JP-7855930-B2 - Index value estimation device, estimation system, index value estimation method, and program

Inventors

• 井原 和紀
• 福司 謙一郎
• オウ シンイ
• 黄 晨暉
• 梶谷 浩司
• 二瓶 史行
• 野崎 善喬
• 中原 謙太郎

Assignees

• 日本電気株式会社

Dates

Publication Date

20260511

Application Date

20220606

Claims (10)

1. A data acquisition means for acquiring feature data, which includes feature quantities extracted from sensor data related to the user's foot movements and used to estimate an index value indicating the user's knee condition; A storage means for storing an estimation model that outputs an index value corresponding to the input of the aforementioned feature data, An estimation means that estimates the output obtained by inputting the acquired feature data into the estimation model as the index value indicating the user's knee condition, The system includes an output means that outputs information regarding the index value indicating the estimated knee condition of the user , The feature data includes features extracted from walking phase clusters, which consist of at least one temporally consecutive walking phase, in walking waveform data extracted from the sensor data for one walking cycle, normalized so that the stance phase accounts for 60 percent, the swing phase accounts for 40 percent, and the timing of toe-off coincides with 60 percent. The estimation model is an index value estimation device that is a machine learning model trained to output parameters related to the knee joint flexion angle, including the timing of the peak appearing in the swing phase of the two peaks that appear in the time series data of the knee joint flexion angle for one step period, and the timing of toe-off, in response to the input of the feature data.
2. The aforementioned storage means is The estimation model generated by learning using training data, in which the feature quantities used to estimate the index value indicating the knee condition extracted from the sensor data obtained from the walking verification of multiple subjects are used as explanatory variables, and the measured values of the index value indicating the knee condition measured in the walking verification of multiple subjects are used as the dependent variable, is stored. The estimation means is, The index value estimation device according to claim 1, wherein the output obtained by inputting the feature data acquired with respect to the user into the estimation model is estimated as the index value indicating the knee condition of the user.
3. The aforementioned storage means is The estimation model, which estimates parameters related to the knee joint flexion angle as the index value indicating the condition of the knee, is stored. The estimation means is, The index value estimation device according to claim 2, wherein parameters related to the knee joint flexion angle obtained by inputting the feature data acquired with respect to the user into the estimation model are estimated as index values indicating the knee condition of the user.
4. The aforementioned storage means is The estimation model is stored, which estimates parameters related to the knee joint flexion angle that appear in the time-series data of the knee joint flexion angle for one gait cycle, The estimation means is, An index value estimation device according to claim 3, which inputs the feature data acquired in accordance with the user's walking into the estimation model, and estimates the index value indicating the user's knee condition according to the index value of the user output from the estimation model.
5. The estimation means is, An index value estimation device according to claim 3, which inputs the feature data acquired in accordance with the user's walking into the estimation model, and estimates the index value indicating the user's knee condition according to the index value of the user output from the estimation model.
6. The aforementioned storage means is A second estimation model is stored to estimate the cost indicating the smoothness of knee movement as the aforementioned index value indicating the condition of the knee. The estimation means is, The index value estimation device according to claim 2 , wherein the cost indicating the smoothness of knee movement, obtained by inputting the feature data acquired with respect to the user into the second estimation model, is estimated as the index value indicating the knee condition of the user.
7. The aforementioned storage means is For each of the multiple segments included in the stance phase, the second estimation model is stored, which estimates the cost indicating the smoothness of knee movement as an index value indicating the state of the knee. The estimation means is, The index value estimation device according to claim 6, wherein the cost indicating the smoothness of knee movement, obtained by inputting the feature data acquired for the user into the second estimation model for at least one of the plurality of intervals, is estimated as the index value indicating the knee condition of the user.
8. An index value estimation device according to any one of claims 1 to 7, It comprises a measuring device installed on the user's footwear, which is the target of estimating an index value indicating the condition of the knee, The aforementioned measuring device is A sensor that measures spatial acceleration and spatial angular velocity, generates sensor data related to foot movement using the measured spatial acceleration and spatial angular velocity, and outputs the generated sensor data, An estimation system comprising: a feature data generation means that acquires time-series data of sensor data including gait characteristics; extracts walking waveform data for one walking cycle from the time-series data of sensor data; normalizes the extracted walking waveform data; extracts features used for estimating the index value indicating the knee state from the normalized walking waveform data from a walking phase cluster consisting of at least one temporally consecutive walking phase; generates feature data including the extracted features; and outputs the generated feature data to the index value estimation device.
9. Computers Obtain feature data, which includes features extracted from sensor data related to the user's foot movements and used to estimate an index value indicating the user's knee condition. The acquired feature data is input to an estimation model that outputs an index value corresponding to the input of the feature data. The output obtained by inputting it into the estimation model is estimated as the index value indicating the user's knee condition. Output information regarding the index value indicating the estimated knee condition of the user , The feature data includes features extracted from walking phase clusters, which consist of at least one temporally consecutive walking phase, in walking waveform data extracted from the sensor data for one walking cycle, normalized so that the stance phase accounts for 60 percent, the swing phase accounts for 40 percent, and the timing of toe-off coincides with 60 percent. The estimation model is a machine learning model that, in response to the input of the feature data, outputs parameters related to the knee joint flexion angle, including the timing of the peak appearing in the swing phase of the two peaks that appear in the time series data of the knee joint flexion angle for one step period, and the timing of toe-off .
10. A process for obtaining feature data, which includes features extracted from sensor data related to the user's foot movements and used to estimate an index value indicating the user's knee condition, The estimation model outputs an index value corresponding to the input of the feature data, and the process involves inputting the acquired feature data into the estimation model. The process involves estimating the output obtained by inputting it into the estimation model as the index value indicating the user's knee condition, The computer is made to perform a process that outputs information regarding the index value indicating the estimated knee condition of the user , The feature data includes features extracted from walking phase clusters, which consist of at least one temporally consecutive walking phase, in walking waveform data extracted from the sensor data for one walking cycle, normalized so that the stance phase accounts for 60 percent, the swing phase accounts for 40 percent, and the timing of toe-off coincides with 60 percent. The estimation model is a machine learning model program that, in response to the input of the feature data, outputs parameters related to the knee joint flexion angle, including the temporal relationship between the timing of the peak appearing in the swing phase of the two peaks that appear in the time-series data of the knee joint flexion angle for one step period and the timing of toe- off.

Description

This disclosure relates to an index value estimation device, etc., for estimating index values indicating the condition of the knee. With the growing interest in healthcare, services that provide information based on gait patterns are attracting attention. For example, technologies are being developed that analyze gait patterns using sensor data measured by sensors embedded in footwear such as shoes. Time-series sensor data reveals features associated with walking events related to physical condition. By analyzing gait data that includes these features, it is possible to estimate the physical condition of the subject. For example, if the condition of a subject's knees can be estimated, early detection and prevention of diseases such as osteoarthritis of the knee becomes possible. Patent Document 1 discloses a knee condition determination system that determines the user's knee condition by focusing on the movement of the knee area during footstepping. The system in Patent Document 1 comprises multiple sensor devices and a knee condition estimation device. The multiple sensor devices are attached to the waist, the thighs of both legs, and the lower legs of both legs. The multiple sensor devices measure the angular velocity generated by the rotational movement of the thighs and lower legs accompanying the user's footstepping. The multiple sensor devices transmit the rotational angular velocity, reflecting the measured angular velocity, to the knee condition determination device. The knee condition determination device determines the user's knee condition by analyzing the data transmitted from the sensor devices. Specifically, the knee condition determination device uses the yaw component around the axis of gravity output from each sensor device attached to the thighs and lower legs of both legs to determine if there is an abnormality in the user's knee. Patent Document 2 discloses a detection device used for estimating the state of a person during movement. The device in Patent Document 2 includes sensors such as an acceleration sensor and an angular velocity sensor. The device in Patent Document 2 is attached to the knee or around the knee of a subject. The acceleration detected by the device in Patent Document 2 is used to estimate the state of the knee. Patent Document 2 also discloses using the detected acceleration to estimate the degree and prognosis of osteoarthritis of the knee. Non-Patent Documents 1-4 report various cases of verification of knee conditions related to diseases such as osteoarthritis of the knee. Non-Patent Document 1 reports on the development of diagnostic methods for knee diseases such as osteoarthritis of the knee. Non-Patent Document 1 lists height, leg length, range of motion, and lower limb alignment as factors influencing gait patterns. Non-Patent Document 2 reports the results of a gait analysis conducted on multiple subjects for the purpose of quantitatively evaluating lateral thrust observed in patients with osteoarthritis of the knee. Non-Patent Document 3 reports the results of evaluating gait abnormalities caused by osteoarthritis of the knee in multiple subjects. Non-Patent Document 4 reports the results of an investigation into the muscles that affect knee flexion velocity during the bilateral support phase of walking. Non-Patent Document 4 states that if the knee flexion velocity is sufficient at toe-off, appropriate knee flexion can be obtained during the swing phase. Japanese Patent Publication No. 2016-106948Japanese Patent Publication No. 2022-051451 Yuki Ishikawa et al., "A Proposal for the Construction of a Diagnostic Method for Knee Disease Using Individual Modeling - Towards Elucidating the Pathogenesis of Osteoarthritis of the Knee," Proceedings of the 2012 JSME Conference on Robotics and Mechatronics, Hamamatsu, Japan, May 27-29, (2012), pp. 2P1-I02(1)-2P1-I02(2).Takashi Komura et al., "A Study on Gait Analysis of Patients with Medial-Type Osteoarthritis of the Knee," Journal of the Faculty of Medicine, Kobe University, 61(4), (2001), pp. 89-94.Shunsuke Yamashina, "Development of an observational method for evaluating gait abnormalities in patients with osteoarthritis of the knee undergoing conservative treatment and verification of its relationship with reduced physical activity," Doctoral dissertation, Kibi International University, 2019.S. R. Goldberg, et al., Journal of Biomechanics, 37, (2004), pp.1189-1196. This is a block diagram showing an example of the configuration of the estimation system according to the first embodiment.This is a block diagram showing an example of the configuration of a measuring device included in the estimation system according to the first embodiment.This is a conceptual diagram showing an example of the arrangement of a measuring device according to the first embodiment.This is a conceptual diagram illustrating an example of the relationship between the local coordinate system and the world coordinate system set in the meas