment. The rows and columns of the structural connectivity network matrix represent different nodes, and each network element represents the similarity between any pair of nodes.

5. The method for predicting the efficacy of electroconvulsive therapy for adolescents with major depression according to claim 1, wherein: Based on the brain structure connection network and graph theory, the specific steps for extracting the topological features of the brain structure connection network are as follows: Based on graph complex network theory, the GRETNA toolbox was used to comprehensively investigate the changes in the topological properties of the brain structure connection network in adolescent MDD patients before and after ECT. Topological attributes include seven global network topology indicators: global efficiency Eglob, local efficiency Eloc, clustering coefficient Cp, characteristic path length Lp, standardized clustering coefficient γ, standardized characteristic path length λ, small-worldness index σ, and 270 node-level attributes: degree, efficiency, betweenness centrality × 90 brain regions; The sparsity S of all correlation matrices was set in the range of 0.05 to 0.34, with a step size of 0.01, based on criteria such as average node degree > 2×log(90) and small-worldness index σ > 1.

0. To sensitively capture changes in network topology, the area under the curve (AUC) of each indicator within this sparsity range was calculated as a comprehensive indicator of network topology properties.

6. The method for predicting the efficacy of electroconvulsive therapy for adolescent patients with major depression according to claim 1, wherein: The two-stage machine learning model is specifically: Phase 1: Based on the high-dimensional nature of neuroimaging data, feature dimensionality reduction helps retain key features and improve the accuracy of the prediction model. Three dimensionality reduction algorithms are used: kernel principal component analysis (KPCA), factor analysis (FA), and autoencoder (AE). Phase II: To predict the efficacy of ECT, the topological features after dimensionality reduction were imported into the prediction model. The prediction model used one of the support vector machine (SVM), random forest (RF), and penalized logistic regression (PLR). The feature interpretability of the prediction results was analyzed using the SHAP method.

7. The method for predicting the efficacy of electroconvulsive therapy for adolescents with major depression according to claim 1, wherein: According to the prediction results, adjust the position and parameters of the electrodes of the ECT treatment device. The specific steps are as follows: During the first treatment, electrodes were placed on the patient's bilateral temporal lobes. The initial electrical stimulation energy was determined based on the patient's age and gender: initial energy percentage = age × 50%. Male patients were increased by 5%-10%, while female patients were decreased by 5%-10%. Seizures were monitored by electroencephalography, and if there was no response, the stimulation energy was gradually increased; all patients received propofol during ECT: 1.4 mg / kg for anesthesia, and succinylcholine (0.5 mg / kg) for muscle relaxation; If abnormalities in nodal centrality in specific brain regions are identified that are associated with efficacy, the electrode positions and parameters of the ECT device can be readjusted: If the predicted results show abnormal prefrontal connectivity, then additional frontal electrodes will be added to enhance regulation of the default mode network; if the abnormality is concentrated in the right hemisphere, then left brain stimulation will be reduced to reduce cognitive side effects.

8. A system for predicting the efficacy of electroconvulsive therapy for adolescent patients with major depression, characterized by: It includes a bed body, and auxiliary mechanisms, protective mechanisms, ECT treatment mechanisms, human-computer interaction modules and prediction modules arranged on the bed body; The auxiliary mechanism includes a groove, a pillow and an elevator. The groove is set on the bed at a position corresponding to the patient's thoracic vertebra, the pillow is placed in the groove, and the telescopic end of the elevator is connected to the bottom of the pillow to control the up and down displacement of the pillow. The protective mechanism includes an elastic band, a socket and a tongue depressor. One end of the elastic band and the socket are respectively installed on both sides of the bed corresponding to the position of the patient's head. The other end of the elastic band is connected to a plug that can be snapped into the socket. The tongue depressor is installed on the elastic band. The human-computer interaction module is used to collect structural magnetic resonance imaging data of adolescent MDD patients before and after ECT treatment. The output end of the human-computer interaction module is connected to the input end of the prediction module. The prediction module executes the method described in one of claims 1-7, obtains the prediction results and outputs them to the human-computer interaction module for display.

9. The system for predicting the efficacy of electroconvulsive therapy for adolescent patients with major depression according to claim 8, characterized in that: The protection mechanism also includes gauze and a clamping piece. The gauze can be wound on the tongue depressor. The clamping piece is installed at both ends of the tongue depressor and is used to clamp the two ends of the gauze.

10. The system for predicting the efficacy of electroconvulsive therapy for adolescent patients with major depression according to claim 9, characterized in that: The protection mechanism also includes a first airbag, a second airbag and an inflation and deflation mechanism; A clamping block is provided at one end of the first airbag, and a clamping slot is provided on the tongue depressor to engage with the clamping block. The first airbag is expanded to fill the patient's oral cavity; The second airbag is a long strip airbag, and both ends of the second airbag can be clamped and connected with the clamping member; The inflation and deflation mechanism is installed on the bed body and is connected to the first airbag and the second airbag in a one-to-one correspondence through the first air pipe and the second air pipe. The first air pipe and the second air pipe are both provided with electric valves.

Images