Developmental disorder diagnosis support apparatus and developmental disorder diagnosis support method

Abstract

To provide a developmental disorder diagnosis support device capable of supporting diagnosis of a child having a developmental disorder.SOLUTION: The processor 22 analyzes the parent-child activity video data 41 to acquire time-series data indicating the characteristics of the child's movements, and calculates a movement index that evaluates the child's movements related to the presence or absence of developmental disorders based on the time-series data. The processor 22 then estimates the likelihood that the child has a developmental disorder on the basis of the motion index.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to a developmental disorder diagnostic support device and a developmental disorder diagnostic support method. [Background technology]

[0002] In recent years, the number of children with developmental disorders has been increasing, and early diagnosis, detection, and early intervention have become social issues. However, due to a shortage of specialists in developmental disorders, it often takes a long time, sometimes several months, to diagnose a developmental disorder.

[0003] In addition, experts diagnose whether a child has a developmental disorder by observing the child's developmental state based on their movements, facial expressions, and communication skills. Generally, the diagnostic criteria for developmental disorders are based on qualitative rules of thumb, such as "children with developmental disorders have little change in facial expressions" or "children with developmental disorders have sudden changes in movements," so the accuracy of the diagnosis depends on the expert.

[0004] In response to this, Patent Documents 1 to 4 disclose techniques for quantitatively measuring human movements, facial expressions, and physical communication through sports and the like. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Publication No. 2018-081406 [Patent Document 2] Japanese Patent Application Publication No. 2023-064238 [Patent Document 3] Japanese Patent Application Publication No. 2024-054747 [Patent Document 4] Japanese Patent Application Laid-Open No. 2023-108290 Summary of the Invention [Problem to be solved by the invention]

[0006] However, Patent Documents 1 to 4 do not describe the problem of diagnosing developmental disorders in children, and in fact, the techniques described in Patent Documents 1 to 4 cannot be directly applied to diagnosing developmental disorders in children.

[0007] An object of the present disclosure is to provide a developmental disorder diagnostic support device and a developmental disorder diagnostic support method that can support the diagnosis of children with developmental disorders. [Means for solving the problem]

[0008] A developmental disorder diagnostic support device according to one aspect of the present disclosure is a developmental disorder diagnostic support device that supports the diagnosis of a developmental disorder in a subject, and has a processor, which analyzes video data showing the subject to obtain time series data indicating characteristics of the subject's movements, calculates a movement index that evaluates the movement in relation to the presence or absence of the developmental disorder based on the time series data, and estimates the possibility that the subject has a developmental disorder based on the movement index. [Effects of the Invention]

[0009] According to the present invention, it is possible to assist in the diagnosis of developmental disorders. [Brief explanation of the drawings]

[0010] [Figure 1] 1 is a diagram illustrating an example of a developmental disorder diagnostic support system according to an embodiment of the present disclosure. [Figure 2] FIG. 10 is a diagram illustrating another example of a developmental disorder diagnostic support system according to an embodiment of the present disclosure. [Figure 3] 1 is a flowchart illustrating an example of a first calculation process and a second calculation process. [Figure 4] 10 is a flowchart illustrating an example of a third calculation process. [Figure 5] 10 is a flowchart illustrating an example of a fourth calculation process. [Figure 6] 10 is a flowchart illustrating an example of a synchronization determination process. [Figure 7] 10 is a flowchart illustrating an example of an estimation process. [Figure 8] FIG. 10 is a diagram illustrating an example of a display screen. DETAILED DESCRIPTION OF THE INVENTION

[0011] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings. In the following description, a process will be described with a "program" as the operating subject, but since the program is executed by a processor to perform a predetermined process while appropriately using at least one of a storage unit and an interface unit, the subject of the process may be the processor (or a computer or computer system having a processor).

[0012] 1 and 2 are diagrams illustrating a developmental disorder diagnostic support system according to an embodiment of the present disclosure. The developmental disorder diagnostic support system illustrated in Fig. 1 and Fig. 2 includes a smartphone 1 and a PC (Personal Computer) 2. Fig. 1 illustrates a configuration in which the smartphone 1 and the PC 2 are directly connected to each other so as to be able to communicate with each other, while Fig. 2 illustrates a configuration in which the smartphone 1 and the PC 2 are connected to each other so as to be able to communicate with each other via a network 3.

[0013] The smartphone 1 functions as a user terminal used by a user of the developmental disorder diagnostic support system. The smartphone 1 acquires video data showing a child who is the subject of a developmental disorder diagnosis, and transmits the video data to the PC 2. The video data may show only the child, but in this embodiment, it is parent-child activity video data 41 that shows not only the child but also the parent who is communicating with the child. Note that the person communicating with the child is not limited to the parent, but may also be another child, a teacher, etc. Furthermore, the user terminal is not limited to the smartphone 1.

[0014] The PC 2 is a computer that functions as a developmental disorder diagnostic support device that estimates the possibility that a child has a developmental disorder based on parent-child activity video data 41 from the smartphone 1. Note that the developmental disorder diagnostic support device is not limited to the PC 2, and may be a tablet terminal, a server, or the like.

[0015] The PC 2 includes a storage unit 21, a processor 22, a memory 23, an input / output unit 24, and an external communication unit 25.

[0016] The storage unit 21 is a device that records data in a writable and readable manner, and stores a program that defines the operation of the processor 22, and various data that are used and generated by the program.

[0017] In the illustrated example, the programs stored in the storage unit 21 include a skeleton recognition program 31, a facial expression recognition program 32, a synchronization determination program 33, a body part synchronization calculation program 34, a synchronization probability calculation program 35, a synchronization frequency calculation program 36, an estimation program 37, and a display program 38. The data stored in the storage unit 21 include parent-child activity video data 41, skeleton feature point data 42, facial expression feature point data 43, feature point synchronization data 44, synchronization network time series data 45, body part synchronization data 46, center of gravity synchronization data 47, and center of gravity synchronization probability time series data 48.

[0018] The processor 22 is configured with a processor, reads a program recorded in the storage unit 21 into the memory 23, and executes processing according to the program using the memory 23. For example, when the processor 22 receives an analysis request from the smartphone 1, it acquires time-series data indicating characteristics of a child's movements based on the parent-child activity video data 41, and calculates, based on the time-series data, action indices for evaluating the child's movements related to the presence or absence of a developmental disorder. Then, based on the action indices, the processor 22 estimates the possibility that the child appearing in the parent-child activity video data 41 has a developmental disorder, and displays the estimation result on the smartphone 1.

[0019] The child's movements include actions, which are the child's physical movements, and facial expressions, which are the child's facial movements. There may be multiple types of movement indices. In this embodiment, the movement indices include a first index representing the complexity of the synchrony (linkage) of each part of the child's body, a second index representing the complexity of the child's facial expressions, a third index representing the synchrony of the facial expressions of the child and parent, and a fourth index representing the synchrony of the body movements of the child and parent.

[0020] The input / output unit 24 receives various data from and displays various data to the user of the PC 2. The external communication unit 25 is communicably connected to an external device such as the smartphone 1, and transmits and receives data to and from the external device.

[0021] FIG. 3 is a flowchart illustrating an example of a first calculation process and a second calculation process for calculating a first index and a second index for evaluating a child's movements related to the presence or absence of a developmental disorder.

[0022] In the first calculation process, the skeleton recognition program 31 first analyzes the parent-child activity video 41 to recognize skeletal feature points, which are characteristic points of the skeletons of the child and parent appearing in the parent-child activity video 41, and generates skeletal feature point data 42 indicating the positions of the skeletal feature points for each frame of the parent-child activity video 41, and stores the data in the storage unit 21 (step S101). The skeletal feature point data 42 is an example of skeletal feature point time-series data indicating changes in the positions of skeletal feature points as characteristics of the child's movement. Skeletal feature points are parts such as joints, and multiple skeletal feature points exist for both the child and the parent. Note that the parent's skeletal feature points are not necessary to calculate the first index, but are required to calculate the fourth index, so are calculated here as well.

[0023] Based on the skeleton feature point data 42, the skeleton recognition program 31 calculates the movement amount of each of the child's skeleton feature points between successive frames of the parent-child activity video data 41 as the movement amount between frames (step S102).

[0024] Based on the inter-frame movement of each child's skeletal feature point, the synchronization determination program 33 calculates the movement of the child's skeletal feature points for each predetermined synchronization determination interval (here, T seconds) as time-series data of the movement amount. Based on the time-series data of the movement amount, the synchronization determination program 33 generates feature point synchronization data 44, which is synchronized time-series data indicating the children's skeletal feature points that are synchronized (linked) with each other, for each synchronization determination interval, and stores the data in the storage unit 21 (step S103). Note that in calculating the time-series data of the movement amount, the synchronization determination program 33 calculates the movement amount during the synchronization determination interval of T seconds, shifting it by S seconds, which is shorter than T seconds. Furthermore, synchronization has directionality; for example, it is determined separately that a first feature point is synchronized with a second feature point and that a second feature point is synchronized with the first feature point. Synchronization determination will be described in more detail with reference to FIG. 6.

[0025] Based on the feature point synchronization data 44, the body part synchronization calculation program 34 generates time series data indicating a network for each synchronization determination interval, in which each skeletal feature point of the child is defined as a node and lines connecting skeletal feature points that are synchronized with each other are defined as edges (links), and stores this data in the storage unit 21 as skeletal feature point synchronization network time series data 45 (step S104). Note that edges may be represented by arrows indicating the synchronization direction.

[0026] The part synchronization calculation program 34 calculates a body synchronism evaluation value that evaluates the synchronism of each part of the child's body based on the synchronization network time series data 45 of skeletal feature points, and stores part synchronization data 46 indicating the synchronism evaluation value in the memory unit 21 (step S105).

[0027] In this embodiment, the synchrony evaluation value is a value based on the probability of occurrence of a phenomenon called a 3-clique, in which all three skeletal feature points are synchronized (all connected by an edge) for each combination of three skeletal feature points of the children. More specifically, the synchrony evaluation value is the ratio of the total occurrence probability k, which is the sum of the occurrence probabilities p of the 3-cliques, to the entropy difference of the occurrence probability p (1-H / H max ) The ratio of the entropy difference is specifically the maximum value H of the entropy H of the occurrence probability p. maxIts maximum value H max and the difference between the entropy H (H max The sum of the occurrence probabilities represents the activity of the connections between the child's skeletal features (activity of the movements), and the ratio of the entropy differences represents the complexity of the network structure of the child's skeletal features (complexity of the movements).

[0028] The body part synchronization calculation program 34 calculates a first index for evaluating the child's movements related to the presence or absence of a developmental disorder based on the body part synchronization data 46 (step S106). Specifically, the first index is calculated by dividing the sum of the occurrence probabilities of the three cliques k by the ratio of the entropy differences of the three cliques (1-H / H max This index represents the change in the complexity of movements in which three skeletal features are linked, and has a negative correlation with the scores of evaluation indices for assessing the presence or absence of developmental disorders (such as the CBCL (Child Behavior Check List) and SRS (Social Responsiveness Scale)). This suggests that the behavior of children who may have developmental disorders includes unpredictable movements.

[0029] In the second calculation process, the facial expression recognition program 32 analyzes the parent-child activity video 41 to recognize facial expression feature points, which are characteristic points of the faces of the child and parent appearing in the parent-child activity video 41, and generates facial expression feature point data 43 indicating the positions of the facial expression feature points for each frame of the parent-child activity video 41, and stores the data in the storage unit 21 (step S111). The facial expression feature points are areas that form facial expressions, such as around the eyes and mouth. Note that the parent's facial expression feature points are not necessary to calculate the second index, but are required to calculate the third index, so are calculated here as well.

[0030] The facial expression recognition program 32 calculates the amount of movement of each facial expression feature point of the child between successive frames of the parent-child activity video data 41 as the amount of movement between frames based on the facial expression feature point data 43 (step S112).

[0031] Thereafter, steps S112 to S115 are performed in which the skeletal features in steps S103 to S106 are replaced with facial feature points, and in step S116, body part synchronization calculation program 34 calculates a second index for evaluating the child's movements related to the presence or absence of a developmental disorder based on body part synchronization data 46.

[0032] The second index, like the first index, is calculated as (the sum of the occurrence probabilities of the three cliques, k) ÷ (the ratio of the entropy differences of the three cliques, (1-H / Hmax)). This index represents the change in the complexity of the movements in which the three facial feature points are linked, and has a positive correlation with the value of the evaluation index that evaluates the presence or absence of developmental disorders. This suggests that children who may have developmental disorders have little change in their facial expressions.

[0033] FIG. 4 is a flowchart illustrating an example of a third calculation process for calculating a third index that evaluates a child's movement related to the presence or absence of a developmental disorder.

[0034] In the third calculation process, first, the facial expression recognition program 32 analyzes the parent-child activity video 41 to recognize facial expression feature points, which are feature points of the faces of the child and the parent appearing in the parent-child activity video 41, and generates facial expression feature point data 43 indicating the positions of the facial expression feature points for each frame of the parent-child activity video 41, and stores the data in the storage unit 21 (step S211). Note that the processing of step S211 is the same as the processing of step S111 in the second calculation process (see FIG. 3), and therefore can be omitted if the processing of step S111 has already been executed in the second calculation process.

[0035] Based on the facial feature point data 43, the facial expression recognition program 32 calculates the position of the center of gravity of the facial feature points of each of the child and the parent as the center of gravity of the face for each frame of the parent-child activity moving image data 41 (step S212).

[0036] The synchronization determination program 33 calculates the movement speed of the center of gravity of the child's and parent's faces for each specified synchronization determination period based on the center of gravity of each of the child's and parent's faces as time series data of the movement speed, and generates face center of gravity synchronization data 47 indicating whether or not there is synchronization between the center of gravity of the child's face and the center of gravity of the parent's face for each synchronization determination interval based on the time series data of the movement speed, and stores this in the memory unit 21 (step S213).

[0037] Based on the face centroid synchronization data 47, the synchronization determination program 33 generates time series data showing a network for each synchronization determination interval, with the centroid positions of the child's and parent's faces as nodes and lines connecting the centroid positions that are synchronized with each other as edges, as face centroid synchronization network time series data 45, and stores this in the memory unit 21 (step S214).

[0038] The synchronization probability calculation program 35 calculates, for each evaluation period, the probability that the parent's facial centroid is synchronized with the child's facial centroid as a first synchronization probability that the parent's facial expression is synchronized with the child's facial expression, and further calculates the probability that the child's facial centroid is synchronized with the parent's facial centroid as a second synchronization probability that the child's facial expression is synchronized with the parent's facial expression, based on the facial centroid synchronization network time series data 45. The synchronization probability calculation program 35 calculates centroid synchronization probability time series data 48 indicating the first synchronization probability and the second synchronization probability, and stores them in the storage unit 21 (step S215).

[0039] The synchronization probability calculation program 35 calculates a third index based on the center of gravity synchronization probability time series data 48 (step S216). Specifically, the third index is a correlation coefficient between the time series data of the first synchronization probability and the time series data of the second synchronization probability. This index represents the phase shift between the time series graph of the "probability that the parent synchronizes with the child" and the time series graph of the "probability that the child synchronizes with the parent," and has a negative correlation with the value of the evaluation index for evaluating the presence or absence of a developmental disorder. This suggests that children who may have a developmental disorder tend to be unable to predict the actions of others in communication and tend to start their own actions only after confirming that the other person's actions have finished.

[0040] FIG. 5 is a flowchart illustrating an example of a fourth calculation process for calculating a fourth index for evaluating a child's movement related to the presence or absence of a developmental disorder.

[0041] In the fourth calculation process, the skeleton recognition program 31 analyzes the parent-child activity moving image data 41, recognizes skeletal feature points that are characteristic points of the skeletons of the child and parent appearing in the parent-child activity moving image data 41, generates skeletal feature point data 42 that indicates the positions of the skeletal feature points for each frame of the parent-child activity moving image data 41, and stores the data in the storage unit 21 (step S301). Note that the process of step S301 is the same as the process of step S101 in the first calculation process (see FIG. 3), and therefore can be omitted if the process of step S101 has already been executed in the first calculation.

[0042] Based on the skeletal feature data 42, the skeleton recognition program 31 calculates the positions of the centers of gravity of the skeletal feature points of the child and the parent as the positions of the center of gravity of the body for each frame of the parent-child activity video data 41 (step S302). The positions of the center of gravity of the body are calculated taking into account the weight ratio of each part of the body.

[0043] The synchronization determination program 33 calculates the movement speed of the center of gravity of the child's and parent's bodies for each specified synchronization determination period based on the center of gravity positions of the child's and parent's bodies as time series data of the movement speed, and generates center of gravity synchronization data 47 indicating whether or not the center of gravity positions of the child's and parent's bodies are synchronized for each synchronization determination interval based on the time series data of the movement speed, and stores this in the memory unit 21 (step S303).

[0044] Based on the body center of gravity synchronization data 47, the synchronization determination program 33 generates time series data showing a network for each synchronization determination interval, with the center of gravity positions of the child's and parent's faces as nodes and lines connecting the center of gravity positions that are synchronized with each other as edges, as skeletal center of gravity synchronization network time series data 45 and stores it in the memory unit 21 (step S304).

[0045] The synchronization frequency calculation program 36 calculates the frequency distribution of the synchronization duration, which is the time during which the center of gravity position of the child's body is synchronized with the center of gravity position of the parent's body, based on the synchronization network time series data 45 of the skeleton center of gravity (step S305).

[0046] The synchronization frequency calculation program 36 calculates a fourth index based on the frequency distribution of synchronization duration (step S306). The fourth index is the coefficient of determination when approximating the frequency distribution of synchronization duration to a power distribution. This index indicates that a child's synchronization with their parent is fragmented and often short, and has a negative correlation with the index used to evaluate the presence or absence of a developmental disorder. This suggests that children who may have a developmental disorder have difficulty maintaining behavior that is in line with the other person's behavior for a long period of time when communicating.

[0047] Fig. 6 is a flowchart illustrating an example of the synchronization determination process for determining whether or not skeletal feature points are synchronized in step S103 in Fig. 3. The following synchronization determination process is performed for each of two skeletal feature values ​​that are the subject of synchronization determination.

[0048] In the synchronization determination process, the synchronization determination program 33 generates a frequency distribution of the amount of movement for each of the two skeletal feature points to be determined, based on time-series data of the amount of movement for each of the two skeletal feature points (step S401).

[0049] The synchronization determination program 33 converts the frequency distribution of the movement amounts of the two skeletal feature points into a frequency histogram (step S402). The number of divisions of the histogram is determined using, for example, Sturges's rule.

[0050] The synchronization determination program 33 normalizes the time series data of the movement amounts of the two skeletal feature points so that the movement amounts can be treated as random variables, and further discretizes the movement amounts according to the number of divisions of the frequency histogram to generate random variable time series data (step S403).

[0051] The synchronization determination program 33 calculates the transfer entropy between the two skeletal feature points based on the random variable time series data of each of the two skeletal feature points (step S404).

[0052] The transfer entropy T(J→I) from one skeletal feature point J to another skeletal feature point I at time n+1 is expressed as a vector i whose elements are the random variables (amount of transfer) of skeletal feature point I from time nm to time n. n (m) (=i n-m ,…,i n ) and vector j whose elements are the random variables (movements) of skeleton feature point J from time nm to time n. n (m) (=j n-m ,…,j n ) is expressed as the following equation 1.

number

[0053] Then, synchronization determination program 33 compares the motion entropy with a predetermined threshold to determine whether the two skeletal feature points are synchronized and, if so, the synchronization direction (step S405). For example, if the absolute value of the difference in motion entropy in each direction is equal to or greater than a threshold, synchronization determination program 33 determines that the two skeletal feature points are synchronized, and sets the synchronization direction as the direction from the skeletal feature point with greater motion entropy to the skeletal feature point with less motion entropy.

[0054] Note that the synchronization determination in step S113 in Fig. 3, step S214 in Fig. 4, and step S303 in Fig. 5 can also be realized by processing similar to the above. For example, when applied to the processing in step S113 in Fig. 3, skeletal feature points can be read as facial expression feature points in the above synchronization determination processing.

[0055] FIG. 7 is a flowchart illustrating an example of an estimation process for estimating the presence or absence of a developmental disorder based on the first to fourth indices.

[0056] In the estimation process, the estimation program 37 inputs the first to fourth indices calculated in the first to fourth calculation processes, respectively, into a developmental diagnosis estimation model, which is a machine learning model for estimating a diagnosis result of a developmental disorder (step S501). The developmental diagnosis estimation model is a machine learning model that receives as input movement indices for evaluating a child's movements related to the presence or absence of a developmental disorder and outputs an estimation result of the presence or absence of a developmental disorder. The presence or absence of a developmental disorder is determined by whether or not the CBCL value and SRS value, which are evaluation indices, exceed a threshold value. The developmental diagnosis estimation model is, for example, a binary prediction model constructed using a neural network model. The threshold value may be set by the user or may be set to a value standardized by clinical practice, etc.

[0057] The estimation program 37 acquires the evaluation result output from the development diagnosis estimation model (step S502).

[0058] The estimation program 37 calculates the accuracy of the evaluation result as the probability that the evaluation index is equal to or greater than a threshold (step S503). The accuracy of the evaluation result is, for example, AUC (Area Under the ROC Curve).

[0059] 8 is a diagram illustrating an example of a display screen displayed on the smartphone 1. As shown in FIG. 8, the display screen includes an analysis request screen 801 before the parent-child activity video data 41 is analyzed, and an analysis result screen 802 that displays the analysis results of the parent-child activity video data 41. The analysis request screen 801 and the analysis result screen 802 are generated by the display program 38 and transmitted to the smartphone 1.

[0060] The analysis request screen 801 has a video display section 811, a shooting start / stop button 812, and an image analysis start button 813. The video display section 811 displays video data acquired by shooting. The shooting start / stop button 812 is a button that is tapped to start and stop shooting. The parent-child activity video data 41 is acquired by taking pictures with a camera (not shown) provided in the smartphone 1 between the time the shooting start / stop button 812 is tapped and the time the button is next tapped. The image analysis start button 813 is a button that is tapped to start analysis of the acquired parent-child activity video data 41. When the image analysis start button 813 is tapped, an analysis request including the parent-child activity video data 41 is sent to the PC 2.

[0061] The analysis result screen 802 includes a video display section 821, a seek bar 822, an index value display section 823, and an analysis result display section 824. The video display section 821 displays related video data related to the parent-child activity video data 41. The related video data is, for example, the synchronous network time series data 45, or an image in which the synchronous network time series data 45 is superimposed on the parent-child activity video data 41. The seek bar 822 is responsible for displaying and adjusting the playback position of the related video data. The index value display section 823 displays the first to fourth index values ​​calculated by the PC2. The analysis result display section 824 displays the probability that each of the CBCL value and SRS value, which are evaluation indexes, exceeds a threshold, as the analysis result.

[0062] The inventors of the present application have discovered that the first to fourth movement indices described above are correlated with evaluation indices conventionally used in clinical settings and are associated with the presence or absence of developmental disorders. Furthermore, the meaning of predictions based on the movement indices is consistent with the qualitative empirical rules of experts, suggesting that the behavior of children who may have developmental disorders includes unpredictable movements, as described above. However, the movement indices are not limited to these examples, and any movement indices may be used as long as they can evaluate children's movements related to the presence or absence of developmental disorders.

[0063] As described above, according to this embodiment, the processor 22 analyzes the parent-child activity video data 41 to obtain time-series data indicating characteristics of the child's movements, and calculates action indices that evaluate the child's movements related to the presence or absence of a developmental disorder based on the time-series data.The processor 22 then estimates the possibility that the child has a developmental disorder based on the action indices.This makes it possible to support the diagnosis of a developmental disorder.

[0064] Furthermore, in this embodiment, the processor 22 generates feature point synchronization data 44 indicating a plurality of skeletal feature points that are synchronized with each other, based on skeletal feature point data 42 indicating changes in the positions of each of a plurality of skeletal feature points of the child, and calculates action indices based on the feature point synchronization data 44. In this case, it is possible to calculate appropriate action indices.

[0065] In this embodiment, the first action indicator is calculated by dividing the sum of occurrence probabilities, which is the sum of occurrence probabilities of three cliques in which the three skeletal feature points are synchronized for each combination of three skeletal feature points, by the ratio of the difference between the maximum value of the entropy of the occurrence probabilities and the entropy to the maximum value of the entropy of the occurrence probabilities. In this case, it is possible to calculate an appropriate action indicator.

[0066] Furthermore, in this embodiment, the processor 22 generates feature point synchronization data 44 indicating a plurality of facial expression feature points that are synchronized with each other, based on facial expression feature point data 43 indicating changes in the positions of each of a plurality of facial expression feature points of the child, and calculates action indices based on the feature point synchronization data 44. In this case, it is possible to calculate appropriate action indices.

[0067] In this embodiment, the second action index is calculated by dividing the sum of the occurrence probabilities, which is the sum of the occurrence probabilities of three cliques in which the three facial feature points are synchronized for each combination of three facial feature points, by the ratio of the difference between the maximum value of the entropy of the occurrence probabilities to the maximum value of the entropy of the occurrence probabilities. In this case, it is possible to calculate an appropriate action index.

[0068] In this embodiment, the third action index is a correlation function between the time series data of the first synchronization probability where the center of gravity of the parent's face is synchronized with the center of gravity of the child's face and the time series data of the second synchronization probability where the center of gravity of the child's face is synchronized with the center of gravity of the parent's face, thereby making it possible to calculate an appropriate action index.

[0069] In this embodiment, the fourth action index is a coefficient of determination when power-law approximation is performed on the frequency distribution of the synchronization duration, which is the time during which the center of gravity of the child's body is synchronized with the center of gravity of the parent's body. In this case, it is possible to calculate an appropriate action index.

[0070] Furthermore, in this embodiment, the processor 22 estimates the probability that the value of the evaluation index for evaluating whether or not a child has a developmental disorder will exceed a threshold value, based on the behavior index.

[0071] The above-described embodiments of the present disclosure are merely illustrative examples of the present disclosure, and are not intended to limit the scope of the present disclosure to these embodiments alone. Those skilled in the art may implement the present disclosure in various other forms without departing from the scope of the present disclosure. [Explanation of symbols]

[0072] 1: Smartphone 2: PC 3: Network 21: Storage unit 22: Processor 23: Memory 24: Input / output unit 25: External communication unit 31: Skeleton recognition program 32: Facial expression recognition program 33: Synchronization determination program 34: Body part synchronization calculation program 35: Synchronization probability calculation program 36: Synchronization frequency calculation program 37: Estimation program 38: Display program 41: Parent-child activity video data 42: Skeleton feature point data 43: Facial expression feature point data 44: Feature point synchronization data 45: Synchronization network time series data 46: Body part synchronization data 47: Center of gravity synchronization data 48: Center of gravity synchronization probability time series data

Claims

1. A developmental disorder diagnostic support device that supports the diagnosis of a developmental disorder for a subject, a processor; The processor: Analyzing video data of the subject to obtain time-series data indicating characteristics of the subject's movements; calculating a movement index that evaluates the movement related to the presence or absence of the developmental disorder based on the time-series data; A developmental disorder diagnostic support device that estimates the possibility that the subject has a developmental disorder based on the behavioral indicators.

2. the time-series data includes skeletal feature point time-series data indicating changes in positions of a plurality of feature points of the subject's skeleton as the movement characteristics; 2. The developmental disorder diagnostic support device according to claim 1, wherein the processor generates synchronized time-series data indicating the plurality of feature points that are synchronized with each other based on the skeletal feature point time-series data, and calculates the action index based on the synchronized time-series data.

3. 3. The developmental disorder diagnostic support device according to claim 2, wherein the processor calculates, as the behavior index, a first index which is a value obtained by dividing a sum of occurrence probabilities, which is the sum of occurrence probabilities of three cliques in which the three feature points are synchronized for each combination of the three feature points, by a ratio of a difference between the maximum value of entropy of the occurrence probabilities and the entropy to the maximum value of the entropy.

4. the time-series data includes facial expression feature point time-series data indicating changes in the positions of a plurality of feature points on the face of the subject as the movement characteristics; 2. The developmental disorder diagnostic support device according to claim 1, wherein the processor generates synchronized time-series data indicating the plurality of feature points that are synchronized with each other based on the skeletal feature point time-series data, and calculates the action index based on the synchronized time-series data.

5. 5. The developmental disorder diagnostic support device according to claim 4, wherein the processor calculates, as the behavior index, a second index which is a value obtained by dividing a sum of occurrence probabilities, which is the sum of occurrence probabilities of three cliques in which the three feature points are synchronized for each combination of the three feature points, by a ratio of a difference between the maximum value of entropy of the occurrence probabilities and the entropy to the maximum value of the entropy.

6. The video data further shows a person communicating with the subject; the time series data includes time series data of movement speeds of centers of gravity of a plurality of feature points on the faces of the target person and the other person, The developmental disorder diagnostic support device of claim 1, wherein the processor calculates, as the behavior index, a third index which is a correlation coefficient between time series data of a first synchronization probability in which the center of gravity of the other person is synchronized with the center of gravity of the subject and time series data of a second synchronization probability in which the center of gravity of the subject is synchronized with the center of gravity of the other person, based on the time series data of the moving speed.

7. The video data further shows a person communicating with the subject; the time series data includes time series data of the movement speed of the center of gravity of each of the target person and the other person; The developmental disorder diagnostic support device of claim 1, wherein the processor calculates, as the behavior index, a fourth index which is a coefficient of determination when approximating the frequency distribution of the synchronization duration, which is the time during which the center of gravity of the subject is synchronized with the center of gravity of the other person, based on the time series data of the moving speed.

8. The developmental disorder diagnostic support device of claim 1 , wherein the processor estimates, as the possibility, the probability that the value of an evaluation index for evaluating whether or not the subject has a developmental disorder exceeds a threshold based on the behavior index.

9. A developmental disorder diagnostic support method using a developmental disorder diagnostic support device that supports the diagnosis of a developmental disorder for a subject, comprising: Analyzing video data of the subject to obtain time-series data indicating characteristics of the subject's movements; calculating a movement index that evaluates the movement related to the presence or absence of the developmental disorder based on the time-series data; A developmental disorder diagnostic support method that estimates the possibility that the subject has a developmental disorder based on the behavioral indicators.

Images