Simple multi-modal index rapid auxiliary screening system for mild cognitive impairment of middle-aged and elderly people

Abstract

The invention belongs to the technical field of auxiliary medical treatment, and particularly relates to a simple multi-modal index rapid auxiliary screening system for middle-aged and elderly mild cognitive impairment. The system includes: an input module configured to input blood indicators, physical examination data, and inquiry variables of a subject; the prediction module is configured to input the blood indexes, the physical examination data and the inquiry variables into a machine learning model to obtain a prediction result whether the middle-aged and elderly mild cognitive impairment risk exists or not; and the output module is configured to output the prediction result. The MCI screening model which can accurately predict the mild cognitive impairment of the middle-aged and elderly people by using simply acquired indexes is successfully constructed. On the basis of ensuring the prediction performance, the model is simple in index acquisition and strong in generalization ability, is beneficial to large-scale popularization of the technology, and has a very good application prospect.

Description

Technical Field

[0001] The present invention belongs to the field of auxiliary medical technology, and in particular relates to a rapid auxiliary screening system for mild cognitive impairment in the elderly using simple multimodal indicators. Background Art

[0002] Mild cognitive impairment (MCI) is the early stage of dementia. How to detect mild MCI patients in apparently healthy community residents at an early stage is the top priority for early prevention and control of dementia. Although large-scale population screening for mild cognitive impairment is not possible due to various factors, how to develop and apply new screening tools to achieve rapid, accurate, and intelligent large-scale screening of mild cognitive impairment in community residents has become the direction of joint efforts of scholars around the world.

[0003] Among the overall cognitive function assessment tools, the 2018 Chinese Guidelines for the Diagnosis and Treatment of Dementia and Cognitive Impairment (V): Diagnosis and Treatment of Mild Cognitive Impairment lists the Montreal Cognitive Assessment (MoCA) scale as an "A-level recommendation" for early screening of mild cognitive impairment, which is the gold standard for MCI screening. The scale was compiled by Nasreddine scholars in 2005 and is specifically used for screening of mild cognitive impairment. The assessment method is peer review, and the assessment content includes eight parts: visual-spatial and executive function, naming, memory, attention, language, abstraction, delayed recall and orientation. It fully considers the educational level of the person being assessed and has high sensitivity (80-100%) and specificity (50-76%).

[0004] In the related research in the interdisciplinary field of cognitive impairment and machine learning, the screening model research of MCI is still in the early stages of development, mainly focusing on the prediction research of late-stage diseases such as Alzheimer's disease, while there are relatively few studies on auxiliary screening and auxiliary diagnosis of mild cognitive impairment in the early stages of the disease. Domestic and foreign scholars have found that the effects of traditional methods are not ideal in the exploration of mild cognitive impairment screening models using traditional regression models and other methods, but with the rapid development of artificial intelligence, machine learning methods may be one of the potential possibilities for solving such problems in the future.

[0005] In current research, the use of machine learning methods for auxiliary diagnosis of mild cognitive impairment usually requires the use of some medical diagnostic indicators, such as cerebrospinal fluid indicators (such as tau, Aβ, etc.), imaging features (such as MRI, PET, etc.), bioelectric signals (such as EEG signals, etc.). The collection of these indicator data has problems such as high difficulty in collection, high collection fees / indicator detection costs, expensive data collection equipment or high-net-value consumables, and patients facing a higher risk of invasive / traumatic infection. These problems limit the widespread application of machine learning technology in the auxiliary diagnosis of mild cognitive impairment.

[0006] Therefore, under the premise of ensuring good diagnostic performance, developing a technology that can use simple and easily available indicators to assist in the diagnosis of mild cognitive impairment is an issue that urgently needs to be addressed in this field. Summary of the invention

[0007] In view of the problems of the prior art, the present invention provides a simple multimodal index system for rapid auxiliary screening of mild cognitive impairment in the elderly.

[0008] A simple multimodal index rapid auxiliary screening system for mild cognitive impairment in the elderly, including: An input module, configured to input blood indicators, physical examination data and inquiry variables of the subject; A prediction module is configured to input the blood index, physical examination data and query variables into a machine learning model to obtain a prediction result of whether the risk of mild cognitive impairment in the middle-aged and elderly is present; The output module is configured to output the prediction result.

[0009] Preferably, the blood index is selected from at least one of the following test results: white blood cell count, neutrophil count, lymphocyte count, hemoglobin, hematocrit, mean corpuscular hemoglobin concentration, indirect bilirubin, platelet count, monocyte count, eosinophil count, and total bilirubin.

[0010] Preferably, the physical examination data is selected from at least one of the following test results: calf circumference / neck circumference, height / calf circumference, height / neck circumference, calf circumference / mid-upper arm circumference, calf circumference, height / mid-upper arm circumference.

[0011] Preferably, the query variable is selected from delayed recall.

[0012] Preferably, the blood index is selected from: white blood cell count, neutrophil count, lymphocyte count, hemoglobin, hematocrit, mean corpuscular hemoglobin concentration, indirect bilirubin, platelet count, monocyte count, eosinophil count, and total bilirubin; The physical examination data is selected from: calf circumference / neck circumference, height / calf circumference, height / neck circumference, calf circumference-upper arm circumference, calf circumference, height / upper arm circumference; The query variable is selected from delayed recall.

[0013] Preferably, the algorithm of the machine learning model is selected from the extreme gradient boosting algorithm.

[0014] Preferably, the parameter settings of the machine learning model include: sampling rate = 1, learning rate = 0.001, maximum depth = 6, subsample = 0.5.

[0015] The present invention also provides a computer-readable storage medium, on which is stored: a computer program for implementing the above-mentioned simple multimodal index rapid auxiliary screening system for mild cognitive impairment in the elderly.

[0016] In the present invention, the symbol " / " represents a division symbol, for example: "calf circumference / neck circumference" refers to the ratio of the calf circumference divided by the neck circumference.

[0017] The present invention realizes a system for machine learning-assisted diagnosis of mild cognitive impairment (MCI screening model) using simple and readily available indicators. By optimizing the indicator type and model algorithm, a high prediction performance is achieved, which can meet the application needs of rapid auxiliary screening of mild cognitive impairment in the middle-aged and elderly. Specifically, the technical solution of the present invention can achieve the following beneficial technical effects: 1. The input indicators selected by the present invention are simple to collect, low in cost, and do not require the purchase of additional large equipment. Medical institutions at all levels, including community health service centers and township health centers, can complete the collection, which solves the difficulties in collecting indicators used in the prior art (such as cerebrospinal fluid indicators, imaging features, and bioelectric signals), and is conducive to the large-scale promotion and application of machine learning methods.

[0018] 2. The MCI screening model of the present invention has better results in cross-validation of the model establishment data set than in the past (AUC>0.8). At the same time, the screening model was independently validated externally in different years and on the basis of populations different from those in the model establishment data set, and its AUC result was still higher than 0.8, proving that the MCI screening model has good generalization ability.

[0019] In short, the MCI screening model of the present invention simplifies the difficulty of collecting input indicators on the basis of ensuring the prediction performance of the model, which is conducive to its promotion. The application scenarios of the present invention for early community screening of mild cognitive impairment may include primary medical and health service centers, biological sample testing companies, physical examination centers, health management centers, etc., and the application prospects are very good.

[0020] Obviously, according to the above contents of the present invention, in accordance with common technical knowledge and customary means in the art, without departing from the above basic technical ideas of the present invention, other various forms of modification, replacement or change may be made.

[0021] The above contents of the present invention are further described in detail below through specific implementation methods in the form of embodiments. However, this should not be understood as the scope of the above subject matter of the present invention being limited to the following examples. All technologies realized based on the above contents of the present invention belong to the scope of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 Establish a technical roadmap for the MCI screening model in Experimental Example 1; Figure 2 A path diagram for the establishment and external validation of the MCI screening model in Experimental Example 1; Figure 3 The ROC curve diagram of the best MCI screening model obtained in Experimental Example 1 in the model establishment data set; Figure 4 The figure is the ROC curve diagram of the evaluation of the optimal MCI screening model obtained in Experimental Example 1 in an external independent validation data set. DETAILED DESCRIPTION

[0023] It should be noted that the algorithms of data collection, transmission, storage and processing steps not specifically described in the embodiments, as well as the hardware structure, circuit connection, etc. not specifically described can all be implemented through the contents disclosed in the prior art.

[0024] Example 1 Simple multimodal index rapid auxiliary screening system for mild cognitive impairment in the elderly The system of this embodiment includes: An input module, configured to input blood indicators, physical examination data and inquiry variables of the subject; A prediction module is configured to input the blood index, physical examination data and inquiry variables into a machine learning model (hereinafter referred to as an MCI screening model) to obtain a prediction result of whether there is a risk of mild cognitive impairment in the middle-aged and elderly; The output module is configured to output the prediction result.

[0025] The blood index is selected from the group consisting of: white blood cell count, neutrophil count, lymphocyte count, hemoglobin, hematocrit, mean corpuscular hemoglobin concentration, indirect bilirubin, platelet count, monocyte count, eosinophil count, and total bilirubin; The physical examination data is selected from: calf circumference / neck circumference, height / calf circumference, height / neck circumference, calf circumference-upper arm circumference, calf circumference, height / upper arm circumference; The inquiry variable is selected from delayed recall. The quantitative determination method of delayed recall is: before starting the body dimension measurement, select 5 words, record and repeat reading twice, for example: "face", "velvet", "church", "chrysanthemum", "red", and then start the body dimension measurement. These 5 words are not prompted during the measurement. After all the measurements are completed (more than 5 minutes), ask whether the 5 words are remembered. No prompts are given, and there is no need to recall them in order. Each time one is recalled, 1 point is recorded, and the full score is 5 points. Just enter the score.

[0026] In this embodiment, the algorithm of the machine learning model is selected from the extreme gradient boosting algorithm. The parameter settings of the machine learning model include: sampling rate=1, learning rate=0.001, maximum depth=6, subsample=0.5.

[0027] The technical solution of the present invention is further illustrated below through experimental data.

[0028] Experimental Example 1 Establishment and external validation of MCI screening model 1. Experimental Methods This experimental case constructed a cognitive health cohort of the natural population in the community, adopted a snowball multi-stage sample enrollment strategy, established and evaluated the community MCI screening model with the help of the first-stage baseline sample population data, and used the second-stage snowball baseline additional sample population data to conduct external independent verification and evaluation of the community MCI screening model to examine the model's generalization ability.

[0029] The overall technical roadmap and model technical roadmap of the present invention are shown in Figure 1 and Figure 2 Specifically, the technical path for establishing the model mainly includes three aspects: 1. Model construction and external validation data collection technology based on community cohort studies: Build a cognitive health cohort of natural populations in the community, obtain cognitive-related assessment data of cohort study samples, and provide real-world data support for the construction and external validation of rapid screening models for mild cognitive impairment. This study recruited samples of cohort study populations in the natural population community of Chengdu. Taking the fixed-group research strategy as the starting point, a comprehensive cognitive health assessment was conducted on all subjects. The basic demographic information, cognitive and influencing factor assessment data, physical examination and biological specimen detection information of the sample population were systematically collected to establish a high-quality cognitive health cohort of natural populations in the community. The "three-level quality control" strategy was adopted throughout the establishment of the community cohort, and quality control work was carried out at the three levels of preparation, execution, and maintenance, laying a solid foundation for the accurate collection of real-world data, the conduct of cognitive-related scientific research, and sustainable cohort development.

[0030] 2. Construction of community MCI screening model and internal evaluation technology: Based on the biological specimens and evaluation data of the sample population of the cohort study, with the help of different model training methods, a rapid screening model for mild cognitive impairment for middle-aged and elderly residents in the community is established and internally evaluated. Using the baseline data of the first phase of the community natural population cognitive health cohort population sample enrollment, try to use a variety of methods to screen variables such as biological specimen indicators and demographic information, and combine professional field knowledge injection and other methods to select highly correlated variables; by comparing the predictive effectiveness between traditional models and machine learning models, select models or model combinations with better generalization ability for screening and prediction of mild cognitive impairment in the community, and complete the internal evaluation of model prediction.

[0031] 3. External validation and feature interpretability analysis technology of community MCI screening model: With the help of the snowballing of new sample data in the cohort study, the established rapid screening model for mild cognitive impairment is externally validated and the variables are explained. With the help of the diagnostic test design idea, the second phase of the community natural population cognitive health cohort snowballing population baseline sample enrollment assessment data is used to evaluate the prediction effect of the established model, and the differences between the prediction results and the traditional gold standard scale screening results are compared and the consistency is evaluated.

[0032] 2. Experimental Results In this experimental example, in terms of community MCI screening model establishment, effect evaluation and screening results, focusing on the problem of building a screening model suitable for mild cognitive impairment in the community, a total of 274 alternative MCI screening models were established; during the internal evaluation process within the MCI screening model establishment data set, it was found that according to the screening model optimization rules (condition one: AUC ≥ 0.8; condition two 0.75 ≤ AUC < 0.8, and sensitivity is greater than specificity), a total of 38 preferred MCI screening models entered the external validation stage; in the context of an external validation independent data set, based on the expert consultation opinion of the optimal screening model of "AUC ≥ 0.8 and sensitivity is greater than specificity", while taking into account the comprehensive effectiveness of the screening model in the model establishment data set and the model external validation data set, after expert consultation, it was finally determined that the MCI screening model described in Example 1 had the best predictive performance.

[0033] In the cross-validation evaluation of the model establishment data set, the sensitivity, specificity, accuracy and AUC of the above-mentioned optimal MCI screening model were 79.31%, 81.72%, 80.79% and 80.52%, respectively; the performance of the relevant evaluation indicators of this method in the external independent validation data set included sensitivity, specificity, accuracy and AUC of 81.13%, 64.29%, 75.31% and 84.03%, respectively; the relevant ROC curve Figure 3 and Figure 4 shown.

[0034] Through the above embodiments and experimental examples, it can be seen that the present invention has successfully constructed an MCI screening model that can accurately predict mild cognitive impairment in the middle-aged and elderly using simple indicators. On the basis of ensuring the prediction performance, the model has simple indicator collection and strong generalization ability, which is conducive to the large-scale promotion of technology and has a good application prospect.

Claims

1. A simple multimodal index rapid auxiliary screening system for mild cognitive impairment in the elderly, characterized by: include: An input module, configured to input blood indicators, physical examination data and inquiry variables of the subject; A prediction module is configured to input the blood index, physical examination data and query variables into a machine learning model to obtain a prediction result of whether the risk of mild cognitive impairment in the middle-aged and elderly is present; The output module is configured to output the prediction result.

2. The simplified multimodal index rapid auxiliary screening system for mild cognitive impairment in the elderly according to claim 1, characterized in that: The blood index is selected from at least one of the following test results: white blood cell count, neutrophil count, lymphocyte count, hemoglobin, hematocrit, mean corpuscular hemoglobin concentration, indirect bilirubin, platelet count, monocyte count, eosinophil count, and total bilirubin.

3. The simplified multimodal index rapid auxiliary screening system for mild cognitive impairment in the elderly according to claim 1, characterized in that: The physical examination data is selected from at least one of the following test results: calf circumference / neck circumference, height / calf circumference, height / neck circumference, calf circumference / mid-upper arm circumference, calf circumference, height / mid-upper arm circumference.

4. The simplified multimodal index rapid auxiliary screening system for mild cognitive impairment in the elderly according to claim 1, characterized in that: The query variable is selected from delayed recall.

5. The system for rapid auxiliary screening of mild cognitive impairment in the elderly using simplified multimodal indicators according to any one of claims 1 to 4, characterized in that: The blood index is selected from: white blood cell count, neutrophil count, lymphocyte count, hemoglobin, hematocrit, mean corpuscular hemoglobin concentration, indirect bilirubin, platelet count, monocyte count, eosinophil count, and total bilirubin; The physical examination data is selected from: calf circumference / neck circumference, height / calf circumference, height / neck circumference, calf circumference-upper arm circumference, calf circumference, height / upper arm circumference; The query variable is selected from delayed recall.

6. The simplified multimodal index rapid auxiliary screening system for mild cognitive impairment in the elderly according to claim 1, characterized in that: The algorithm of the machine learning model is selected from the extreme gradient boosting algorithm.

7. The system for rapid auxiliary screening of mild cognitive impairment in the elderly using simple multimodal indicators according to claim 1 or 6, characterized in that: The parameter settings of the machine learning model include: sampling rate = 1, learning rate = 0.001, maximum depth = 6, subsample = 0.

5.

8. A computer-readable storage medium, characterized in that: The computer program is stored thereon for implementing the system for rapid auxiliary screening of mild cognitive impairment in the elderly using the simple multimodal index as described in any one of claims 1 to 7.

Images