CN117594193A - Transcranial direct current personalized stimulation target positioning method based on deep learning - Google Patents
Transcranial direct current personalized stimulation target positioning method based on deep learning Download PDFInfo
- Publication number
- CN117594193A CN117594193A CN202410063719.3A CN202410063719A CN117594193A CN 117594193 A CN117594193 A CN 117594193A CN 202410063719 A CN202410063719 A CN 202410063719A CN 117594193 A CN117594193 A CN 117594193A
- Authority
- CN
- China
- Prior art keywords
- deep learning
- electric field
- direct current
- transcranial direct
- field simulation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000013135 deep learning Methods 0.000 title claims abstract description 82
- 230000000638 stimulation Effects 0.000 title claims abstract description 73
- 238000000034 method Methods 0.000 title claims abstract description 44
- 230000005684 electric field Effects 0.000 claims abstract description 76
- 238000004088 simulation Methods 0.000 claims abstract description 76
- 210000003484 anatomy Anatomy 0.000 claims abstract description 42
- 238000012549 training Methods 0.000 claims abstract description 32
- 238000004070 electrodeposition Methods 0.000 claims abstract description 20
- 238000010586 diagram Methods 0.000 claims abstract description 15
- 238000009826 distribution Methods 0.000 claims abstract description 15
- 238000007781 pre-processing Methods 0.000 claims abstract description 12
- 210000004556 brain Anatomy 0.000 claims description 23
- 238000004364 calculation method Methods 0.000 claims description 10
- 238000012360 testing method Methods 0.000 claims description 10
- 239000011159 matrix material Substances 0.000 claims description 9
- 238000005457 optimization Methods 0.000 claims description 9
- 230000033228 biological regulation Effects 0.000 claims description 8
- 230000004913 activation Effects 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 5
- 238000012937 correction Methods 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 4
- 230000006870 function Effects 0.000 claims description 4
- 230000003920 cognitive function Effects 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 claims description 3
- 230000007246 mechanism Effects 0.000 claims description 3
- 208000020016 psychiatric disease Diseases 0.000 claims description 3
- 210000003128 head Anatomy 0.000 description 10
- 238000002595 magnetic resonance imaging Methods 0.000 description 7
- 210000001519 tissue Anatomy 0.000 description 7
- 230000008569 process Effects 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 3
- 238000003384 imaging method Methods 0.000 description 3
- 210000004761 scalp Anatomy 0.000 description 3
- 206010001497 Agitation Diseases 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000013136 deep learning model Methods 0.000 description 2
- 239000003814 drug Substances 0.000 description 2
- 238000002372 labelling Methods 0.000 description 2
- 230000000926 neurological effect Effects 0.000 description 2
- 230000001105 regulatory effect Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 210000003625 skull Anatomy 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 235000002566 Capsicum Nutrition 0.000 description 1
- 239000006002 Pepper Substances 0.000 description 1
- 235000016761 Piper aduncum Nutrition 0.000 description 1
- 235000017804 Piper guineense Nutrition 0.000 description 1
- 244000203593 Piper nigrum Species 0.000 description 1
- 235000008184 Piper nigrum Nutrition 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000003925 brain function Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 210000003710 cerebral cortex Anatomy 0.000 description 1
- 210000001175 cerebrospinal fluid Anatomy 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000001276 controlling effect Effects 0.000 description 1
- 230000001054 cortical effect Effects 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 229940079593 drug Drugs 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 210000004884 grey matter Anatomy 0.000 description 1
- 230000005764 inhibitory process Effects 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 150000003839 salts Chemical class 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 210000004885 white matter Anatomy 0.000 description 1
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
- G16H20/30—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance relating to physical therapies or activities, e.g. physiotherapy, acupressure or exercising
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/0033—Features or image-related aspects of imaging apparatus classified in A61B5/00, e.g. for MRI, optical tomography or impedance tomography apparatus; arrangements of imaging apparatus in a room
- A61B5/004—Features or image-related aspects of imaging apparatus classified in A61B5/00, e.g. for MRI, optical tomography or impedance tomography apparatus; arrangements of imaging apparatus in a room adapted for image acquisition of a particular organ or body part
- A61B5/0042—Features or image-related aspects of imaging apparatus classified in A61B5/00, e.g. for MRI, optical tomography or impedance tomography apparatus; arrangements of imaging apparatus in a room adapted for image acquisition of a particular organ or body part for the brain
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/05—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves
- A61B5/055—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves involving electronic [EMR] or nuclear [NMR] magnetic resonance, e.g. magnetic resonance imaging
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61N—ELECTROTHERAPY; MAGNETOTHERAPY; RADIATION THERAPY; ULTRASOUND THERAPY
- A61N1/00—Electrotherapy; Circuits therefor
- A61N1/18—Applying electric currents by contact electrodes
- A61N1/20—Applying electric currents by contact electrodes continuous direct currents
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
- G06N3/0455—Auto-encoder networks; Encoder-decoder networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/0985—Hyperparameter optimisation; Meta-learning; Learning-to-learn
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/50—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for simulation or modelling of medical disorders
Landscapes
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Theoretical Computer Science (AREA)
- Molecular Biology (AREA)
- Public Health (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Animal Behavior & Ethology (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Veterinary Medicine (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Pathology (AREA)
- Radiology & Medical Imaging (AREA)
- Heart & Thoracic Surgery (AREA)
- Surgery (AREA)
- Epidemiology (AREA)
- Primary Health Care (AREA)
- Neurology (AREA)
- Databases & Information Systems (AREA)
- Physical Education & Sports Medicine (AREA)
- High Energy & Nuclear Physics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physiology (AREA)
- Psychiatry (AREA)
- Signal Processing (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a method for positioning a transcranial direct current personalized stimulation target point based on deep learning, which comprises the following steps: acquiring an individual anatomy structure magnetic resonance image, and preprocessing the individual anatomy structure magnetic resonance image to obtain a high-quality structural image; performing tDCS inverse electric field simulation on the high-quality structural image to obtain inverse electric field simulation data comprising an electric field simulation distribution diagram, tDCS electrode positions and corresponding current magnitudes; constructing individual anatomy magnetic resonance images and inverse electric field simulation data into a deep learning data set; and training the pre-constructed deep learning network by taking part of data in the deep learning data set as a training set so as to obtain the position of the transcranial direct current personalized stimulation target point and corresponding current information by utilizing the trained network. The method can realize rapid and accurate prediction of the location of the tDCS stimulation target point, saves time cost and increases the feasibility of customizing the tDCS personalized scheme in clinical application.
Description
Technical Field
The invention belongs to the technical field of transcranial direct current stimulation, and particularly relates to a method for positioning a transcranial direct current personalized stimulation target point based on deep learning.
Background
Transcranial direct current stimulation (transcranial direct current stimulation, tDCS) is one of the important non-invasive brain stimulation techniques, typically modulating brain function by applying weak direct current to the scalp surface to affect the excitability and inhibition of the cerebral cortex. tDCS can be realized by placing two electrodes (anode and cathode) on the scalp, and the direction and intensity of the current can be flexibly adjusted according to the use requirement.
When tDCS acts, it generates a certain electric field in the brain, which causes a change in cortical excitability. The specific distribution of the electric field is related to the individual head volume, skull thickness and cortex thickness, so that the same tDCS stimulation of different patients will produce different electric field distributions, which in turn will lead to individual variability in the efficacy. And different electrode montages, different stimulation electrode positions all result in different currents flowing through the brain. The accurate determination of the stimulation electrode position (i.e. the stimulation target point) is important to optimize the position and stimulation parameters of the tDCS treatment and to customize the tDCS personalized scheme.
The traditional transcranial direct current stimulation regulation position determination is mainly carried out by wearing 10-20 brain electrical positioning caps, and the target area marked by the positioning caps is directly used. And the specific position for regulating and controlling the electrode placement in the corresponding brain region is mainly obtained through previous experience. However, since there are individual anatomical structures, differences in brain morphology and size due to age, sex, etc., individual differences exist in regulation of a fixed position using a standard 10-20 electroencephalogram positioning cap, and thus a brain region of interest associated with a certain disease cannot be stimulated accurately.
Later, related personnel propose can be before carrying out tDCS stimulation, calculate accurate tDCS electrode position and the electric current size that specific target point was regulated and control through reverse electric field simulation. However, this method is costly in time, requires about 4 hours per test, and requires high demands on the clinical operators, and is not feasible for clinical use.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides a method, a device and a system for positioning a transcranial direct current personalized stimulation target point based on deep learning. The technical problems to be solved by the invention are realized by the following technical scheme:
in a first aspect, the invention provides a method for positioning a transcranial direct current personalized stimulation target point based on deep learning, which comprises the following steps:
acquiring an individual anatomy structure magnetic resonance image, and preprocessing the individual anatomy structure magnetic resonance image to obtain a high-quality structural image;
performing tDCS reverse electric field simulation on the high-quality structural image to obtain reverse electric field simulation data; the inverse electric field simulation data comprise an electric field simulation distribution diagram, tDCS electrode positions and corresponding current magnitudes;
constructing individual anatomy magnetic resonance images and inverse electric field simulation data into a deep learning data set;
and training the pre-constructed deep learning network by taking part of data in the deep learning data set as a training set so as to obtain the position of the transcranial direct current personalized stimulation target point and corresponding current information by utilizing the trained deep learning network.
In a second aspect, the invention provides a deep learning-based transcranial direct current personalized stimulation target point positioning device, which is used for realizing the deep learning-based transcranial direct current personalized stimulation target point positioning method provided by the first aspect, and comprises the following steps:
the image processing module is used for acquiring individual anatomy structure magnetic resonance images and preprocessing the individual anatomy structure magnetic resonance images to obtain high-quality structural images;
the simulation module is used for performing tDCS reverse electric field simulation on the high-quality structural image to obtain reverse electric field simulation data; the inverse electric field simulation data comprise an electric field simulation distribution diagram, tDCS electrode positions and corresponding current magnitudes;
the data set construction module is used for constructing the individual anatomy structure magnetic resonance image and the inverse electric field simulation data into a deep learning data set;
the model training module is used for training the pre-constructed deep learning network by taking part of data in the deep learning data set as a training set so as to obtain the position and corresponding current information of the transcranial direct current personalized stimulation target point by utilizing the trained deep learning network.
In a third aspect, the present invention also provides a deep learning-based transcranial direct current personalized stimulation system, comprising:
the target point positioning device is used for obtaining the position of the transcranial direct current stimulation target point and corresponding current information according to the individual anatomy structure magnetic resonance image of the object to be stimulated;
the stimulation device is used for making a personalized regulation and control scheme according to the position of the transcranial direct current stimulation target point and corresponding current information so as to perform personalized transcranial direct current stimulation treatment on the to-be-stimulated object;
the target positioning device comprises the transcranial direct current individuation stimulation target positioning device based on deep learning.
The invention has the beneficial effects that:
the invention provides a transcranial direct current personalized stimulation target positioning method based on deep learning, which comprises the steps of firstly acquiring individual anatomy structure magnetic resonance images, preprocessing and performing inverse electric field simulation calculation to obtain an inverse electric field simulation data set; and then combining the inverse electric field simulation data and the individual anatomy structure magnetic resonance image to perform deep learning training, and exploring the relation between the individual anatomy structure magnetic resonance image and the inverse electric field simulation result to obtain a deep learning network capable of rapidly and accurately predicting the tDCS stimulation target point position and current. The method avoids the complicated calculation process of the personalized reverse electric field simulation positioning required before the accurate positioning of the tDCS in the existing method, saves the time cost of the accurate positioning before the tDCS treatment, constructs a prediction network model based on the accurate target regulation of the tDCS of the individual anatomy structure magnetic resonance image by using deep learning, omits the technical requirement of clinical operators on the reverse electric field simulation technology learning, increases the feasibility of the accurate positioning of the tDCS and the customization of the tDCS personalized scheme in clinical application, and ensures that the personalized tDCS scheme based on the individual anatomy structure has practical significance.
The present invention will be described in further detail with reference to the accompanying drawings and examples.
Drawings
FIG. 1 is a schematic flow chart of a method for locating a target point of transcranial direct current personalized stimulation based on deep learning according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart of reverse electric field simulation according to an embodiment of the present invention;
FIG. 3 is a schematic flow chart of another method for locating a target point of transcranial direct current personalized stimulation based on deep learning according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of a device for positioning a target point of transcranial direct current personalized stimulation based on deep learning according to an embodiment of the present invention;
fig. 5 is a block diagram of a transcranial direct current personalized stimulation system based on deep learning according to an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to specific examples, but embodiments of the present invention are not limited thereto.
Example 1
The positioning method of the transcranial direct current personalized stimulation target point based on deep learning provided by the embodiment firstly collects individual anatomy structure magnetic resonance (Magnetic Resonance Imaging, MRI) images (also called individual nuclear magnetic structure images, MRI images or simply structural images in the embodiment), secondly pre-processes the individual nuclear magnetic structure images, then carries out tDCS inverse electric field simulation calculation, and finally carries out network model training on the inverse electric field simulation result and the individual nuclear magnetic structure images by using a deep learning algorithm, wherein the trained network can rapidly carry out tDCS accurate position prediction based on the nuclear magnetic structure images.
Referring to fig. 1, fig. 1 is a schematic flow chart of a method for positioning a target point of transcranial direct current personalized stimulation based on deep learning according to an embodiment of the present invention. The positioning method of the transcranial direct current personalized stimulation target point based on deep learning provided by the embodiment mainly comprises the following steps:
step 1: and acquiring a magnetic resonance image of the individual anatomy structure, and preprocessing the magnetic resonance image of the individual anatomy structure to obtain a high-quality structural image.
Specifically, a large number of individual nuclear magnetic structural images can be acquired through the MRI equipment, so that the subsequent training of the network is facilitated. It will be appreciated that for ease of subsequent processing, it is also necessary to convert DICOM (Digital Imaging and Communications in Medicine, digital imaging and communications in medicine, which is an international standard for medical images and related information) files of the original individual nuclear magnetic structural images into NIfTI (a file format for medical image processing and analysis) files for ease of preprocessing of the individual nuclear magnetic structural images.
Furthermore, after the magnetic resonance image of the individual anatomy is obtained, it is also required to perform a preprocessing to obtain a high-quality structural image.
In particular, the individual anatomy magnetic resonance image may be preprocessed as follows:
firstly, median filtering is carried out on individual anatomy structure magnetic resonance images, and filtered images are obtained. And then, carrying out offset field correction on the filtered image to obtain a high-quality structural image.
In the process of acquiring the MRI image, the acquired MRI image is interfered by noise and distortion due to the random interference and other factors in the imaging equipment, so that median filtering is needed to remove the salt and pepper noise in the MRI image, the image is clearer, and the edges and details of the image can be well reserved. In addition, since the gray scale distribution is not uniform when using the magnetic resonance scanning, the bias field correction technique is also required, and a high-quality structural image is finally obtained.
The specific implementation of the median filtering technique and the offset field correction technique mentioned herein may be implemented with reference to the related art, and this embodiment will not be described in detail herein.
Step 2: performing tDCS reverse electric field simulation on the high-quality structural image to obtain reverse electric field simulation data; the inverse electric field simulation data comprises an electric field simulation distribution diagram, tDCS electrode positions and corresponding current magnitudes.
Referring to fig. 2, fig. 2 is a schematic flow chart of reverse electric field simulation provided in an embodiment of the present invention, and step 2 may be implemented specifically by the following substeps:
21 Pre-processing the high-quality structural image to divide it into a plurality of partial tissue structures, constructing a head model based on the plurality of partial tissue structures, and generating a finite element file of the head model.
In general, the structural image can be divided into five tissue structures, namely scalp, skull, cerebrospinal fluid, gray matter and white matter; and then constructing a head model according to the five-part organization structure, and generating a finite element file.
22 The activation region of the standard brain map and the brain activation region related to cognitive functions and mental diseases are adopted as target point coordinates (also called target brain region coordinates), and the target point coordinates are converted from the standard space to the individual space, so that the target point coordinates under the individual space are obtained. The standard space is coordinate space corresponding to the standard brain atlas, and generally refers to MNI (Montreal Neurological Institute, montreal neuroscience) space. Reference is made to the related art for how to convert coordinates from standard space to individual space, and this embodiment is not specifically described herein.
Alternatively, the standard brain atlas in this embodiment may use AAL (Anatomical Automatic Labeling, automatic anatomical labeling) template, brodmann template, or brain network group atlas provided by the institute of automation of the department of chinese sciences, or the like. And brain activation areas associated with cognitive functions and mental disorders may be determined based on known research results.
23 The finite element file of the head model is subjected to inverse electric field simulation calculation based on the target point coordinates of interest in the individual space, so that an electric field simulation distribution diagram of each target point of interest, a tDCS electrode position corresponding to the maximum electric field value of the target brain region and a corresponding current are obtained.
In general, the inverse electric field simulation calculation is performed on the finite element file of the head model, and the method mainly comprises the following two steps.
a) And calculating a lead field matrix based on the constructed head model.
The specific calculation method of the lead field matrix comprises the following steps: a number of electrodes are placed on the constructed head model, the fields induced by each pair of electrodes (cathode-anode pair) are calculated separately and the constant return electrode is maintained, by which simulation a lead field matrix can be formed for calculating the electric field induced by any combination of electrodes.
Alternatively, the lead field matrix in this embodiment may be calculated from EEG10-10 (Electroencephalo-gram 10-10, electroencephalogram standard 10-10 system) electrode positions.
b) Setting a tDCS optimization structure based on a lead field matrix, selecting a lead field for optimization, and optimally designing by utilizing coordinates of interested targets in an individual space to obtain an electric field simulation distribution diagram of each interested target, and a corresponding tDCS electrode position and a corresponding current when an electric field value of a target brain region is maximum.
Specifically, the tDCS optimization structure is set first, and it is necessary to select a lead field to be used for optimization, set safety constraints, and limit the number of electrodes. Wherein, setting safety constraint is to set the simulation current and tissue conductivity; the simulation current can be set according to experience and user requirements, and in the tDCS technology, the simulation current is generally 1-2mA; the tissue conductivity is set prior to simulation, and is usually set empirically as the conductivity of the head model varies from tissue to tissue. The number of the electrodes can be set according to the number of the electrodes optimized according to actual needs, for example, 5 electrodes are required to be optimized, and the number of the electrodes is limited to be 5. Then, an optimization target, namely, a coordinate of a target of interest determined before is defined by using a tDCS target structure, a spherical ROI (region of interest ) with the diameter of 10mm is manufactured by taking the coordinate as a circle center based on MNI (Montreal Neurological Institute, montreal nerve research institute) coordinate of a tested space, and when the field intensity in the ROI reaches the maximum, the corresponding tDCS electrode position and current size are calculated, namely, the optimal electrode position and current.
It will be appreciated that single-objective optimization, where 1 spherical ROI is determined, as well as multi-objective optimization, where multiple spherical ROIs are determined, may be performed.
Through inverse electric field simulation, an electric field simulation distribution diagram based on each target point of the brain region of interest can be obtained, and the position and the current magnitude of the tDCS electrode corresponding to the maximum electric field value of the target brain region can be obtained.
Step 3: the individual anatomy magnetic resonance image and the inverse electric field simulation data are constructed as a deep learning dataset.
In general, the deep learning dataset may be divided into training and testing sets in a ratio of 7:3 or 8:2 to facilitate subsequent training and testing of the network model.
Step 4: and training the pre-constructed deep learning network by taking part of data in the deep learning data set as a training set so as to obtain the position of the transcranial direct current personalized stimulation target point and corresponding current information by utilizing the trained deep learning network.
First, a deep learning network model is constructed.
Specifically, the embodiment builds an improved Unet network model based on an attention mechanism based on a deep learning framework, and the network adds a residual convolution layer with one residual connection to each layer of convolution on the basis of an original Unet network. The attention mechanism can enable the deep learning model to pay more attention to the most relevant area or feature in the medical image, and the prediction performance is improved. The added residual convolution layer can further improve the training and convergence effects of the network.
Then, performing network training, wherein the training data is the training set divided in the step 3.
Specifically, in the embodiment, accurate tDCS positioning data calculated by inverse electric field simulation and individual anatomy structure magnetic resonance images are used for training a deep learning network model, model parameters are optimized through an Adam optimizer to provide better convergence performance and generalization capability, a cross entropy loss function (Cross Entropy Loss) is used as a loss function, the difference between a simulation result output by the model in the training process and a real simulation result label can be measured, and a learning rate is set to enable the model to be converged rapidly and obtain better performance. The accurate tDCS positioning data is data including tDCS electrode positions and current sizes, which are calculated by performing inverse electric field simulation on finite element files of different head models of testees based on target point coordinates of interest in individual space in the step 2. It can be understood that the electrode position and the current magnitude calculated by the inverse electric field simulation are also different for the same brain region of different subjects.
Through the training, a trained deep learning network is obtained, namely an optimal deep learning model for accurately realizing tDCS personalized stimulation target positioning.
Further, after obtaining the trained deep learning network, the method further comprises testing the model.
Specifically, the test set is used for performing performance evaluation on the trained deep learning network so as to test the accuracy of the network in predicting the position of the personalized tDCS electrode. The relevant evaluation index and the like can be realized by referring to the prior art.
After the test is passed, the network model can be used for prediction application.
In the subsequent prediction application, the structural image data of the patient can be directly acquired, the structural image data is input into a network model, and the optimal tDCS position coordinate and the current magnitude based on the target point of interest are output. And finally, according to the tDCS position coordinates and the current magnitude calculated by the prediction model, a personalized regulation and control scheme is formulated for personalized transcranial direct current stimulation treatment, and the whole flow chart is shown in figure 3.
The invention provides a transcranial direct current personalized stimulation target positioning method based on deep learning, which comprises the steps of firstly acquiring individual anatomy structure magnetic resonance images, preprocessing and performing inverse electric field simulation calculation to obtain an inverse electric field simulation data set; and then combining the inverse electric field simulation data and the individual anatomy structure magnetic resonance image to perform deep learning training, and exploring the relation between the individual anatomy structure magnetic resonance image and the inverse electric field simulation result to obtain a deep learning network capable of rapidly and accurately predicting the tDCS stimulation target point position and current. The method avoids the complicated calculation process of the personalized reverse electric field simulation positioning required before the accurate positioning of the tDCS in the existing method, saves the time cost of the accurate positioning before the tDCS treatment, constructs a prediction network model based on the accurate target regulation of the tDCS of the individual anatomy structure magnetic resonance image by using deep learning, omits the technical requirement of clinical operators on the reverse electric field simulation technology learning, increases the feasibility of the accurate positioning of the tDCS and the customization of the tDCS personalized scheme in clinical application, and ensures that the personalized tDCS scheme based on the individual anatomy structure has practical significance.
Example two
Based on the first embodiment, the present embodiment provides a transcranial direct current personalized stimulation target positioning device based on deep learning based on the same inventive concept. Fig. 4 is a schematic structural diagram of a deep learning-based transcranial direct current personalized stimulation target positioning device according to an embodiment of the present invention. The device specifically comprises:
the image processing module is used for acquiring individual anatomy structure magnetic resonance images and preprocessing the individual anatomy structure magnetic resonance images to obtain high-quality structural images;
the simulation module is used for performing tDCS reverse electric field simulation on the high-quality structural image to obtain reverse electric field simulation data; the inverse electric field simulation data comprise an electric field simulation distribution diagram, tDCS electrode positions and corresponding current magnitudes;
the data set construction module is used for constructing the individual anatomy structure magnetic resonance image and the inverse electric field simulation data into a deep learning data set;
the model training module is used for training the pre-constructed deep learning network by taking part of data in the deep learning data set as a training set so as to obtain the position and corresponding current information of the transcranial direct current personalized stimulation target point by utilizing the trained deep learning network.
The transcranial direct current personalized stimulation target positioning device provided by the embodiment can be used for realizing the method provided by the first embodiment, and the detailed process is described in the first embodiment. Therefore, the device can also realize accurate determination of the position and the current of the tDCS stimulation target point.
Example III
On the basis of the second embodiment, the present embodiment provides a transcranial direct current personalized stimulation system based on deep learning. Referring to fig. 5, fig. 5 is a block diagram of a deep learning-based transcranial direct current personalized stimulation system according to an embodiment of the present invention, including:
the target point positioning device is used for obtaining the position of the transcranial direct current stimulation target point and corresponding current information according to the individual anatomy structure magnetic resonance image of the object to be stimulated;
the stimulation device is used for making a personalized regulation and control scheme according to the position of the transcranial direct current stimulation target point and corresponding current information so as to perform personalized transcranial direct current stimulation treatment on the to-be-stimulated object;
the target positioning device comprises the transcranial direct current individuation stimulation target positioning device based on deep learning, which is provided by the second embodiment.
The foregoing is a further detailed description of the invention in connection with the preferred embodiments, and it is not intended that the invention be limited to the specific embodiments described. It will be apparent to those skilled in the art that several simple deductions or substitutions may be made without departing from the spirit of the invention, and these should be considered to be within the scope of the invention.
Claims (10)
1. The method for locating the transcranial direct current personalized stimulation target point based on deep learning is characterized by comprising the following steps of:
acquiring an individual anatomy structure magnetic resonance image, and preprocessing the individual anatomy structure magnetic resonance image to obtain a high-quality structural image;
performing tDCS reverse electric field simulation on the high-quality structural image to obtain reverse electric field simulation data; the reverse electric field simulation data comprise an electric field simulation distribution diagram, tDCS electrode positions and corresponding current magnitudes;
constructing the individual anatomy magnetic resonance image and the inverse electric field simulation data as a deep learning dataset;
and training the pre-constructed deep learning network by taking part of data in the deep learning data set as a training set so as to obtain the position of the transcranial direct current personalized stimulation target point and corresponding current information by utilizing the trained deep learning network.
2. The deep learning based transcranial direct current personalized stimulation target positioning method according to claim 1, wherein the individual anatomy magnetic resonance image is preprocessed according to the following steps:
median filtering is carried out on the individual anatomy structure magnetic resonance image to obtain a filtered image;
and carrying out offset field correction on the filtered image to obtain the high-quality structural image.
3. The method for locating a target point of transcranial direct current personalized stimulation based on deep learning according to claim 1, wherein the method is characterized in that the high-quality structural image is subjected to tDCS inverse electric field simulation to obtain inverse electric field simulation data, and specifically comprises the following steps:
dividing the high-quality structural image into a plurality of partial tissue structures, constructing a head model based on the partial tissue structures, and generating a finite element file of the head model;
the method comprises the steps of adopting an activation region of a standard brain map and a brain activation region related to cognitive functions and mental diseases as target point coordinates of interest, and converting the target point coordinates of interest from a standard space to an individual space to obtain target point coordinates of interest in the individual space;
and performing inverse electric field simulation calculation on the finite element file of the head model based on the target point coordinates of interest in the individual space to obtain an electric field simulation distribution map of each target point of interest, and a tDCS electrode position and a corresponding current size corresponding to the maximum target brain region electric field value.
4. The deep learning-based transcranial direct current personalized stimulation target positioning method according to claim 3, wherein the inverse electric field simulation calculation is performed on the finite element file of the head model based on the target point coordinate of interest in the individual space, and specifically comprises the following steps:
calculating a lead field matrix based on the constructed head model;
and setting a tDCS optimization structure based on the lead field matrix, selecting a lead field for optimization, and optimally designing by utilizing the coordinates of the target points of interest in the individual space to obtain an electric field simulation distribution diagram of each target point of interest, and the corresponding tDCS electrode position and the corresponding current when the electric field value of the target brain region is maximum.
5. The deep learning-based transcranial direct current personalized stimulation target positioning method according to claim 4, wherein the calculating of the lead field matrix based on the constructed head model specifically comprises:
electrodes are placed on the constructed head model, the fields induced by each pair of electrodes are calculated separately, and a constant return electrode is maintained, thereby forming a lead field matrix for calculating the electric field induced by any combination of electrodes.
6. The method for positioning the personalized stimulation target spot of the transcranial direct current based on deep learning according to claim 1, wherein the deep learning network adopts an improved Unet network model based on an attention mechanism; wherein the improved Unet network model includes a residual convolution layer that adds a layer of residual connection to each layer of convolution of the original Unet network.
7. The deep learning-based transcranial direct current personalized stimulation target positioning method according to claim 1, wherein the loss function adopted in training the pre-built deep learning network is a cross entropy loss function.
8. The deep learning-based transcranial direct current personalized stimulation target positioning method according to claim 1, further comprising, after obtaining the trained deep learning network:
performing performance evaluation on the trained deep learning network by using a test set to test the accuracy of the network prediction individuation tDCS electrode position; wherein the test set is data in the deep learning data set that does not include the training set.
9. A deep learning-based transcranial direct current personalized stimulation target positioning device for implementing a deep learning-based transcranial direct current personalized stimulation target positioning method according to any one of claims 1-8, comprising:
the image processing module is used for acquiring individual anatomy structure magnetic resonance images and preprocessing the individual anatomy structure magnetic resonance images to obtain high-quality structural images;
the simulation module is used for performing tDCS reverse electric field simulation on the high-quality structural image to obtain reverse electric field simulation data; the reverse electric field simulation data comprise an electric field simulation distribution diagram, tDCS electrode positions and corresponding current magnitudes;
a data set construction module for constructing the individual anatomy magnetic resonance image and the inverse electric field simulation data into a deep learning data set;
and the model training module is used for training the pre-constructed deep learning network by taking part of data in the deep learning data set as a training set so as to obtain the position and corresponding current information of the transcranial direct current personalized stimulation target point by using the trained deep learning network.
10. A deep learning-based transcranial direct current personalized stimulation system, comprising:
the target point positioning device is used for obtaining the position of the transcranial direct current stimulation target point and corresponding current information according to the individual anatomy structure magnetic resonance image of the object to be stimulated;
the stimulation device is used for making a personalized regulation and control scheme according to the position of the transcranial direct current stimulation target point and corresponding current information so as to perform personalized transcranial direct current stimulation treatment on the to-be-stimulated object;
wherein the target positioning device comprises the deep learning based transcranial direct current personalized stimulation target positioning device of claim 9.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410063719.3A CN117594193B (en) | 2024-01-17 | 2024-01-17 | Transcranial direct current personalized stimulation target positioning method based on deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410063719.3A CN117594193B (en) | 2024-01-17 | 2024-01-17 | Transcranial direct current personalized stimulation target positioning method based on deep learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN117594193A true CN117594193A (en) | 2024-02-23 |
CN117594193B CN117594193B (en) | 2024-06-18 |
Family
ID=89910193
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202410063719.3A Active CN117594193B (en) | 2024-01-17 | 2024-01-17 | Transcranial direct current personalized stimulation target positioning method based on deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117594193B (en) |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050004617A1 (en) * | 2003-04-28 | 2005-01-06 | Dawant Benoit M. | Apparatus and methods of optimal placement of deep brain stimulator |
CN110507904A (en) * | 2019-08-22 | 2019-11-29 | 西安八水健康科技有限公司 | A kind of method that computation-intensive array tDCS orientation modulates optimal electrode stimulating mode |
CN112232293A (en) * | 2020-11-09 | 2021-01-15 | 腾讯科技(深圳)有限公司 | Image processing model training method, image processing method and related equipment |
CN113058159A (en) * | 2021-06-03 | 2021-07-02 | 杭州回车电子科技有限公司 | Electrode wearing condition detection method and device for transcranial electrical stimulation |
CN113289249A (en) * | 2021-06-01 | 2021-08-24 | 西安科悦医疗股份有限公司 | Multi-target brain region accurate electrical stimulation method based on dense electrode array |
KR102321009B1 (en) * | 2021-07-06 | 2021-11-03 | 뉴로핏 주식회사 | Method, server and computer program for designing customized headgear for transcranial direct current stimulation |
CN114463493A (en) * | 2022-01-18 | 2022-05-10 | 武汉工程大学 | Transcranial magnetic stimulation electric field rapid imaging method and model based on coding and decoding structure |
CN116090294A (en) * | 2022-12-21 | 2023-05-09 | 中国科学院自动化研究所 | Brain response method, device, equipment and medium for personalized transcranial direct current stimulation |
CN116433967A (en) * | 2023-03-21 | 2023-07-14 | 南京脑科医院 | Personalized target spot selection method oriented to noninvasive nerve regulation technology |
CN117197521A (en) * | 2023-06-19 | 2023-12-08 | 深圳市联影高端医疗装备创新研究院 | Simulation image preprocessing method, local specific absorption rate estimation method and device |
-
2024
- 2024-01-17 CN CN202410063719.3A patent/CN117594193B/en active Active
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050004617A1 (en) * | 2003-04-28 | 2005-01-06 | Dawant Benoit M. | Apparatus and methods of optimal placement of deep brain stimulator |
CN110507904A (en) * | 2019-08-22 | 2019-11-29 | 西安八水健康科技有限公司 | A kind of method that computation-intensive array tDCS orientation modulates optimal electrode stimulating mode |
CN112232293A (en) * | 2020-11-09 | 2021-01-15 | 腾讯科技(深圳)有限公司 | Image processing model training method, image processing method and related equipment |
CN113289249A (en) * | 2021-06-01 | 2021-08-24 | 西安科悦医疗股份有限公司 | Multi-target brain region accurate electrical stimulation method based on dense electrode array |
CN113058159A (en) * | 2021-06-03 | 2021-07-02 | 杭州回车电子科技有限公司 | Electrode wearing condition detection method and device for transcranial electrical stimulation |
KR102321009B1 (en) * | 2021-07-06 | 2021-11-03 | 뉴로핏 주식회사 | Method, server and computer program for designing customized headgear for transcranial direct current stimulation |
CN114463493A (en) * | 2022-01-18 | 2022-05-10 | 武汉工程大学 | Transcranial magnetic stimulation electric field rapid imaging method and model based on coding and decoding structure |
CN116090294A (en) * | 2022-12-21 | 2023-05-09 | 中国科学院自动化研究所 | Brain response method, device, equipment and medium for personalized transcranial direct current stimulation |
CN116433967A (en) * | 2023-03-21 | 2023-07-14 | 南京脑科医院 | Personalized target spot selection method oriented to noninvasive nerve regulation technology |
CN117197521A (en) * | 2023-06-19 | 2023-12-08 | 深圳市联影高端医疗装备创新研究院 | Simulation image preprocessing method, local specific absorption rate estimation method and device |
Non-Patent Citations (2)
Title |
---|
徐子良: "高精度脑部经颅直流电刺激系统中关键问题研究", 《中国博士学位论文全文数据库 医药卫生科技辑》, 15 July 2020 (2020-07-15) * |
耿跃华 等: "前庭电刺激下大脑皮层电场强度分布仿真分析", 《实验室研究与探索》, vol. 42, no. 8, 31 August 2023 (2023-08-31), pages 88 - 92 * |
Also Published As
Publication number | Publication date |
---|---|
CN117594193B (en) | 2024-06-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Huang et al. | Automated MRI segmentation for individualized modeling of current flow in the human head | |
Seibt et al. | The pursuit of DLPFC: non-neuronavigated methods to target the left dorsolateral pre-frontal cortex with symmetric bicephalic transcranial direct current stimulation (tDCS) | |
Nourski et al. | Functional organization of human auditory cortex: investigation of response latencies through direct recordings | |
US9307925B2 (en) | Methods and systems for generating electrical property maps of biological structures | |
CN113367679B (en) | Target point determination method, device, equipment and storage medium | |
US11775698B2 (en) | Method, server and computer program for designing customized headgear for transcranial direct current stimulation | |
CN113367681B (en) | Target point determination method, device, equipment and storage medium | |
WO2023280003A1 (en) | Target determination method and apparatus, electronic device, storage medium and neuromodulation device | |
CN111311703B (en) | Electrical impedance tomography image reconstruction method based on deep learning | |
CN106485039B (en) | A kind of Chinese brain language distinguishes the construction method of Butut | |
CN105395194B (en) | A kind of brain electric channel system of selection of functional mri auxiliary | |
Rashed et al. | End-to-end semantic segmentation of personalized deep brain structures for non-invasive brain stimulation | |
CN113289249A (en) | Multi-target brain region accurate electrical stimulation method based on dense electrode array | |
Conte et al. | Cortical source analysis of event-related potentials: a developmental approach | |
CN117594193B (en) | Transcranial direct current personalized stimulation target positioning method based on deep learning | |
US20230011442A1 (en) | Optimal stimulation position combination determination method, server, and computer program using preset guide system | |
US20230038541A1 (en) | Brain stimulation simulation system and method according to preset guide system using anonymized data-based external server | |
CN117934726B (en) | Three-dimensional visualization method, apparatus and system, and readable storage medium | |
CN117454692A (en) | tDCS electric field simulation image generation method and device based on deep learning network | |
US20230010674A1 (en) | Electric stimulation simulation method, server, and computer program for determining optimal stimulation position combination | |
Pancholi et al. | Analysis of electric field strengths and focality for healthy and neurologically impaired subjects upon multiple tDCS stimulation protocols | |
KR20230007650A (en) | Method, server and computer program for designing customized headgear for transcranial direct current stimulation | |
Baeken et al. | POTENTIAL OF ELECTRIC FIELD SIMULATIONS IN CLINICAL PRACTICE | |
Cheng et al. | Design and Implementation of Integrated Simulation Software for Multichannel Modified Electroconvulsive Therapy | |
CN117562552A (en) | Transcranial magnetic stimulation electroencephalogram positioning method based on multi-feature fusion |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |