TWI840680B - Sarcopenia assessment system and method and model establishing method - Google Patents

Abstract

The present invention relates to a sarcopenia assessment method, comprising: providing a sarcopenia self-inspection for a user and acquiring an inspection result; uploading and feeding the inspection result to a sarcopenia assessment model established on the basis of a machine learning classifier; and executing the sarcopenia assessment model in a remote end to performing a sarcopenia assessment, to assess whether the user belongs to a high risk group of sarcopenia or not.

Description

Sarcopenia assessment system and method and assessment model establishment method

The present invention relates to a sarcopenia assessment system and method, and a method for establishing a sarcopenia assessment model based on artificial intelligence machine learning technology, in particular, a system and method that can provide users with a simple operation of their own user equipment to independently conduct sarcopenia risk assessment.

Sarcopenia refers to a persistent and widespread loss of skeletal muscle weight and function throughout the body, which may lead to disability, decreased quality of life, or even inability to take care of oneself and increased risk of death. Studies have shown that human skeletal muscle decreases with age. After the age of 40, muscle mass decreases at a rate of 8% per decade, and after the age of 70, the rate of loss accelerates to 15% per decade. Generally speaking, the body's muscle mass decreases by about 1% each year from the age of 30, and by the age of 80, muscle mass and strength may have been lost by as much as 50%.

The word sarcopenia first originated in 1989 when Irwin Rosenberg proposed combining the Greek roots for meat (sarx) and lack (penia) into the word "sarcopenia" to describe the phenomenon of skeletal muscle loss due to aging. The definition of sarcopenia generally adopts the definition of sarcopenia by the European Working Group on Sarcopenia (EWGSOP) in 2010, which is: "Progressive loss of muscle mass and decreased muscle strength or physiological activity, and sarcopenia is divided into pre-sarcopenia, sarcopenia and severe sarcopenia".

However, due to different definitions of sarcopenia, there are still different definitions of sarcopenia. For example, Baumgartner et al. defined sarcopenia as muscle mass index (appendicular muscle mass/height 2) lower than two standard deviations of the average value of young people. Later, Newman, Janssen et al. also used a similar method, but used skeletal muscle as correction and defined sarcopenia as percentage or standard deviation. The International Working Group on Sarcopenia (IWGS) proposed that in addition to the above-mentioned reduction in muscle mass and strength, physiological performance should also be reduced. The Society for Sarcopenia, Cachexia and Wasting (SCWD) believes that in addition to the reduction in skeletal muscle, there should also be limitations in activity, such as a reduction in walking speed. The Special Interest Group (SIG) on cachexia-anorexia in chronic wasting diseases believes that sarcopenia is not limited to the elderly population, and symptoms include a reduction in muscle mass and strength.

Sarcopenia is believed to be caused by many factors, including hormonal changes caused by age and aging, neuron degeneration, decreased number of neuromuscular junctions, insufficient blood vitamin D, nutritional deficiencies, reduced activity, reduced physical activity, sedentary lifestyle, and increased oxidative stress in the body, all of which can promote the occurrence of sarcopenia. Gender, history of chronic diseases, high or low body mass index (BMI), previous falls, poor physical function, lack of exercise, long-term medication use (such as steroids), insufficient calorie or protein intake, gastrointestinal diseases, kidney diseases, retirement, low education, bedridden, and hospitalization can accelerate muscle loss and are considered a high-risk group for sarcopenia.

Sarcopenia can lead to adverse prognoses such as the risk of falling and fractures, functional impairment of physical activities, frailty and disability of the elderly, and even increase the risk of death in severe cases. Currently, Taiwanese people are still unclear about the understanding and definition of sarcopenia, and many elderly people are diagnosed with sarcopenia only after functional impairment has occurred. In particular, the National Nutrition Survey in the United States conducted from 1999 to 2010 found that the prevalence of sarcopenia among adults over 60 years old is as high as 29.9%, and skeletal muscle mass will gradually lose with age, and changes in body composition due to aging will lead to a decrease in muscle strength or activity function.

On the other hand, with the development and progress of information and communication technology (ICT), the World Health Organization (WHO) has proposed the eHealth initiative in 2010, believing that any application of ICT technology in health-related fields to promote personal or community health can be called eHealth. According to the eHealth Global Survey Report published in 2016, more than half of WHO member states have formulated eHealth policies. The so-called eHealth refers to the use of communication technology in the health field, such as applications (Apps) is one of the applications.

Therefore, with the development of technology and the times, it is necessary to build an assessment model for sarcopenia based on ITC technology, combined with artificial intelligence and machine learning technology, and provide it to the general public through a very simple application or browser that everyone can operate, so that they can independently assess whether they belong to the high-risk group for sarcopenia.

Therefore, in view of the shortcomings of the conventional technology, the inventor has carefully tried and studied, and with a spirit of perseverance, finally conceived the present invention "Sarcopenia Assessment System and Method and Assessment Model Establishment Method", which can overcome the above shortcomings. The following is a brief description of the present invention.

In view of the shortcomings of the applied technology, the present invention proposes a sarcopenia assessment system and method, as well as a method for establishing a sarcopenia assessment model based on artificial intelligence machine learning technology, especially a system and method that can provide users with a simple operation of their own user equipment and remotely execute the sarcopenia assessment model to easily conduct sarcopenia risk assessment.

Based on this, the present invention proposes a sarcopenia assessment method, which includes: providing users with self-testing of sarcopenia through a front-end program and obtaining test results; uploading the test results to a sarcopenia assessment model built based on a machine learning classifier; and remotely executing the sarcopenia assessment model to perform sarcopenia assessment, so as to assess whether the user belongs to a high-risk group for sarcopenia based on the test results.

The present invention further proposes a sarcopenia assessment system, which includes: a system server, which is installed with a sarcopenia assessment intelligent platform including a sarcopenia assessment model built based on a machine learning classifier; and a user device, which is separately configured and communicatively connected to the system server and is installed with a front-end program of the sarcopenia assessment intelligent platform, so as to implement self-detection of sarcopenia for the user through a quick interface and obtain the detection result, wherein the user device uploads the detection result and inputs it into the sarcopenia assessment model, so as to automatically assess whether the user belongs to a high-risk group for sarcopenia according to the detection result.

The present invention further proposes a method for establishing a sarcopenia assessment model, which includes: implementing a sarcopenia action questionnaire survey program, providing a quick interface to multiple respondents through a front-end program to conduct a sarcopenia action questionnaire survey, and obtaining sarcopenia survey results; selectively implementing an annotation program, annotating the sarcopenia survey results according to predefined rules, distinguishing the sarcopenia survey results into at least two categories of results, and obtaining an annotation data set; and inputting the sarcopenia survey results or the annotation data set into a machine learning classifier to train the machine learning classifier to establish a sarcopenia assessment model. .

The above invention content is intended to provide a simplified summary of the present disclosure so that readers can have a basic understanding of the present disclosure. This invention content does not disclose a complete description of the present invention, and is not intended to point out the important/key elements of the embodiments of the present invention or to define the scope of the present invention.

10: The sarcopenia assessment system of the present invention

100: User equipment

101: Processor unit

103: Wireless RF communication module

105: Non-volatile memory unit

107: Touch screen unit

109: Applications

111: Quick operation program components

110: Mobile device

120:Desktop computer

125: Laptop computer

130: Smartphone

135: Tablet device

150:User

200: System server

250: Sarcopenia Database

260: Internet

310: Self-test instruction interface

312: Detect a shortcut key

314: Detection 2 shortcut keys

320: Detection-Self-detection interface

321:DAX numerical shortcut input field

323: Calf perimeter quick input field

325: Muscle mass quick input field

327: Mobility quick input field

329: Execute test shortcut keys

330: Sarcopenia assessment result interface

500: The sarcopenia assessment method of the present invention

501-509: Implementation steps

Figure 1 shows a schematic diagram of the system architecture of the sarcopenia assessment system of the present invention;

Figure 2 reveals the hardware network equipment architecture view of the system server and user equipment included in the sarcopenia assessment system of the present invention;

Figure 3 reveals the data feature distribution diagram revealed by the unsupervised machine learning classifier of the present invention extracting the original data of calf circumference below the normal value;

Figure 4 reveals the data feature distribution diagram revealed by the original data of muscle strength (hand grip strength) lower than normal value extracted by the unsupervised machine learning classifier of the present invention;

Figure 5 reveals the data feature distribution diagram revealed by the original data whose action power extracted by the unsupervised machine learning classifier of the present invention is lower than the normal value;

Figure 6 shows a schematic diagram of the sarcopenia self-testing instruction interface displayed on the user device of the sarcopenia assessment system of the present invention;

Figure 7 shows a schematic diagram of the detection-self-detection interface of the sarcopenia assessment system of the present invention displayed on the user device;

Figure 8 shows a schematic diagram of the sarcopenia assessment result interface displayed on the user device by the sarcopenia assessment system of the present invention; and

Figure 9 discloses a flowchart of the implementation steps of the sarcopenia assessment method of the present invention.

The present invention can be fully understood by the following embodiments, so that people familiar with the art can complete it accordingly. However, the implementation of the present invention is not limited by the following implementation cases; the drawings of the present invention do not include limitations on size, dimensions and scale. The size, dimensions and scale of the present invention in actual implementation are not limited by the drawings of the present invention.

The term "preferably" used herein is non-exclusive and should be understood as "preferably but not limited to". Any steps described or recorded in any specification or claim may be performed in any order, not limited to the order described in the claim. The scope of the invention shall be determined only by the attached claims and their equivalents, and shall not be determined by the embodiments exemplified in the implementation method. The term "including" and its variations herein, when appearing in the specification and claim, is an open term, has no restrictive meaning, and does not exclude other features or steps.

FIG. 1 is a schematic diagram of the system architecture of the sarcopenia assessment system of the present invention; FIG. 2 is a view of the hardware network equipment architecture of the system server and user equipment included in the sarcopenia assessment system of the present invention; the sarcopenia assessment system 10 of the present invention includes a system server 200 configured at the remote end, i.e., the server end, using a client-server model, and multiple user equipment (UE) 100 configured at the front end, i.e., the client end. The user equipment 100 is preferably a mobile device 110, a desktop computer 120, a laptop computer 125, a smart phone 130, or a tablet device 135, etc. A communication link is established between the system server 200 and the multiple user equipment 100 via the Internet 260, so that two-way communication and data exchange can be performed. The user device 100 includes hardware units such as a processor unit 101, a wireless RF communication module 103, a non-volatile memory unit (flash memory) 105, and a touch screen unit 107, which are electrically connected to each other, and software components such as an application 109 and a shortcut operation program component 111 included in the application 109. The application 109 as a front-end program is installed on the user device 100 and is loaded and executed through the processor unit 101.

When the user 150 opens the application 109 on the user device 100 and runs it on the processor unit 101, the application 109 loads the shortcut operation program component 111 into the processor unit 101 for execution. The application 109 generates a shortcut interface on the touch screen unit 107 through the shortcut operation program component 111. The shortcut interface is preferably, for example but not limited to: a series of graphical user interfaces (GUIs), a quick menu based on one-click operations, a quick operation web page, a quick list, a quick key, a graphical shortcut interface, a floating quick option menu window, a drop-down option quick menu, or a quick option menu, so as to provide the user 150 with the touch screen unit 107 for selection and operation.

Front-end applications can be applied to the two major device platforms of smartphones on the market, Android and iOS. For Android devices, front-end applications are developed using the Android Studio design editor. For remote server programs and software systems, it is better to use, for example but not limited to PHP scripts, to access the relational database MySQLDB to send and receive data. MySQLDB is a multi-user, multi-thread SQL database server that can effectively organize, file, and tabulate a database software for more efficient subsequent query, organization, transmission and reception of data. For iOS devices, applications are preferably developed using, for example but not limited to, Xcode and UIKit basic components to design the application (App) interface, and developed using Objective-C programming language or Swift programming language.

The processor unit 101 receives instructions issued by the user through the front-end application, and according to the instructions issued by the user, connects to the system server 200 through the wireless radio frequency communication module 103, and accesses the sarcopenia database 250, including reading (reading) required data or writing (writing) data, and downloads it to the client user device 100, through the application 109 and the shortcut operation program component 111 contained in the application 109, and finally displays it to the user on the touch screen unit 107 through a graphical user interface or a shortcut menu. The sarcopenia database 250 can be a storage unit built and stored inside the system server 200, or it can be built separately from the system server 200 and stored separately on other separate devices. Between the user device 100 application 109 and the system server 200 sarcopenia database 250, a communication protocol such as but not limited to HTTP/HTTPS is selected, and a format such as but not limited to JSON is used for data exchange and two-way communication.

Application 109 and shortcut program components 111 or other program components contained in application 109, as well as remote server programs or software systems, can be programmed and implemented using any programming language, such as but not limited to C, C++, C#, Java, VBScript, Macromedia Cold Fusion, COBOL, Microsoft Active Server Web, assembly language, PERL, PHP, awk, Python, Visual Basic, SQL stored procedure, PL/SQL, any UNIX command description language, and extensible markup language (XML) with various algorithms implemented in any combination of data structures, objects, procedures, routines or other program components.

In one embodiment, the program components, modules, or software engines configured on the user device 100 can be implemented in the form of applications 109 or micro applications (micro Apps), and these applications 109 or micro applications can be executed in a mobile operating system environment, and the mobile operating system includes, for example, Palm mobile operating system, Windows mobile operating system, Android operating system, Apple iOS, Blackberry operating system, etc.

The sarcopenia assessment system proposed in the present invention can be built as an application for user operation in the front-end program part, and can also be built through an off-premises method using cloud technology such as software as a service (SaaS) or platform as a service (PaaS), and directly provided to users for operation through a web browser (internet browser) running on the user device 100 as a front-end user interface, and provide users with on-the-spot operation services. When using SaaS or PaaS services, users only need to obtain permissions to access and use the system online, and users do not need to install applications on their mobile devices.

The sarcopenia assessment system of the present invention and the applications included therein are composed of various hardware in the physical device layer and application/software platform/computer program products in the application layer. According to the international open system interconnection communication reference model (OSI/RM) architecture, the medical material point-to-point operation application of the present invention and the shortcut function program components included therein are software application services executed and operated on the 7th layer (application layer) of the OSI/RM architecture. Application services can independently select various communication protocols in the 4th layer transmission layer, form data packets and determine the transmission path in the 3rd layer network layer, add logical link control (LLC) and media access control (MAC) through the 2nd layer data link layer, and then establish the required upload and download communication links with various devices on the 1st layer physical layer, such as but not limited to multiple user devices 100, smart phones 130, tablet devices 135, system servers 200, etc.

The sarcopenia assessment model proposed in the present invention is preferably based on an artificial intelligent machine learning classifier, covering supervised classifiers and non-supervised classifiers. The construction steps at least include sarcopenia action questionnaire survey procedures, data pre-processing procedures, labeling procedures, training procedures, verification procedures, and testing procedures.

First, we will implement the Sarcopenia Action Questionnaire Survey to collect a large amount of raw data as model training data. With the penetration of information and communication technologies such as smartphones and tablets into our daily lives, providing mobile questionnaires through applications or web browsers can effectively reach a large number of respondents, help systematically and efficiently collect a large amount of raw data, and overcome the problem of large-scale questionnaire surveys being difficult to implement.

This invention refers to the guidance provided by the MARS technology to build the application and user interface (UI) required for the Sarcopenia Action Questionnaire Survey, as well as the user interface for web browser loading, and designs the relevant Sarcopenia Action Questionnaire. After providing a large number of users with the ability to download and answer the questionnaire through mobile devices, the data collection work for a large number of original samples of Sarcopenia can be effectively completed in a relatively short period of time.

MARS technology divides applications into five quality levels, namely functionality, information quality, aesthetics, subjective quality and engagement. Through scientifically based assessments, it is possible to evaluate the suitability of mobile health applications for patients' disease cognition and nursing guidance and care. The results can also serve as a reference for application developers in designing related applications in the future.

The content of the Sarcopenia Action Questionnaire is designed according to the 5-point scoring method proposed by the Likert Scale principle, where strongly disagree is 1 point, disagree is 2 points, average is 3 points, agree is 4 points, and strongly agree is 5 points. The designed questionnaire scale was also handed over to a number of experts, such as medical information experts, engineering professors, and information engineers, to test and revise the appropriateness and clarity of the questionnaire content to ensure the expert validity of the questionnaire scale. The overall validity index (Total CVI) of the action questionnaire scale of the present invention is controlled to at least 0.90, and the Cronbach's α value of internal consistency is greater than 0.9.

For example, the sarcopenia action questionnaire designed by the present invention according to the MARS technology and Likert scale principle is preferably as shown in Table 1 below:

The sarcopenia action questionnaire designed by the present invention has a total scoring method of MARS, with the functionality and subjective quality aspects divided into two points, and the cumulative total score is 0-20 points. The average total score of the functionality aspect is 19.16±1.838 points, and the average total score of the subjective quality aspect is 19.15±19.13 points. Through the use of action questionnaire technology, a large-scale action questionnaire survey can be efficiently carried out, overcoming the difficulty of large-scale questionnaire surveys in the application of technology.

After the Sarcopenia Action Questionnaire Survey is completed, the collected raw data must first undergo data pre-processing, including outlier removal, erroneous data removal, incomplete record removal, field recognition, text recognition, semantic recognition, format conversion, standardization or formatting, etc., to remove incorrect and missing records, unreliable and possibly false answers, etc., and only retain correct and reliable records to complete the data pre-processing operation. Since the present invention uses an application or web browser to provide a mobile questionnaire survey, the original data collected has been stored and recorded in the cloud database in the form of digital files, which can omit the traditional paper questionnaire survey, which requires the process of converting paper documents into digital data, and the possible data errors, loss, recognition errors, etc.

The results of the sarcopenia survey will include responses from many respondents. The more respondents, the better. It should preferably cover those who have been diagnosed with sarcopenia, those at high risk of sarcopenia, those at low risk of sarcopenia, and those without sarcopenia, as well as those with and without chronic diseases, types of chronic diseases, and those who take long-term medications. This will help expand the number of parent samples and make the data characteristics contained in the overall data closer to unbiased. In this example, a total of 112 people were enrolled, and the presence or absence of chronic diseases, types of chronic diseases, and long-term medications were collected. A total of 10 types of chronic diseases and 10 types of long-term medications were collected. They were respectively correlated with the three major items of sarcopenia screening: calf circumference, muscle strength (hand grip strength), and mobility. The chi-square test of statistical analysis was initially used to analyze the correlation and risk estimation of sarcopenia. The results showed that diabetes had a significant impact on the probability of low normal values of calf circumference, muscle strength (hand grip strength), and mobility.

The characteristics of the collected raw data are as follows. The average values of the respondents' calf circumference, muscle strength (hand grip strength) and mobility are listed in Table 2. The average calf circumference of the respondents is 34.88±4.62 cm, of which 77.7% of the respondents have calf circumferences within the normal range, accounting for the majority. There are 25 respondents with calf circumferences below the normal range (less than 34 cm for boys and less than 32 cm for girls), accounting for 22.3% of the respondents. The average muscle strength (hand grip strength) was 23.88±10.25 kg, most of them were within the normal range, accounting for 68.8%, and 35 people were below the normal value (less than 26 kg for boys and less than 18 kg for girls), accounting for 31.3%; the average mobility was 1.01±0.23 m/s, 18 people were below the normal value, accounting for 16.1%, and the rest were within the normal range, accounting for 83.9%.

The risk estimates of the calf circumference of the respondents and chronic diseases and long-term medications are listed in Tables 3 and 4 below. Among the 10 chronic diseases and long-term medications, diabetes has a significant effect on the probability of having a calf circumference below normal (p=0.005, OR=6.378), which means that patients with diabetes have a 6.378 times higher risk of having a calf circumference below normal than those without diabetes. Long-term use of hypoglycemic drugs also has a significant effect on the probability of having a calf circumference below normal (p=0.025, OR=5.188), which means that patients taking hypoglycemic drugs have a 5.188 times higher risk of having a calf circumference below normal than those not taking hypoglycemic drugs.

The following table 5 lists the risk estimates of the relationship between the respondents' muscle strength (hand grip strength) and chronic diseases. Among the 10 types of chronic diseases, diabetes has a significant effect on the probability of low normal muscle strength (hand grip strength) (p=0.009, OR=5.407), which means that patients with diabetes have a 5.407 times higher risk of muscle strength (hand grip strength) below normal than those without diabetes. And hypertension also has a significant effect on the probability of low normal muscle strength (hand grip strength) (p=0.006, OR=3.375), which means that patients with hypertension have a 3.375 times higher risk of muscle strength (hand grip strength) below normal than those without hypertension.

The risk estimates of the respondents' muscle strength (hand grip strength) and long-term medication are listed in Table 6. Among the 10 types of long-term medications, long-term use of antihypertensive drugs also has a significant effect on the probability of low normal muscle strength (p=0.001, OR=4.500), indicating that patients taking antihypertensive drugs have a 4.5 times higher risk of having a calf circumference below the normal value than those who do not take antihypertensive drugs.

The risk estimates of the respondents' mobility and chronic diseases and long-term medications are listed in Tables 7 and 8. Among the 10 chronic diseases and long-term medications, diabetes has a significant effect on the probability of low normal mobility (p=0.024, OR=4.780), which means that patients with diabetes have a 4.78 times higher risk of low normal mobility than those without diabetes. Long-term use of hypoglycemic drugs also has a significant effect on the probability of low normal mobility (p=0.036, OR=5.086), which means that patients with diabetes have a 5.086 times higher risk of low normal mobility than those without diabetes.

Figure 3 reveals the data feature distribution diagram revealed by the original data of calf circumference lower than normal value extracted by the unsupervised machine learning classifier of the present invention; Figure 4 reveals the data feature distribution diagram revealed by the original data of muscle strength (hand grip strength) lower than normal value extracted by the unsupervised machine learning classifier of the present invention; Figure 5 reveals the data feature distribution diagram revealed by the original data of movement force lower than normal value extracted by the unsupervised machine learning classifier of the present invention.

In this embodiment, the original data is input into a machine learning unsupervised classifier to extract the features contained in the original data, analyze the calf circumference, muscle strength (hand grip strength) and mobility risk estimation of the respondents, and establish a basic classification and prediction model for sarcopenia risk assessment; for example, calf circumference less than normal, muscle strength (hand grip strength) less than normal and mobility less than normal are set as the target variables of the classifier, and age, gender, BMI, hypertension, glaucoma, liver disease, sciatica Pain, diabetes, hyperlipidemia, asthma, rheumatoid arthritis, gastroesophageal reflux, cardiovascular disease, cardiovascular vascular drugs, antihypertensive drugs, liver drugs, stomach drugs, hypoglycemic drugs, hormone preparations, antihyperlipidemic drugs, sleeping pills, analgesics, and Chinese medicine are set as independent variables of the classifier. The overall accuracy of the model is 73% for calf circumference below normal values, 66.7% for muscle strength (hand grip strength) below normal values, and 75.7% for mobility below normal values. Then the original data is input into the machine learning classifier to extract the data feature distribution map. The extracted feature maps are shown in Figures 3 to 5.

According to the disclosure in Figure 3, it is obvious that for the group with calf circumference below normal, the most important data characteristics are age and body mass index (BMI). According to the disclosure in Figure 4, it is obvious that for the group with muscle strength (hand grip strength) below normal, the most important data characteristics are also age and body mass index. According to the disclosure in Figure 5, it is obvious that for the group with mobility below normal, the most important data characteristics are also age and body mass index.

Then the annotation process is implemented. The annotation process is to use the correct data retained after data pre-processing, and for each respondent's answer, according to predefined rules, such as but not limited to: the European Union Working Group on Sarcopenia (EWGSOP) operational definition of sarcopenia, the International Working Group on Sarcopenia (IWGS) operational definition of sarcopenia, the Asian Working Group on Sarcopenia (AWGS) operational definition of sarcopenia, preliminary statistical analysis results, data feature distribution diagrams or user-defined operational definitions of sarcopenia, etc., to give each respondent a different sarcopenia risk level label.

For example, after selecting the predefined rules to be followed, in accordance with the EWGSOP sarcopenia operational definition rules, the sarcopenia risk is further divided into two levels such as common and high risk, or the sarcopenia risk is divided into three levels such as low risk, medium risk and high risk, or the sarcopenia risk is divided into more risk levels, and each respondent in the original data is labeled according to the original data EWGSOP sarcopenia operational definition.

Tables 9 and 10 list the operational definitions and recommended screening methods for sarcopenia in EWGSOP, IWGS, and AWGS.

The above statistical analyses are preferably performed using statistical software (SPSS 24.0/G-power3.1), with the α value set to 0.05, the power to 0.8, and the effect size to 0.3. The estimated sample size is 100 people, and considering the attrition rate of 20%, the estimated total sample size is 120 people.

Finally, the training dataset and the test dataset are split, where the training dataset includes the validation dataset. The annotated original data text is split into the training dataset and the test dataset at a ratio of approximately 1:1 for fine-tuning the sarcopenia assessment model. After the model learns and interprets the annotated answer result text, the construction of the sarcopenia assessment model can be completed. In this embodiment, more than 1,000 answer results are collected for training.

The machine learning classifier used in the present invention is preferably, for example but not limited to: supervised classifier, unsupervised classifier, regression tree classifier, random forest classifier, decision tree classifier, boost tree classifier, gradient boost tree classifier, strong gradient boosting machine classifier, weak gradient boosting machine classifier, ensemble learning classifier, or support vector machine (SVM), etc.

In summary, the sarcopenia assessment model proposed in the present invention is essentially a multi-class classification. The supervised machine learning technology used in the sarcopenia assessment model is mainly based on the clustering method. The data is distinguished by grouping, and the data is clustered by establishing similar clusters, so that the data differences in the same cluster are minimized and the similarities are maximized, while the differences between different clusters are maximized and the similarities are minimized. That is, this technology is used to reduce the complexity of the data and find the common features in the same cluster. The main purpose of clustering is to find the differences between groups and the similarities in the same group, so that the differences within the group are small and the differences outside the group are large. The most commonly used clustering analysis algorithm techniques currently include two categories: segmentation clustering algorithm and hierarchical clustering algorithm. The number of clusters in the present invention is known, and the segmentation clustering method is suitable. With such a system model, it can be used to estimate, predict, make decisions, diagnose sarcopenia, and assess the risk level for each user.

The sarcopenia assessment system of the present invention preferably applies SaaS and PaaS technologies, and provides users with a web browser running on the user device 100 as a front-end user interface, or provides users with an application running on the user device 100 as a front-end user interface, and generates a series of graphical user interfaces as shown in Figures 6 to 8 through the browser or application, and provides users with a one-click operation and a quick menu based on a quick interface, so as to provide users with a sarcopenia self-assessment service.

FIG. 6 is a schematic diagram of a sarcopenia self-testing description interface displayed on a user device by the sarcopenia assessment system of the present invention; FIG. 7 is a schematic diagram of a self-testing interface displayed on a user device by the sarcopenia assessment system of the present invention; FIG. 8 is a schematic diagram of a sarcopenia assessment result interface displayed on a user device by the sarcopenia assessment system of the present invention; in this embodiment, the user device is preferably illustrated by, for example but not limited to, a smart phone, and an application installed on the mobile phone.

The user clicks on the sarcopenia application on the touch screen of the smart phone 130 and clicks on the option to enter the self-test. First, a self-test instruction interface 310 as shown in FIG. 6 will be displayed. The self-test instruction interface 310 shown in FIG. 6 is a graphical user interface (GUI). The system explains the correct various sarcopenia detection methods to the user through the content displayed in the self-test instruction interface 310, so that the user can operate and perform self-testing. After the user understands the sarcopenia detection method, he can choose to perform self-testing of test one or test two. The self-test instruction interface 310 provides a test one shortcut key 312 and a test two shortcut key 314. In this embodiment, the user chooses to perform test one and presses the test one shortcut key 312.

The system provides a self-test interface 320 for test 1 as shown in FIG. 7 to assist the user in performing test 1. The self-test interface 320 for test 1 provides multiple quick input fields, including at least a DAX value quick input field 321, a calf circumference quick input field 323, a muscle mass quick input field 325, and a mobility quick input field 327, for the user to input the information required by the sarcopenia assessment system and model according to the prompts of the quick input fields. The system will continue to use other interfaces to ask the user whether he has chronic diseases and whether he takes long-term medications, etc. When the user completes all input, he presses, for example but not limited to, the execution detection shortcut key 329, and the application returns the input information to the remote system server 200 and sends a model execution instruction to request the sarcopenia assessment smart platform to execute the sarcopenia assessment model. By executing the sarcopenia assessment model, it is assessed whether the user belongs to the high-risk group for sarcopenia.

The system server 200 returns the assessment result to the smartphone 130 application, and prompts the user with the assessment result of the artificial intelligence model through the sarcopenia assessment result interface 330 shown in FIG. 8. In this embodiment, the sarcopenia assessment model determines whether the user is in the high-risk group for sarcopenia based on the information input by the user, and displays such assessment result and related details including probability value and risk factors to the user on the sarcopenia assessment result interface 330, so that the user can read and understand his or her own sarcopenia risk status.

FIG. 9 is a flowchart of the implementation steps of the sarcopenia assessment method of the present invention. In summary, the sarcopenia assessment method 500 of the present invention preferably includes the following steps: implementing a sarcopenia action questionnaire survey program, providing a quick interface to multiple respondents through a front-end program to conduct a sarcopenia action questionnaire survey, and obtaining sarcopenia survey results (step 501); inputting the sarcopenia survey results or the annotated data set into a machine learning classifier to train the machine learning classifier; The sarcopenia assessment model is established by the classifier (step 503); the user is provided with a self-test of sarcopenia through the front-end program, and the test result is obtained (step 505); the test result is uploaded to the sarcopenia assessment model established based on the machine learning classifier (step 507); and the sarcopenia assessment model is remotely executed to perform sarcopenia assessment, so as to assess whether the user belongs to a high-risk group for sarcopenia according to the test result (step 509).

The above embodiments of the present invention can be arbitrarily combined or replaced with each other to derive more embodiments, but they are within the scope of protection of the present invention. More embodiments of the present invention are further provided as follows:

Embodiment 1: A sarcopenia assessment method, comprising: providing a user with a self-test for sarcopenia through a front-end program and obtaining the test result; uploading the test result to a sarcopenia assessment model built based on a machine learning classifier; and remotely executing the sarcopenia assessment model to perform sarcopenia assessment, so as to assess whether the user belongs to a high-risk group for sarcopenia based on the test result.

Embodiment 2: The sarcopenia assessment method as described in Embodiment 1, wherein the sarcopenia assessment model is established according to the sarcopenia assessment model establishment method, and the sarcopenia assessment model establishment method comprises: implementing a sarcopenia action questionnaire survey procedure, providing a quick interface to multiple respondents through a front-end program to conduct a sarcopenia action questionnaire survey, and obtaining sarcopenia survey results; selectively implementing an annotation procedure, annotating the sarcopenia survey results according to predefined rules, distinguishing the sarcopenia survey results into at least two categories of results, and obtaining an annotation data set; and inputting the sarcopenia survey results or the annotation data set into a machine learning classifier to train the machine learning classifier to establish a sarcopenia assessment model.

Example 3: The method for establishing a sarcopenia assessment model as described in Example 2, wherein the predefined rule is selected from one of the European Union Working Group on Sarcopenia (EWGSOP) operational definition of sarcopenia, the International Working Group on Sarcopenia (IWGS) operational definition of sarcopenia, the Asian Working Group on Sarcopenia (AWGS) operational definition of sarcopenia, and a user-defined operational definition of sarcopenia.

Example 4: The method for establishing a sarcopenia assessment model as described in Example 2, wherein the machine learning classifier is selected from a supervised classifier, an unsupervised classifier, a regression tree classifier, a random forest classifier, a decision tree classifier, a boosted tree classifier, a gradient boosted tree classifier, a strong gradient boosting machine classifier, a weak gradient boosting machine classifier, an ensemble learning classifier, a support vector machine, and a combination thereof.

Embodiment 5: A sarcopenia assessment system, comprising: a system server, which is installed with a sarcopenia assessment intelligent platform including a sarcopenia assessment model built based on a machine learning classifier; and a user device, which is separately configured and communicatively connected to the system server and is installed with a front-end program of the sarcopenia assessment intelligent platform, so as to implement a self-detection of sarcopenia for the user through a quick interface and obtain the detection result, wherein the user device uploads the detection result and inputs it into the sarcopenia assessment model, so as to automatically assess whether the user belongs to a high-risk group for sarcopenia according to the detection result.

Embodiment 6: In the sarcopenia assessment system described in Embodiment 5, the user device is one of a mobile device, a smart phone, a tablet device, a desktop computer and a laptop computer.

Embodiment 7: In the sarcopenia assessment system described in Embodiment 5, the front-end program is one of an application and a web browser.

Embodiment 8: In the sarcopenia assessment system described in Embodiment 5, the shortcut interface is selected from a graphical user interface, a shortcut menu, a shortcut operation web page, a shortcut list, a shortcut key, a graphical shortcut interface, a floating shortcut option menu window, a drop-down option shortcut menu, a shortcut menu interface, a shortcut option interface, a shortcut option menu, and a shortcut selection key.

Embodiment 9: A method for establishing a sarcopenia assessment model, comprising: implementing a sarcopenia action questionnaire survey program, providing a quick interface to multiple respondents through a front-end program to conduct a sarcopenia action questionnaire survey, and obtaining sarcopenia survey results; selectively implementing a labeling program, labeling the sarcopenia survey results according to predefined rules, distinguishing the sarcopenia survey results into at least two categories of results, and obtaining a labeling data set; and inputting the sarcopenia survey results or the labeling data set into a machine learning classifier to train the machine learning classifier to establish a sarcopenia assessment model.

The various embodiments of the present invention can be combined or replaced with each other at will, thereby deriving more implementation forms, but they are within the scope of protection of the present invention. The definition of the protection scope of the present invention shall be based on the scope of the patent application of the present invention.

500: The sarcopenia assessment method of the present invention

501-509: Implementation steps

Claims (5)

A sarcopenia assessment method, comprising: providing a quick operation interface through an application program executed on a smart phone, so as to provide a user with a sarcopenia self-motor test based on an electronic questionnaire through the quick operation interface, so that the user can input at least one of age information, BMI information, DXA value, calf circumference, muscle mass and mobility, and obtain a test result; uploading the test result to a sarcopenia assessment model built based on a supervised random forest classifier, wherein the sarcopenia assessment model is a sarcopenia survey result of a sarcopenia action questionnaire designed with an action application rating scale and a Likert scale as a guide, wherein the sarcopenia survey result is based on A labeled sarcopenia data set is generated by labeling according to an Asian Working Group on Sarcopenia (AWGS) operational definition of sarcopenia, wherein the supervised random forest classifier is trained with the labeled sarcopenia data set to learn the labeled sarcopenia data set, and the trained supervised random forest classifier is converted into an assessment model with the ability to assess sarcopenia. To establish the sarcopenia assessment model, wherein the supervised random forest classifier determines the age information, the DXA value, the calf circumference, the muscle mass, the mobility and the BMI information as the main causes of sarcopenia after learning; and remotely execute the sarcopenia assessment model to assess whether the user belongs to the high-risk group for sarcopenia based on the test results. The sarcopenia assessment method as described in claim 1, wherein the sarcopenia assessment model is established according to a sarcopenia assessment model establishment method, the sarcopenia assessment model establishment method comprising: implementing a sarcopenia action questionnaire survey procedure, providing the quick operation interface to a plurality of respondents through an application to conduct the sarcopenia action questionnaire survey, and obtaining the sarcopenia survey results; selectively implementing a labeling procedure, labeling the sarcopenia survey results according to a predefined rule, distinguishing the sarcopenia survey results into at least two categories of results, and obtaining the labeled sarcopenia data set; and inputting the sarcopenia survey results or the labeled sarcopenia data set into the supervised random forest classifier to train the supervised random forest classifier to establish the sarcopenia assessment model. The sarcopenia assessment method as described in claim 2, wherein the predefined rule is selected from one of a European Union Working Group on Sarcopenia (EWGSOP) sarcopenia operational definition, an International Working Group on Sarcopenia (IWGS) sarcopenia operational definition, the Asian Working Group on Sarcopenia (AWGS) sarcopenia operational definition, and a user-defined sarcopenia operational definition. A sarcopenia assessment system, comprising: a system server, on which is installed a sarcopenia assessment intelligent platform including a sarcopenia assessment model based on a supervised random forest classifier; and a smart phone, which is separately configured and communicatively linked to the system server and is installed with an application of the sarcopenia assessment intelligent platform to provide a quick operation interface, so as to implement a sarcopenia autonomous action test based on an electronic questionnaire for a user through a quick operation interface, and obtain a test result, wherein the quick operation interface allows the user to input at least one of age information, BMI information, DXA value, calf circumference, muscle mass and mobility, wherein the smart phone uploads the test result and inputs it into the sarcopenia assessment model to automatically evaluate the user according to the test result, especially the age information and the BMI information. Whether they belong to the high-risk group for sarcopenia, wherein the sarcopenia assessment model is built based on a sarcopenia survey result of a sarcopenia action questionnaire designed with an action application rating scale and a Likert scale as a guide, wherein the sarcopenia survey result is annotated according to the operational definition of sarcopenia of the Asian Working Group on Sarcopenia (AWGS) to generate an annotated sarcopenia data set, wherein the supervised random forest classifier is based on The labeled sarcopenia data set is trained to learn the labeled sarcopenia data set, and accordingly the trained supervised random forest classifier is converted into an assessment model with the ability to assess sarcopenia to establish the sarcopenia assessment model, wherein the supervised random forest classifier determines the age information, the DXA value, the calf circumference, the muscle mass, the mobility and the BMI information as the main causes of sarcopenia after learning. The sarcopenia assessment system as described in claim 4, wherein the quick operation interface is selected from a graphical user interface, a quick menu, a quick operation web page, a quick list, a quick key, a graphical quick interface, a floating quick option menu window, a drop-down option quick menu, a quick menu interface, a quick option interface, a quick option menu, and a quick selection key.

Images