Apparatus and method for predicting the effect of acupuncture treatment using brain imaging signals

Abstract

The present invention relates to an apparatus and method for predicting the effect of acupuncture treatment using brain imaging signals. The method and apparatus according to the present invention can accurately predict the effect of acupuncture treatment appearing from a subject solely through brain imaging, thereby enabling the presentation of appropriate treatment to the patient and serving as a basis for the effect of non-pharmacological treatments such as acupuncture.

Description

Technology Field

[0001] The present invention relates to an apparatus and method for predicting the effect of acupuncture treatment using brain imaging signals. The method according to the present invention makes it possible to accurately predict the effect of acupuncture treatment on a subject using only brain imaging signals without using invasive methods. Background Technology

[0002] In Korean traditional medicine, acupuncture has been used for a thousand years to treat various diseases and alleviate symptoms. Acupuncture involves inserting needles into specific acupoints, which are traditionally used to regulate Qi, the body's energy flow. In modern medicine, acupuncture is recognized for its effectiveness in various applications, including pain management, stress relief, and the treatment of neurological disorders.

[0003] However, the effectiveness of acupuncture treatment varies significantly from patient to patient, depending on various factors such as the practitioner's skill level, treatment methods, and the patient's individual response. Consequently, it is difficult to guarantee predictable outcomes. Therefore, there is a need for methods to accurately predict and optimize the effectiveness of acupuncture treatment.

[0004] In this regard, brain imaging technology has shown significant progress in recent years. Brain imaging techniques, such as functional magnetic resonance imaging (fMRI) or electroencephalography (EEG), are used to visualize and understand the brain's real-time activity, enabling an understanding of the brain's health status, structure, and function. These technologies are being used as important tools for the diagnosis and treatment of brain and nervous system diseases.

[0005] Therefore, it is important to utilize these brain imaging technologies to more accurately predict the effects of acupuncture treatment and, through this, develop enhanced personalized treatment methods. This increases the predictability of treatment outcomes and helps in selecting and applying the acupuncture treatment most suitable for the patient's health condition.

[0006] However, this approach remains challenging due to the complexity of brain imaging signals and the difficulty of interpretation. Brain imaging data is high-dimensional and features complex temporal and spatial patterns. Analyzing and interpreting such data requires a high level of expertise and complex computational tools.

[0007] Therefore, reliable technology is required to effectively analyze brain imaging data and accurately predict the effects of acupuncture treatment based on this analysis; such technology can play a crucial role in complementing and improving existing acupuncture methods and maximizing their efficacy. Prior art literature

[0008] Korean Published Patent No. 2021-0016171

[0009] Yin, Tao, et al. "Predicting acupuncture efficacy for functional dyspepsia based on functional brain network features: a machine learning study." Cerebral Cortex 33.7 (2023): 3511-3522. The problem to be solved

[0010] The inventors have implemented an apparatus and method for predicting the effects of acupuncture treatment using brain imaging signals, and confirmed that the accuracy of predicting the effects of acupuncture treatment through this is excellent.

[0011] Accordingly, one objective of the present invention is to provide a method for predicting the effect of acupuncture treatment.

[0012] Another objective of the present invention is to provide a computer program stored in a storage medium that predicts the effect of acupuncture treatment.

[0013] Another objective of the present invention is to provide a computing device for predicting the responsiveness of acupuncture treatment. means of solving the problem

[0014] The present invention relates to an apparatus and method for predicting the effect of acupuncture treatment using brain imaging signals. The apparatus and method according to the present invention make it possible to accurately predict the effect of acupuncture treatment on a subject using only brain imaging.

[0015] The present invention will be described in more detail below.

[0017] One aspect of the present invention is a method for predicting the responsiveness of acupuncture treatment from medical imaging data by a computing device, comprising: a processing step of generating input data based on medical data including a brain image of a subject who has received acupuncture treatment; a generation step of generating one or more prediction information by inputting the input data into a pre-trained responsiveness prediction model; and a classification step of classifying a high-response group and a low-response group based on the prediction information.

[0018] The term “medical image data” in this specification may include multidimensional medical image data composed of discrete image elements (e.g., pixels in the case of a two-dimensional image). Specifically, the medical data may include a visible object or a digital representation of the object (e.g., a file corresponding to the pixel output of a CT, MRI detector, etc.). The medical data may include medical images of a subject collected by computed tomography (CT), magnetic resonance imaging (MRI), functional magnetic resonance imaging (fMRI), fundus imaging, ultrasound, or any other medical imaging system known in the art of the present invention.

[0019] The term “subject” in this specification may be an object (or subject) having a specific disease or suspected of having a specific disease. In this case, the specific disease may be atopic dermatitis. In one embodiment of the present invention, the subject may be a mammal, and for example, the subject may be one or more selected from the group consisting of humans, monkeys, dogs, cats, mice, rats, cattle, horses, pigs, goats, and sheep.

[0020] In one embodiment of the present invention, the learned reactivity prediction model may be a deep neural network.

[0021] The term “deep neural network” as used herein may refer to a neural network comprising a plurality of hidden layers in addition to an input layer and an output layer. Throughout this specification, the term “deep neural network” may be used interchangeably with the terms “neural network,” “network function,” and “neural network.” Using a deep neural network makes it possible to identify the latent structures of data. That is, it is possible to identify the latent structures of photos, text, video, voice, and music (e.g., what objects are in a photo, what is the content and emotion of the text, what is the content and emotion of the voice, etc.). In the present invention, the deep neural network may include, but is not limited to, a convolutional neural network (CNN), a recurrent neural network (RNN), a long-short-term memory algorithm, a restricted boltzmann machine (RBM), a deep belief network (DBN), a Q network, a U network, a sham network, etc.

[0022] In one embodiment of the present invention, the medical imaging data may include brain magnetic resonance imaging (MRI) data.

[0023] In one embodiment of the present invention, the magnetic resonance image may be functional magnetic resonance imaging (fMRI) data.

[0024] In one embodiment of the present invention, a reactive prediction model is supervised based on one or more training images and guide labels corresponding to the training images and including prediction results, and the supervised learning may be performed based on the comparison result between the training information generated for the training images using the reactive prediction model and the guide labels.

[0025] In one embodiment of the present invention, the brain image may be a brain image in a resting state.

[0026] The term “resting-state brain imaging” as used in this specification refers to images of neural activity of the brain obtained when a subject is in a resting state without performing a specific task. Resting-state brain imaging can be measured using brain imaging techniques such as functional magnetic resonance imaging (fMRI) or electroencephalography (EEG).

[0027] In one embodiment of the present invention, the classification step may classify targets whose SCORAD value decreases by 8.7 or more into a high response group and targets whose SCORAD value decreases by less than 8.7 into a low response group.

[0028] The term “SCORAD” in this specification is an abbreviation for “SCORing Atopic Dermatitis,” which is a scale for evaluating the severity of atopic dermatitis and is expressed numerically by considering the range of symptoms, severity of symptoms, and subjective discomfort regarding symptoms.

[0029] In one embodiment of the present invention, the brain image may include one or more brain images selected from the group consisting of the lingual-parahippocampal-fusiform gyrus, superior frontal gyrus, mid. temporal gyrus, supplementary motor area, superior mid. frontal gyrus, posterior cingulate cortex-precuneus, paracentral lobule, and primary somatosensory cortex / primary motor cortex.

[0030] The term “brain region” in this specification refers to a part of the brain having a certain volume that is partitioned according to specific criteria. In the present invention, brain regions may be divided according to arbitrary criteria such as function, association, location, and cellular structure and composition, considering the diagnostic purpose and characteristics of the subject, or may be divided according to existing partitioning methods. For example, brain regions may be classified according to location and function into hippocampus, parahippocampal gyrus, frontal lobe, insula, primary somatosensory cortex, primary motor cortex, inferior temporal gyrus, superior temporal gyrus, inferior frontal gyrus, anterior cingulate cortex, perirhinal cortex, entorhinal cortex, etc. Alternatively, the brain region may be partitioned according to Brodmann areas and divided into Brodmann area 1 to Brodmann area 52.

[0031] In one embodiment of the present invention, the processing step may generate data on changes in the activity of one or more brain regions over time based on the medical data.

[0032] In one embodiment of the present invention, the processing step may include a preprocessing step of preprocessing medical data by performing one or more operations among noise removal, motion correction, spatial smoothing, and standard templating.

[0033] Another aspect of the present invention is a computer program stored in a storage medium, wherein the computer program, when executed on one or more processors, performs the following operations for predicting the responsiveness of acupuncture treatment, the operations comprising: a processing operation for generating input data based on medical data including brain images of a subject who has received acupuncture treatment; a generating operation for generating one or more prediction information by inputting the input data into a pre-trained responsiveness prediction model; and a classification operation for classifying high-response groups and low-response groups based on the prediction information.

[0035] Another aspect of the present invention is a computing device for predicting the responsiveness of acupuncture treatment, comprising: a processor including one or more cores; and memory; wherein the processor generates input data based on medical data including brain images of a subject who has received acupuncture treatment, inputs the input data into a pre-trained responsiveness prediction model to generate one or more prediction information, and classifies a high-response group and a low-response group based on the prediction information. Effects of the invention

[0036] The present invention relates to an apparatus and method for predicting the effect of acupuncture treatment using brain imaging signals. The method and apparatus according to the present invention can accurately predict the effect of acupuncture treatment appearing from a subject solely through brain imaging, thereby enabling the presentation of appropriate treatment to the patient and serving as a basis for the effect of non-pharmacological treatments such as acupuncture. Brief explanation of the drawing

[0037] FIG. 1 is a block diagram illustrating a computing device that performs an operation to predict the responsiveness of acupuncture treatment according to one embodiment of the present invention. FIG. 2 is a block diagram illustrating the process of predicting the responsiveness of acupuncture treatment according to one embodiment of the present invention. FIG. 3 is a schematic diagram showing a deep neural network according to one embodiment of the present invention. FIG. 4 is a diagram showing an experimental process for predicting the responsiveness of acupuncture treatment from brain images according to one embodiment of the present invention. Figure 5 is a diagram showing a brain region with excellent prediction accuracy of acupuncture treatment responsiveness according to one embodiment of the present invention. Specific details for implementing the invention

[0038] Embodiments of the present invention are described below with reference to the attached drawings so that those skilled in the art can easily implement them. However, the present invention may be embodied in various different forms and is not limited to the embodiments described herein. Furthermore, in order to clearly explain the present invention in the drawings, parts unrelated to the explanation have been omitted, and similar parts throughout the specification are denoted by similar reference numerals.

[0039] Throughout the specification, when a part is described as "comprising" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components. The terms "and / or" and "and / or" include all combinations of the items listed in relation thereto and any one thereof.

[0040] Additionally, the terms “…part” and “…module” as described in the specification refer to a unit that processes at least one function or operation, and this may be implemented in hardware, software, or a combination of hardware and software.

[0041] It should be recognized that the various exemplary logical blocks, configurations, modules, circuits, means, logics, and algorithmic steps described in connection with the embodiments herein may be implemented in electronic hardware, computer software, or a combination of both. To clearly exemplify the interchangeability of hardware and software, various exemplary components, blocks, configurations, means, logics, modules, circuits, and steps have been generally described above in terms of their functionality. Whether such functionality is implemented in hardware or software depends on the specific application and design constraints imposed on the overall system. Persons of the art may implement the described functionality in various ways for each specific application. However, such decisions regarding implementation should not be construed as moving out of the scope of this disclosure.

[0043] FIG. 1 is a block diagram illustrating a computing device that performs an operation to predict the responsiveness of acupuncture treatment according to one embodiment of the present invention, and FIG. 2 is a block diagram illustrating a process for predicting the responsiveness of acupuncture treatment according to one embodiment of the present invention.

[0044] Referring to FIGS. 1 and 2, a computing device for performing an operation to predict the responsiveness of acupuncture treatment according to one embodiment may include a processor (100) and a memory (200).

[0045] A processor (100) can generate input data based on medical data including brain images of a subject who has received acupuncture treatment, input the input data into a pre-trained reactivity prediction model to generate one or more prediction information, and perform an operation to classify a high-response group and a low-response group based on the prediction information.

[0046] The processor (100) can perform preprocessing operations by performing one or more of noise removal, motion correction, spatial smoothing and standard templating when generating input data based on medical data.

[0047] Medical data may include brain images. For example, medical data may include images of one or more brain regions. In this case, medical data may include multidimensional medical image data composed of discrete image elements. For example, medical data may be medical images of a subject collected by computed tomography (CT), magnetic resonance imaging (MRI), functional magnetic resonance imaging (fMRI), fundus images, ultrasound, or any other medical imaging system known in the art of the present invention, and may be, for example, functional magnetic resonance imaging.

[0048] The processor (100) can perform an operation to homogenize the medical data when generating input data based on medical data including brain images. For example, when generating input data based on medical data including brain magnetic resonance images, the processor can generate input data by performing an operation to homogenize the brain magnetic resonance images by adjusting the intensity of the brain magnetic resonance images. When the brain magnetic resonance images are homogenized, it may be advantageous for training a reactive prediction model and for prediction through the model.

[0049] The input data may include data on changes in activity of one or more brain regions over time, generated based on medical data. For example, the medical data may include data in which voxels of target regions are defined from functional brain magnetic resonance imaging, and changes in brain activity in each voxel region over time are shown.

[0050] Brain imaging may include one or more brain regions. Brain regions may be appropriately selected based on the purpose of use and predictive accuracy, for example, the brain regions include the primary somatosensory cortex / primary motor cortex, primary motor cortex, M1-superior parietal cortex, paracentral lobule, supplementary motor area, superior parietal gyrus, and anterior insula.Insula), middle insula, thalamus, inferior operculum - superior temporal gyrus - anterial insula, superior-middle frontal gyrus, superior frontal gyrus, middle frontal gyrus, inferior frontal gyrus, orbitofrontal cortex, orbitofrontal cortex - anterior cingulate cortex, anterior cingulate cortex, posterior cingulate cortex, precuneus, posterior cingulate cortex - precuneus, temporo-parietal junction, It may be one or more brain regions selected from the group consisting of the temporal pole, middle temporal gyrus, inferior temporal gyrus, inferior parietal gyrus, and lingual-parahippocampal-fusiform gyrus.

[0051] Meanwhile, the brain image may be a resting brain image. The resting state refers to a brain image captured when the subject is not performing a specific activity, and as shown in the experimental results to be described later, the inventors of the present invention confirmed that predicting responsiveness from subjects who have received acupuncture treatment using brain images in the resting state offers excellent accuracy.

[0052] When a computing device performs an operation of classifying high response groups and low response groups based on prediction information, it can classify targets according to SCORAD values. At this time, the processor can classify the high response group as a group in which SCORAD values ​​decrease significantly and the low response group as a group in which SCORAD values ​​decrease slightly. For example, the computing device can perform an operation of classifying targets in which SCORAD values ​​decrease by 8.7 or more as the high response group and targets in which SCORAD values ​​decrease by less than 8.7 as the low response group.

[0053] The pre-trained reactivity prediction model may be a deep neural network (DNN). The deep neural network may include an input layer and an output layer, or it may include multiple separate hidden layers in addition to the input layer and the output layer. The deep neural network may include a convolutional neural network (CNN), a recurrent neural network (RNN), a Long Short-Term Memory algorithm, a restricted Boltzmann machine (RBM), a deep belief network (DBN), a Q network, a U network, a Siamese network, etc. For example, the deep neural network may be a recurrent neural network, specifically a Long Short-Term Memory algorithm.

[0054] Long Short-Term Memory (LSTM) algorithms are a type of artificial neural network and a variation of the Recurrent Neural Network (RNN). LSTMs were designed to overcome the limitations of RNNs and demonstrate superior performance compared to RNNs, particularly in learning patterns over long periods. LSTMs can handle long-term dependencies by introducing an additional information flow called a cell state, in addition to the hidden state. LSTMs handle long-term dependencies through a 'gate' structure, and the gates determine what information to add or remove from the cell state. The gates consist of three types: an input gate that determines whether to add new information to the cell state, a forget gate that determines which information to remove from the current cell state, and an output gate that determines what output to generate based on the updated cell state. In the present invention, the LSTM model can be used for brain imaging signal analysis. As mentioned above, the input data may include data on changes in the activity of one or more brain regions over time, and since this data exhibits a pattern that changes over time, a recurrent neural network such as LSTM is suitable for modeling it.

[0055] In a computing device according to one embodiment, a reactivity prediction model may include one or more LSTM layers, and for example, the reactivity prediction model may be a model in which two LSTM layers are sequentially connected and a fully connected layer is connected.

[0056] The subject may be a patient with atopic dermatitis. Accordingly, a computing device that performs an operation to predict the responsiveness of acupuncture treatment according to one embodiment may perform an operation to predict the effect of acupuncture treatment on an atopic patient.

[0057] The processor (100) may be composed of one or more cores and may include a processor for data analysis and deep learning, such as a central processing unit (CPU), a graphics processing unit (GPU), or a tensor processing unit (TPU) of a computing device. The processor may read a computer program stored in memory and perform data processing for machine learning according to one embodiment. According to one embodiment, the processor may perform operations for learning a neural network. The processor may perform calculations for learning a neural network, such as processing input data for learning in deep learning (DL), extracting features from input data, calculating errors, and updating the weights of the neural network using backpropagation. At least one of the CPU, GPU, and TPU of the processor (110) may process the learning of a network function.

[0058] The memory (200) may include at least one type of storage medium among flash memory type, hard disk type, multimedia card micro type, card type memory (e.g., SD or XD memory, etc.), RAM (Random Access Memory; RAM), SRAM (Static Random Access Memory), ROM (Read-Only Memory; ROM), EEPROM (Electrically Erasable Programmable Read-Only Memory), PROM (Programmable Read-Only Memory), magnetic memory, magnetic disk, and optical disk.

[0059] A computing device for performing an operation to predict the responsiveness of acupuncture treatment according to one embodiment may further include a network unit for transmitting and receiving brain images and other data as needed, and a database for storing and managing data used in other computing processes. Accordingly, a computing device according to one embodiment may be capable of wired or wireless communication with brain imaging devices, such as MRI and PET scanners. Generated input data and prediction information generated from the input data may be stored in a database, and data stored in the database may be managed, such as by deletion or modification, through external input authorized by a user.

[0060] FIG. 2 is a schematic diagram showing a deep neural network according to one embodiment of the present invention.

[0061] A neural network can represent a model of a machine learning structure designed to extract feature data from input data and provide inference operations using the feature data. In this case, the feature data may represent data regarding features abstracted from the input data. Although FIG. 2 is illustrated as having three hidden layers for convenience of explanation, the hidden layers may include various numbers of layers. A neural network may include one or more layers, and each layer may include one or more nodes.

[0062] A node (or unit) is an element that constitutes each layer, and each layer can be composed of a node or a set of nodes. Within a neural network, nodes in layers other than the output layer can be connected to nodes in the next layer through links to transmit output signals. In this case, nodes in each layer can be interconnected via links, and nodes in the connected layers may be in an input node or output node relationship depending on whether signals are being transmitted or received. Among the nodes connected via links, the value of the output node's data can be determined based on the data input into the input node. Each node included in the hidden layer may receive the output of an activation function regarding the weighted inputs of the nodes included in the previous layer. Weighted inputs are inputs of the nodes included in the previous layer to which weights have been applied. In this case, weights can be variable and may vary depending on the function and algorithm of the neural network. Weights can be referred to as parameters of a neural network, and activation functions may include sigmoid, hyperbolic tangent (tanh), and rectified linear unit (ReLU).

[0063] Initial input nodes may refer to one or more nodes within a neural network to which data is directly input without passing through links in their relationships with other nodes. Alternatively, in terms of link-based relationships between nodes within the neural network, they may refer to nodes that do not have other input nodes connected by links. Similarly, final output nodes may refer to one or more nodes within a neural network that do not have output nodes in their relationships with other nodes. Furthermore, hidden nodes may refer to nodes constituting the neural network that are neither initial input nodes nor final output nodes.

[0064] A deep neural network (DNN) can refer to a neural network that includes multiple hidden layers in addition to input and output layers. Deep neural networks allow for the identification of latent structures in data. In other words, they can identify the latent structures of photos, text, videos, audio, and music (e.g., what objects are present in a photo, what the content and emotions of the text are, what the content and emotions of the audio are, etc.). Deep neural networks may include convolutional neural networks (CNN), recurrent neural networks (RNN), long-short-term memory algorithms, autoencoders, Generative Adversarial Networks (GAN), restricted Boltzmann machines (RBM), deep belief networks (DBN), Q networks, U networks, Siamese networks, and Generative Adversarial Networks (GAN).

[0065] A neural network can be trained using at least one of the following methods: supervised learning, which uses training data labeled with correct answers; unsupervised learning, which uses training data without labeled answers; semi-supervised learning; or reinforcement learning. In this process, an error can be calculated by comparing the output of the neural network with the labels or training data. The calculated error is backpropagated within the neural network (i.e., from the output layer to the input layer), and the connection weights of each node in each layer of the neural network can be updated based on this backpropagation. The amount of change in the connection weights of each updated node can be determined by the learning rate.

[0066] Overfitting is a phenomenon in neural networks where errors increase despite an increase in the number of training iterations due to excessive learning on the training data. Overfitting can act as a cause of increased errors in machine learning algorithms, and various optimization methods can be used to prevent it. To prevent overfitting, methods such as increasing the training data, regularization, dropout (which disables some nodes in the network during training), and the use of batch normalization layers can be applied.

[0068] The present invention will be explained in more detail below through the following examples. However, these examples are merely illustrative of the invention, and the scope of the invention is not limited by these examples.

[0070] Example 1: Experimental Method

[0071] The experiment was approved by the Institutional Review Board of Kyung Hee University Korean Medicine Hospital (KOMCIRB 2020-06-003) and conducted after enrollment at the Korea Clinical Trial-Similar Agency (KCT0005422). Forty-one patients with mild to moderate atopic dermatitis (22 males, mean ± standard deviation: 24.34 ± 5.29 years) and 40 control subjects (20 males, mean ± standard deviation: 26.4 ± 5.32 years) participated. All participants signed informed consent, and the experiment was conducted ethically in accordance with the principles of the Declaration of Helsinki. Resting-state functional magnetic resonance imaging (rs-fMRI) data were obtained prior to acupuncture treatment (manual acupuncture, twice a week for four weeks, 15 minutes per session). Patients with atopic dermatitis were randomly assigned to receive either real acupuncture (n = 27) or sham acupuncture treatment (n = 14). After acupuncture treatment, patients were classified into a high-responder group (11) and a low-responder group (16) based on the difference in minimum therapeutic importance. Specifically, patients were classified as a high-responder group if their SCORing Atopic Dermatitis total score decreased by 8.7 points or more on a 0-10 point scale, and as a low-responder group if it decreased by less than 8.7. Experiments were performed on the 38 predefined regions of interest (ROIs) in Table 1.

[0073] number MNI coordinates (X, Y, Z) area Accuracy 1 -7.12, -52.22, 60.71 Precuneus (left) 0.67 2 -39.63, -19.04, 54.21 Primary somatosensory cortex / primary motor cortex (left) 0.78 3 44.34, -7.55, 56.98 Primary motor cortex (right) 0.69 4 -29.1, -43, 60.66 Primary motor cortex - superior parietal lobe (M1-sup. parietal cortex, left) 0.69 5 2.4, -27.94, 60.15 Paracentral lobule (right) 0.78 6 -2.88, 2.38, 53.21 Supplementary motor area (left) 0.76 7 6.52, 7.69, 50.58 Supplementary motor area (right) 0.87 8 25.34, -58.18, 60.34 Superior parietal gyrus (right) 0.67 9 -16.5, -58.57, 64.46 Superior parietal gyrus (left) 0.67 10 -35.44, 20.03, 0.07 Anterial Insula (Left) 0.67 11 35.91, 21.91, 2.62 Anterial Insula (right) 0.67 12 -44.76, 0.1, 8.83 middle insula (left) 0.69 13 -10.28, -18.48, 7.04 Awards (Thalamus, left) 0.69 14 11.75, -17.18, 7.54 Awards (Thalamus, right) 0.66 15 -51.26, 8.26, -2.06 Inferior operculum - superior temporal gyrus - anterior insula (left) 0.71 16 23.33, 33.07, 47.68 Superior-midline frontal gyrus (right) 0.84 17 -16.4, 28.52, 53.05 Superior frontal gyrus (left) 0.9 18 -35.36, 19.86, 50.8 Mid. frontal gyrus (left) 0.81 19 6.11, 63.98, 21.96 Superior medial frontal gyrus (right) 0.66 20 -2.06, 37.85, 36.34 Superior medial frontal gyrus (left) 0.82 21 26.07, 49.56, 26.58 Mid. frontal gyrus (right) 0.75 22 47.6, 22.16, 9.74 Inferior frontal gyrus (right) 0.71 23 -42.09, -54.98, 44.74 Inferior frontal gyrus (left) 0.67 24 -17.65, 63.19, -9.17 Orbito-frontal cortex (left) 0.73 25 -46.17, 31.26, -13.03 Orbito-frontal cortex (right) 0.66 26 -3.06, 44.41, -9.46 Orbito-frontal cortex- anterior cingulate cortex (left) 0.74 27 -2.5, 41.7, 16.05 anterior cingulate cortex (left) 0.67 28 7.51, 42.49, -5.35 anterior cingulate cortex (right) 0.67 29 12.25, 35.63, 20.3 anterior cingulate cortex (right) 0.67 30 7.94, -48.37, 30.57 posterior cingulate cortex (right) 0.78 31 15.12, -63.09, 25.98 Precuneus (right) 0.73 32 -2.2, -36.68, 43.85 posterior cingulate cortex- precuneus (left) 0.79 33 -45.79, -60.69, 20.85 Temporo-parietal junction (left) 0.74 34 -43.58, 11.99, -34.15 Temporal pole (left) 0.78 35 64.64, -11.8, -19.3 mid-temporal gyrus (right) 0.89 36 58.31, -52.79, -13.61 inferior temporal gyrus (right) 0.72 37 43.93, -52.95, 46.95 Inferior parietal gyrus (right) 0.67 38 -33.93, -38.06, -15.6 Lingual-parahippocampal-fusiform gyrus (left) 0.9

[0075] Resting state data was preprocessed with the FMRIB software library and subjected to denoising, motion correction, spatial smoothing, and standard templates. Long short-term memory (LSTM) was used to extract relevant time-series dynamic information using the scikit-learn, Keras, and TensorFlow (12) libraries to capture dynamic brain activity over time (see Fig. 4). Since LSTM models retain contextual information of the input sequence, the use of LSTM is suitable for capturing dynamic brain activity over time. The preprocessed rs-fMRI data was used as input to a neural network sequential model with two LSTMs and one dense layer. The significance of classification accuracy obtained from bootstrapping (n = 100) and 4-fold cross-validation using LSTM classifiers was tested at the chance level (0.5) using t-tests. The significance threshold was set at p < 0.05.

[0077] Example 2: Results

[0078] To distinguish between high-responders and low-responders for acupuncture treatment, the lingual-parahippocampal-fusiform gyrus (0.90), superior frontal gyrus (0.90, p < 0.01), right mid-temporal gyrus (0.89, p < 0.001), supplementary motor area (0.87), superior mid-frontal gyrus (0.84, p < 0.01), left superior mid-frontal gyrus (0.82 / 0.81), posterior cingulate cortex-precuneus (0.79, p < 0.05), right paracentral lobule (0.78), and left It was confirmed that the primary somatosensory cortex / primary motor cortex (0.78, p < 0.01) showed high predictive accuracy.

Claims

Claim 1 A method for predicting the responsiveness of acupuncture treatment for a patient with atopic dermatitis from medical imaging data by a computing device comprises: a processing step of generating input data based on medical data including functional magnetic resonance imaging (fMRI) of the brain in a resting state of the subject; a generation step of generating one or more prediction information by inputting the input data into a pre-trained responsiveness prediction model; and a classification step of classifying a high-responder group and a low-responder group based on the prediction information; wherein the processing step defines a region of interest including a lingual-parahippocampal-fusiform gyrus region in the medical data, generates a time-series signal representing a change in activity over time for each of a plurality of voxels included in the region of interest, and generates input data composed of the time-series signals of the plurality of voxels; wherein the responsiveness prediction model includes two sequentially connected Long Short-Term Memory (LSTM) layers and a fully connected layer connected to the LSTM layers, and the input data in units of the region of interest A method for predicting the responsiveness of acupuncture treatment, which receives input and generates the above-mentioned prediction information. Claim 2 A method for predicting the responsiveness of acupuncture treatment according to claim 1, wherein the responsiveness prediction model is supervised learning based on one or more training images and guide labels corresponding to the training images and including prediction results, and the supervised learning is performed based on the result of comparing training information generated for the training images using the responsiveness prediction model with the guide labels. Claim 3 A method for predicting the responsiveness of acupuncture treatment, wherein, in claim 1, the classification step classifies subjects whose SCORAD value decreases by 8.7 or more into a high-response group and subjects whose SCORAD value decreases by less than 8.7 into a low-response group. Claim 4 A method for predicting the responsiveness to acupuncture treatment according to claim 1, wherein the medical data comprises brain images of one or more regions selected from a group consisting of the superior frontal gyrus, mid-temporal gyrus, supplementary motor area, superior mid-frontal gyrus, posterior cingulate cortex-precuneus, paracentral lobule, and primary somatosensory cortex / primary motor cortex. Claim 5 A method for predicting atopic dermatitis according to claim 1, wherein the processing step includes a preprocessing step of preprocessing medical data by performing one or more operations among noise removal, motion correction, spatial smoothing, and standard templating. Claim 6 A computer program stored on a storage medium, wherein the computer program, when executed on one or more processors, performs the following operations to predict the responsiveness of acupuncture treatment for a patient with atopic dermatitis, said operations comprising: a processing operation to generate input data based on medical data including functional magnetic resonance imaging (fMRI) of the subject's brain in a resting state; a generating operation to generate one or more prediction information by inputting said input data into a pre-trained responsiveness prediction model; and a classification operation to classify high-responders and low-responders based on said prediction information; said processing operation to define a region of interest including the lingual-parahippocampal-fusiform gyrus region in said medical data, generate a time-series signal representing a change in activity over time for each of a plurality of voxels included in said region of interest, and generate input data composed of said time-series signals of the plurality of voxels; said responsiveness prediction model comprises two sequentially connected Long Short-Term Memory (LSTM) layers and connected to said LSTM layers It includes a fully connected layer, and the input data in units of the region of interest A computer program stored on a storage medium that receives input and generates the above prediction information. Claim 7 A computing device for predicting the responsiveness of acupuncture treatment for a patient with atopic dermatitis, comprising one or more cores; and memory; wherein the processor generates input data based on medical data including functional magnetic resonance imaging (fMRI) of the brain of a subject in a resting state, inputs the input data into a pre-trained reactivity prediction model to generate one or more prediction information, and classifies high-response and low-response groups based on the prediction information, wherein the generation of input data defines a region of interest including the lingual-parahippocampal-fusiform gyrus region in the medical data, generates a time-series signal representing a change in activity over time for each of a plurality of voxels included in the region of interest, and generates input data composed of the time-series signals of the plurality of voxels, and wherein the reactivity prediction model includes two sequentially connected Long Short-Term Memory (LSTM) layers and a fully connected layer connected to the LSTM layers, and the input data in units of the region of interest A computing device that receives input and generates the above prediction information.

Images