Real-time functional magnetic resonance neural feedback regulation and control device based on decoding

Abstract

A decoding-based real-time functional magnetic resonance neural feedback regulation and control device disclosed by the present invention comprises a casing, a first half sleeve and a second half sleeve, a processor and a memory are installed on the inner wall of the casing, a network interface is installed on the outer wall of the casing, an external thread sleeve is fixedly arranged on the outer wall of the casing, and the first half sleeve and the second half sleeve are in threaded connection. The outer wall of the outer threaded sleeve is in threaded connection with an inner threaded sleeve, two rotating shafts are fixedly arranged on the circumferential outer wall of the inner threaded sleeve in a symmetrical structure, the first half sleeve and the second half sleeve are rotationally connected with the two rotating shafts respectively, the sections of the first half sleeve and the second half sleeve are in magnetic attraction connection, and two clamping feet are fixedly arranged on the inner wall of each of the first half sleeve and the second half sleeve. The first half sleeve and the second half sleeve are pinched to drive the internal thread sleeve to rotate, at the moment, the abutting surfaces on the clamping feet can extrude the connecting end of the crystal head and the cable, and therefore the crystal head can be prevented from loosening in the repeated use process.

Description

Technical Field

[0001] The invention relates to the technical field of magnetic resonance neural feedback control devices, in particular to a real-time functional magnetic resonance neural feedback control device based on decoding. Background Art

[0002] The processing of working memory is divided into three stages: encoding, consolidation, and recall. Whether a piece of information is remembered or forgotten depends on brain activity before and during memory encoding. Functional MRI and EEG studies have shown that the activities of many cortical and subcortical brain regions such as the dorsolateral prefrontal cortex, cingulate gyrus, hippocampus, and temporal lobe may predict the information that is remembered and forgotten. Brain regions such as the prefrontal cortex and medial temporal lobe are crucial for sequence memory. For example, patients with PFC damage are often impaired in tasks involving memory of the order in which items are presented. In 2008, a study used the activation patterns of subjects during the memory encoding stage to predict whether the subjects remembered indoor and outdoor scenes, and the prediction accuracy was greater than the chance probability. Other studies have pointed out that the brain state that occurs before the memory encoding stimulus affects the outcome of the memory, and the parahippocampal gyrus has the greatest impact on memory. When the subjects are detected in real time entering their "good" or "bad" memory brain state, different experimental stimulation scenarios are triggered, and the subjects remember the scenarios triggered by the "good" brain state more accurately than the "bad" brain state. Therefore, by using magnetic resonance imaging data from the encoding phase to classify memory outcomes, subjects may become more aware of when they are in and how to enter a better memory state.

[0003] In the existing real-time functional magnetic resonance neural feedback control device based on decoding, the device includes a network interface for sending and receiving data. The network interface is usually connected through a cable, and a crystal head for connecting to the network interface is usually installed at the end of the cable. The crystal head is connected to the network interface through a card connection. During the multiple plugging and unplugging of the crystal head, the wear between the structures may cause looseness during the installation process. In view of this, we propose a real-time functional magnetic resonance neural feedback control device based on decoding. Summary of the invention

[0004] The present application aims to solve at least one of the technical problems existing in the prior art. To this end, the present application proposes a real-time functional magnetic resonance neural feedback control device based on decoding, which drives the internal threaded sleeve to rotate by pinching the first half sleeve and the second half sleeve, and the abutting surfaces on the multiple clamping pins can squeeze the crystal head and the cable connection end, thereby preventing the crystal head from loosening during repeated use.

[0005] To solve the above technical problems, the present invention provides the following technical solutions: a real-time functional magnetic resonance neural feedback control device based on decoding, comprising a casing, a first half sleeve and a second half sleeve, the inner wall of the casing is equipped with a processor and a memory, the outer wall of the casing is equipped with a network interface, the outer wall of the casing is fixedly provided with an external threaded sleeve, the outer wall of the external threaded sleeve is threadedly connected with an internal threaded sleeve, the circumferential outer wall of the internal threaded sleeve is symmetrically structured and fixed with two rotating shafts, the first half sleeve and the second half sleeve are respectively rotatably connected to the two rotating shafts, the first half sleeve and the second half sleeve are magnetically connected in cross-section, and the inner walls of the first half sleeve and the second half sleeve are both fixed with two clamping feet.

[0006] As a preferred solution of the real-time functional magnetic resonance neural feedback control device based on decoding described in the present invention, wherein: the clamping pin includes a connecting portion, an inclined portion and a butt joint portion, the upper end of the inclined portion is connected and fixed to the connecting portion, the bottom of the inclined portion is connected and fixed to the butt joint portion, the butt joint portion is provided with a butt joint surface on the inner side, and the inner side wall of the butt joint portion is provided with a curved surface.

[0007] As a preferred solution of the real-time functional magnetic resonance neural feedback control device based on decoding described in the present invention, it also includes a behavioral module, which is installed inside the casing and is used to measure the working memory level of the subject and obtain the subject's memory accuracy.

[0008] As a preferred solution of the real-time functional magnetic resonance neural feedback control device based on decoding described in the present invention, it also includes a decoding model training module, which is installed inside the casing. The decoding model training module is used to use a magnetic resonance machine to collect short-sequence memory experiment magnetic resonance data and baseline experiment magnetic resonance data of each subject, and obtain memory state and non-memory state as decoding labels respectively, and convert the short-sequence memory experiment magnetic resonance data and the baseline experiment magnetic resonance data into graph model data. The graph model data and the decoding labels are used as inputs of the decoding model for model training to obtain the optimal decoding model.

[0009] As a preferred solution of the real-time functional magnetic resonance neural feedback control device based on decoding described in the present invention, it also includes a decoding model training module, which is installed inside the casing. The decoding model training module includes a sequence experiment unit, a model training unit, a graph model data processing unit and a decoding label setting unit. The sequence experiment unit is used to use a magnetic resonance machine to collect short-sequence memory experiment magnetic resonance data of each subject, and obtain the short-sequence memory experiment magnetic resonance data corresponding to the experimental block in which the relative positions of different pictures arranged side by side are correct by pressing keys as the memory state, and use the magnetic resonance machine to collect baseline experiment magnetic resonance data of each subject, and use the magnetic resonance machine to judge that the direction of the same pictures arranged side by side is correct. The baseline experiment magnetic resonance data corresponding to the experimental block is used as the non-memory state.

[0010] As a preferred embodiment of the real-time functional magnetic resonance neural feedback control device based on decoding described in the present invention, the short sequence memory experiment is specifically as follows: the subject performs sequence encoding memory on several different pictures presented side by side, and the relative position order of the different pictures in parallel is determined according to the key presses of the subject, and the picture sequence results represented by the key presses and the short sequence memory experiment magnetic resonance data are obtained, and the short sequence memory experiment magnetic resonance data corresponding to the sequence encoding memory stage time point of the picture sequence result for the correct experimental block is used as the memory state; the baseline experiment is specifically as follows: the subject watches several identical pictures presented side by side without performing encoding memory, and the direction position of the identical pictures in parallel is determined according to the key presses of the subject, and the picture direction results represented by the key presses and the baseline experiment magnetic resonance data are obtained, and the baseline experiment magnetic resonance data corresponding to the viewing stage time point of the picture direction result for the correct experimental block is used as the non-memory state.

[0011] As a preferred solution of the real-time functional magnetic resonance neural feedback control device based on decoding described in the present invention, wherein: the model training unit is used to preprocess the short sequence memory experiment magnetic resonance data and the baseline test magnetic resonance data, and convert the preprocessed data into graph model data, use the memory state and the non-memory state as decoding labels, use the graph model data and the decoding labels to train the decoding model, obtain the optimal decoding model by adjusting the parameters, and use the graph model data and the decoding labels in the model training unit to train the decoding model.

[0012] As a preferred solution of the real-time functional magnetic resonance neural feedback control device based on decoding described in the present invention, wherein: obtaining the optimal decoding model by adjusting parameters specifically includes a graph model data processing unit, and the graph model data processing unit is used to divide the graph model data into different brain regions using a brain map template, calculate the time series corresponding to each brain region, and generate graph model feature data of the time series corresponding to each subject multiplied by the number of brain regions, and the graph model feature data of each subject is VεR^(t×n), where t is the duration of each trial and n is the number of brain regions.

[0013] As a preferred solution of the real-time functional magnetic resonance neural feedback control device based on decoding of the present invention, wherein: the decoding label setting unit is used to set the memory state to a memory label, the non-memory state to a non-memory label, and use the memory label and the non-memory label as decoding labels;

[0014] The decoding model training unit is used to input the graph model feature data and the decoding label into the decoding model, calculate the accuracy of the decoding model by adjusting the learning rate, the number of parameters and the rounds, and determine the decoding model whose accuracy no longer decreases as the optimal decoding model; the decoding model includes a multi-layer perceptron for deep learning, a graph neural network, a long short-term memory network, and a support vector machine for machine learning.

[0015] As a preferred solution of the real-time functional magnetic resonance neural feedback control device based on decoding described in the present invention, it also includes a real-time decoding neural feedback module, which is installed inside the casing. The real-time decoding neural feedback module is used to use the optimal decoding model to predict the real-time state information and real-time accuracy of the subject including memory state and non-memory state in real time, and feed back to the subject in the form of different pictures. The subject uses the adjustment strategy to adjust the feedback information and autonomously train the working memory ability.

[0016] Compared with the prior art, the present invention has the following beneficial effects:

[0017] 1. By pinching the first half sleeve and the second half sleeve, the internal thread sleeve is driven to rotate. At this time, the abutment surfaces on the multiple clamping feet can squeeze the crystal head and the cable connection end, thereby preventing the crystal head from loosening during repeated use.

[0018] 2. By moving the first half set and the second half set to both sides, the first half set and the second half set are rotated along the rotating shaft respectively. At this time, the first half set and the second half set are in an open state, so that the operator can easily insert the crystal head at the end of the cable into the network interface, thereby facilitating the installation and removal of the crystal head.

[0019] 3. Use decoded real-time functional magnetic resonance neurofeedback technology to enhance or improve the subjects' working memory level and enhance the subjects' self-control over the attention network and working memory network. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a schematic diagram of the overall structure of the present invention;

[0021] Figure 2 It is a schematic diagram of the installation structure of the external threaded sleeve of the present invention;

[0022] Figure 3 This is a schematic diagram of the installation structure of the first half of the present invention;

[0023] Figure 4 It is a schematic diagram of the overall structure of the clamping foot of the present invention.

[0024] Explanation of the numbers in the figure: 1. housing; 2. behavioral module; 3. decoding model training module; 4. real-time decoding neural feedback module; 5. processor; 6. network interface; 7. memory; 8. external threaded sleeve; 9. internal threaded sleeve; 10. shaft; 11. first half sleeve; 12. second half sleeve; 13. clamping foot; 1301. connecting part; 1302. inclined part; 1303. abutting part; 1304. abutting surface; 1305. arc surface. DETAILED DESCRIPTION

[0025] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0026] Embodiment 1

[0027] like Figure 1 , Figure 2 , Figure 3 and Figure 4 As shown, a real-time functional magnetic resonance neural feedback control device based on decoding includes a casing 1, a first half casing 11 and a second half casing 12, a processor 5 and a memory 7 are installed on the inner wall of the casing 1, a network interface 6 is installed on the outer wall of the casing 1, an external threaded sleeve 8 is fixedly provided on the outer wall of the casing 1, an internal threaded sleeve 9 is threadedly connected to the outer wall of the external threaded sleeve 8, and two rotating shafts 10 are fixedly provided on the circumferential outer wall of the internal threaded sleeve 9 in a symmetrical structure, the first half casing 11 and the second half casing 12 are rotatably connected to the two rotating shafts 10 respectively, the first half casing 11 and the second half casing 12 are magnetically connected in cross section, and two clamping feet 13 are fixedly provided on the inner walls of the first half casing 11 and the second half casing 12.

[0028] By moving the first half sleeve 11 and the second half sleeve 12 to both sides, the first half sleeve 11 and the second half sleeve 12 are respectively rotated along the rotating shaft 10. At this time, the first half sleeve 11 and the second half sleeve 12 are in an open state, so that it is convenient for the operator to insert the crystal head at the end of the cable into the network interface 6, thereby facilitating the installation and removal of the crystal head.

[0029] The clamping foot 13 includes a connecting portion 1301, an inclined portion 1302 and a butting portion 1303. The upper end of the inclined portion 1302 is connected and fixed to the connecting portion 1301, and the bottom of the inclined portion 1302 is connected and fixed to the butting portion 1303. The butting portion 1303 is provided with a butting surface 1304 on the inner side, and the inner wall of the butting portion 1303 is provided with a curved surface 1305.

[0030] By pinching the first half sleeve 11 and the second half sleeve 12, the internal thread sleeve 9 is driven to rotate. At this time, the abutment surfaces 1304 on the multiple clamping feet 13 can squeeze the crystal plug and the cable connection end, thereby preventing the crystal plug from loosening during repeated use.

[0031] Embodiment 2

[0032] like Figure 1 As shown, this embodiment is based on the previous embodiment, and differs from the previous embodiment in that: a real-time functional magnetic resonance neural feedback control device based on decoding, also includes a behavioral module 2, and the behavioral module 2 is installed inside the housing 1. The behavioral module 2 is used to measure the working memory level of the subject and obtain the subject's memory accuracy.

[0033] Embodiment 3

[0034] like Figure 1 As shown, this embodiment is based on the previous embodiment, and is different from the previous embodiment in that: a real-time functional magnetic resonance neural feedback control device based on decoding, also includes a decoding model training module 3, the decoding model training module 3 is installed inside the housing 1, and the decoding model training module 3 is used to use a magnetic resonance machine to collect short-sequence memory experiment magnetic resonance data and baseline experiment magnetic resonance data of each subject, respectively obtain memory state and non-memory state as decoding labels, convert the short-sequence memory experiment magnetic resonance data and baseline experiment magnetic resonance data into graph model data, and use the graph model data and the decoding label as the input of the decoding model for model training to obtain the optimal decoding model.

[0035] Embodiment 4

[0036] like Figure 1As shown, this embodiment is based on the previous embodiment, and is different from the previous embodiment in that: a real-time functional magnetic resonance neural feedback control device based on decoding, further comprising a decoding model training module 3, the decoding model training module 3 being installed inside the housing 1, the decoding model training module 3 comprising a sequence experiment unit, a model training unit, a graph model data processing unit and a decoding label setting unit, the sequence experiment unit being used to collect short sequence memory experiment magnetic resonance data of each subject using a magnetic resonance machine, and obtaining short sequence memory experiment magnetic resonance data corresponding to an experiment block in which the relative positions of different pictures arranged side by side are judged to be correct by pressing a button as a memory state, and baseline experiment magnetic resonance data of each subject being collected using a magnetic resonance machine, and baseline experiment magnetic resonance data corresponding to an experiment block in which the directions of the same pictures arranged side by side are judged to be correct by pressing a button as a non-memory state;

[0037] The short sequence memory experiment is as follows: the subjects perform sequence encoding memory on several different pictures presented side by side, and the relative position order of the different pictures in parallel is judged according to the subjects' key presses, and the picture order results represented by the key presses and the short sequence memory experiment magnetic resonance data are obtained, and the short sequence memory experiment magnetic resonance data corresponding to the sequence encoding memory stage time point of the experimental block with the correct picture order result is used as the memory state; the baseline experiment is as follows: the subjects watch several identical pictures presented side by side without performing encoding memory, and the direction position of the identical pictures in parallel is judged according to the subjects' key presses, and the picture direction results represented by the key presses and the baseline experiment magnetic resonance data are obtained, and the baseline experiment magnetic resonance data corresponding to the viewing stage time point of the experimental block with the correct picture direction result is used as the non-memory state.

[0038] The model training unit is used to preprocess the short sequence memory experiment magnetic resonance data and the baseline test magnetic resonance data, and convert the preprocessed data into graph model data, use the memory state and non-memory state as decoding labels, and use the graph model data and decoding labels to train the decoding model. The optimal decoding model is obtained by adjusting the parameters. The model training unit uses the graph model data and decoding labels to train the decoding model.

[0039] The optimal decoding model is obtained by adjusting parameters and specifically includes a graph model data processing unit. The graph model data processing unit is used to divide the graph model data into different brain regions using a brain map template, calculate the time series corresponding to each brain region, and generate graph model feature data of each subject by multiplying the time series corresponding to each subject by the number of brain regions. The graph model feature data of each subject is VεR^(t×n), where t is the duration of each trial and n is the number of brain regions.

[0040] The decoding label setting unit is used to set the memory state to the memory label, the non-memory state to the non-memory label, and use the memory label and the non-memory label as the decoding label;

[0041] The decoding model training unit is used to input the graph model feature data and decoding labels into the decoding model, calculate the accuracy of the decoding model by adjusting the learning rate, number of parameters and rounds, and determine the decoding model whose accuracy no longer decreases as the optimal decoding model; the decoding models include multi-layer perceptrons for deep learning, graph neural networks, long short-term memory networks, and support vector machines for machine learning.

[0042] Embodiment 5

[0043] like Figure 1 As shown, this embodiment is based on the previous embodiment, and the difference from the previous embodiment is: a real-time functional magnetic resonance neural feedback control device based on decoding, also includes a real-time decoding neural feedback module 4, the real-time decoding neural feedback module 4 is installed inside the housing 1, the real-time decoding neural feedback module 4 is used to use the optimal decoding model to predict the real-time state information and real-time accuracy of the subject including memory state and non-memory state in real time, and feed back to the subject in the form of different pictures. The subject uses the adjustment strategy to adjust the feedback information and autonomously train the working memory ability.

[0044] Working principle: by pushing the first half sleeve 11 and the second half sleeve 12 to both sides, the first half sleeve 11 and the second half sleeve 12 are respectively rotated along the rotating shaft 10, at which time the first half sleeve 11 and the second half sleeve 12 are in an open state, and during installation, the crystal head at the end of the cable is inserted into the network interface 6 on the casing 1, and the first half sleeve 11 and the second half sleeve 12 are respectively flipped over so that the first half sleeve 11 and the second half sleeve 12 are attracted to each other, at which time the clamping pins 13 inside the first half sleeve 11 and the second half sleeve 12 can squeeze the outer wall of the cable, and by pinching the first half sleeve 11 and the second half sleeve 12, the internal threaded sleeve 9 is driven to rotate, and at this time the abutting surfaces 1304 on the multiple clamping pins 13 can squeeze the crystal head and the cable connection end, thereby preventing the crystal head from loosening during repeated use.

[0045] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A real-time functional magnetic resonance neural feedback control device based on decoding, characterized in that: include: A casing, a first half sleeve and a second half sleeve: a processor and a memory are installed on the inner wall of the casing, a network interface is installed on the outer wall of the casing, an external threaded sleeve is fixedly provided on the outer wall of the casing, an internal threaded sleeve is threadedly connected to the outer wall of the external threaded sleeve, and two rotating shafts are fixedly provided on the circumferential outer wall of the internal threaded sleeve in a symmetrical structure, the first half sleeve and the second half sleeve are rotatably connected to the two rotating shafts respectively, the first half sleeve and the second half sleeve are magnetically connected in cross section, and two clamping feet are fixedly provided on the inner walls of the first half sleeve and the second half sleeve.

2. The real-time functional magnetic resonance neural feedback control device based on decoding according to claim 1, characterized in that: The clamping foot comprises a connecting portion (1301), an inclined portion (1302) and a contact portion (1303); the upper end of the inclined portion (1302) is connected and fixed to the connecting portion (1301); the bottom of the inclined portion (1302) is connected and fixed to the contact portion (1303); an abutting surface (1304) is provided on the inner side of the contact portion (1303); and an arc surface (1305) is provided on the inner side wall of the contact portion (1303).

3. The real-time functional magnetic resonance neural feedback control device based on decoding according to claim 2, characterized in that: It also includes a behavioral module, which is installed inside the casing and is used to measure the working memory level of the subject and obtain the memory accuracy of the subject.

4. The real-time functional magnetic resonance neural feedback control device based on decoding according to claim 3, characterized in that: It also includes a decoding model training module, which is installed inside the casing. The decoding model training module is used to use a magnetic resonance machine to collect short-sequence memory experiment magnetic resonance data and baseline experiment magnetic resonance data of each subject, respectively obtain memory state and non-memory state as decoding labels, convert the short-sequence memory experiment magnetic resonance data and the baseline experiment magnetic resonance data into graph model data, and use the graph model data and the decoding labels as inputs of the decoding model for model training to obtain the optimal decoding model.

5. The real-time functional magnetic resonance neural feedback control device based on decoding according to claim 4, characterized in that: It also includes a decoding model training module, which is installed inside the casing. The decoding model training module includes a sequence experiment unit, a model training unit, a graph model data processing unit and a decoding label setting unit. The sequence experiment unit is used to use a magnetic resonance machine to collect short-sequence memory experiment magnetic resonance data of each subject, and obtain the short-sequence memory experiment magnetic resonance data corresponding to the experimental block in which the relative positions of different pictures placed side by side are correct by pressing keys as the memory state. The baseline experiment magnetic resonance data of each subject is collected by the magnetic resonance machine, and the baseline experiment magnetic resonance data corresponding to the experimental block in which the directions of the same pictures placed side by side are correct by pressing keys is used as the non-memory state.

6. The real-time functional magnetic resonance neural feedback control device based on decoding according to claim 5, characterized in that: The short sequence memory experiment is specifically as follows: the subjects perform sequence encoding memory on several different pictures presented side by side, and the relative position order of the different pictures in parallel is determined according to the subjects' key presses, and the picture sequence results represented by the keys and the short sequence memory experiment magnetic resonance data are obtained, and the short sequence memory experiment magnetic resonance data corresponding to the sequence encoding memory phase time point of the experimental block with the picture sequence result being the correct one is used as the memory state; the baseline experiment is specifically as follows: the subjects view several identical pictures presented side by side without performing encoding memory, and the direction position of the identical pictures in parallel is determined according to the subjects' key presses, and the picture direction results represented by the keys and the baseline experiment magnetic resonance data are obtained, and the baseline experiment magnetic resonance data corresponding to the viewing phase time point of the experimental block with the picture direction result being the correct one is used as the non-memory state.

7. The real-time functional magnetic resonance neural feedback control device based on decoding according to claim 6, characterized in that: The model training unit is used to preprocess the short sequence memory experiment magnetic resonance data and the baseline test magnetic resonance data, and convert the preprocessed data into graph model data, use the memory state and the non-memory state as decoding labels, use the graph model data and the decoding labels to train the decoding model, and obtain the optimal decoding model by adjusting parameters. The model training unit uses the graph model data and the decoding labels to train the decoding model.

8. The real-time functional magnetic resonance neural feedback control device based on decoding according to claim 7, characterized in that: The optimal decoding model obtained by adjusting parameters specifically includes a graph model data processing unit, which is used to divide the graph model data into different brain regions using a brain map template, calculate the time series corresponding to each brain region, and generate graph model feature data of each subject by multiplying the time series corresponding to each subject by the number of brain regions. The graph model feature data of each subject is VεR^(t×n), where t is the duration of each trial, and n is the number of brain regions.

9. The real-time functional magnetic resonance neural feedback control device based on decoding according to claim 8, characterized in that: The decoding tag setting unit is used to set the memory state to a memory tag, the non-memory state to a non-memory tag, and use the memory tag and the non-memory tag as decoding tags; The decoding model training unit is used to input the graph model feature data and the decoding label into the decoding model, calculate the accuracy of the decoding model by adjusting the learning rate, the number of parameters and the rounds, and determine the decoding model whose accuracy no longer decreases as the optimal decoding model; the decoding model includes a multi-layer perceptron for deep learning, a graph neural network, a long short-term memory network, and a support vector machine for machine learning.

10. The real-time functional magnetic resonance neural feedback control device based on decoding according to claim 9, characterized in that: It also includes a real-time decoding neural feedback module, which is installed inside the casing. The real-time decoding neural feedback module is used to use the optimal decoding model to predict the real-time state information and real-time accuracy of the subject including memory state and non-memory state, and feed back to the subject in the form of different pictures. The subject uses the adjustment strategy to adjust the feedback information and autonomously train the working memory ability.

Images