Fatigue assessment method and system based on medical data of PBC patient

Abstract

The invention relates to the technical field of medical health, in particular to a fatigue assessment method and system based on medical data of a PBC patient, and the method comprises the steps: collecting the medical data of the patient through a multi-modal terminal, and taking the medical data of a single examination of a single patient as a data set; screening the data set according to a preset screening condition, and storing the screened data set into a database; acquiring a fatigue-related biomarker in the serum sample; evaluating the fatigue degree according to the biomarker, and matching the evaluation result with the corresponding data set; training according to the data group and the evaluation result to obtain a fatigue degree evaluation model; and evaluating the fatigue degree of the patient according to the fatigue degree evaluation model. According to the method, a biomarker quantitative analysis result is used as an evaluation basis, so that the limitation of a subjective scale is effectively made up, more accurate basic data is provided for the training of a subsequent fatigue degree evaluation model, and the reliability of the model is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of medical and health technology, and in particular to a fatigue assessment method and system based on PBC patient medical data. Background Technology

[0002] Fatigue is the most common symptom of primary biliary cholangitis (PBC), affecting approximately [number missing] patients.

[0003] Fatigue can occur in 50% to 80% of PBC patients, and it can even further affect the patient's physical, cognitive, and emotional functions. Fatigue is also the symptom that has the greatest impact on the patient's quality of life.

[0004] Currently, there are many limitations in the clinical methods for assessing fatigue levels in PBC patients:

[0005] The assessment methods are too simplistic and lack accuracy: Traditional methods mainly rely on fatigue scales (such as the PBC-PL scale) filled out by patients subjectively. The assessment results are easily affected by subjective factors such as patients' emotions and cognitive levels, lack the support of objective physiological indicators, and have insufficient accuracy and repeatability.

[0006] Invasive testing is highly dependent: Some assessment methods rely on serum biomarker testing, but frequent blood draws increase patient suffering and medical costs, resulting in low patient compliance; for patients who undergo multiple tests, there is a lack of strategies for optimizing the use of serum samples, leading to a waste of resources.

[0007] Therefore, existing technologies have shortcomings and urgently need improvement. Summary of the Invention

[0008] The purpose of this invention is to provide a fatigue assessment method and system based on PBC patient medical data, in order to solve the problems of single assessment methods, insufficient accuracy, and strong dependence on invasive detection in the existing technology.

[0009] This invention provides a fatigue assessment method based on PBC patient medical data, comprising:

[0010] Patient medical data is collected through multimodal terminals, and the medical data of a single patient's single examination is treated as a data group.

[0011] The data groups are filtered according to preset filtering conditions, and the filtered data groups are stored in the database.

[0012] To obtain fatigue-related biomarkers from serum samples;

[0013] The degree of fatigue is assessed based on the biomarkers, and the assessment results are matched with the corresponding data sets.

[0014] A fatigue assessment model is obtained by training based on the data set and the evaluation results.

[0015] The patient's fatigue level is assessed based on the fatigue assessment model described above.

[0016] As a preferred technical solution for fatigue assessment methods based on PBC patient medical data, the medical data includes external characterization data and internal examination data, wherein:

[0017] External characterization data, including: grip strength data and muscle circumference data, or, facial expression data;

[0018] Internal examination data, including: abdominal ultrasound data, or serum liver function test data, or treatment plan data.

[0019] As a preferred technical solution for fatigue assessment methods based on PBC patient medical data, the fatigue assessment model is built by using the correspondence between the external characterization data or internal examination data and the assessment results representing the fatigue level as input data. If external characterization data or internal examination data is input, the assessment result is output.

[0020] As a preferred technical solution for fatigue assessment based on PBC patient medical data, the acquisition of fatigue-related biomarkers in serum samples includes:

[0021] In response to the fact that a single patient has multiple serum samples, only one data set corresponding to the serum sample is generated, and the fatigue level of the remaining data sets is assessed manually using the PBC fatigue scale.

[0022] As a preferred technical solution for fatigue assessment based on PBC patient medical data, the data preprocessing step is included before screening the data set: standardizing the external characterization data and converting grip strength data and muscle circumference data into standardized scores that match the patient's gender and age.

[0023] As a preferred technical solution for fatigue assessment based on PBC patient medical data, the fatigue-related biomarkers include at least one of interleukin-6, tumor necrosis factor-α, cortisol, and serotonin.

[0024] As a preferred technical solution for fatigue assessment based on PBC patient medical data, the treatment plan data includes the type, dosage and duration of medication used, and the change in fatigue level before and after medication adjustment is used as incremental features for model training.

[0025] As a preferred technical solution for fatigue assessment based on PBC patient medical data, the features extracted from the facial expression data include periorbital muscle activity, corner of mouth drooping angle, and duration of facial micro-expressions, with the rate of change of muscle activity intensity in consecutive frames used as fatigue-related features.

[0026] The present invention also provides a fatigue assessment system based on PBC patient medical data, including: a data collection module, used to collect patient medical data through a multimodal terminal, and to treat the medical data of a single patient's single examination as a data group;

[0027] The data cleaning module is used to filter data groups according to preset filtering conditions and store the filtered data groups into the database.

[0028] The serum processing module is used to obtain fatigue-related biomarkers from serum samples.

[0029] An assessment module is used to assess the degree of fatigue based on the biomarkers and match the assessment results with the corresponding data set;

[0030] The modeling module trains a fatigue assessment model based on the data set and the evaluation results.

[0031] The output module is used to assess the patient's fatigue level based on the fatigue assessment model.

[0032] Compared with existing technologies, the beneficial effects of this invention lie in its introduction of a biomarker detection mechanism from serum samples to construct a new benchmark for fatigue assessment. Traditional methods, which rely solely on manually completed PBC-PL (Physical Body Comprehension and Fatigue Scale), are easily influenced by patients' subjective feelings, emotional fluctuations, and cognitive differences, often resulting in discrepancies between the assessment results and the actual fatigue state. This invention, however, utilizes the quantitative analysis results of biomarkers (such as IL-6 and TNF-α) as the assessment basis, effectively overcoming the limitations of subjective scales and providing more accurate foundational data for the subsequent training of fatigue assessment models, thus ensuring the reliability of the model from the outset.

[0033] Furthermore, by collecting a single serum sample from only one patient, this invention significantly reduces the frequency and sample size of blood tests. This not only reduces the burden of invasive testing on patients during medical treatment and eliminates unnecessary testing steps, but also simplifies clinical procedures, thereby significantly improving patient comfort and compliance.

[0034] Furthermore, this invention categorizes medical data into two main types: external characterization data and internal examination data, and deeply explores the intrinsic correlation between these two types and fatigue levels. In clinical practice, there is a clear one-to-one correspondence between the external characterization parameters (such as grip strength and facial features), internal examination data (such as liver function indicators and imaging results), and fatigue levels of the same patient at the same time point. That is, a specific external characterization parameter necessarily corresponds to a specific fatigue level, and a specific internal examination data necessarily points to a specific fatigue level. Based on this principle, this invention achieves the function of "outputting fatigue level with a single data input": when the existing data includes either external characterization data or internal examination data, the assessment can be completed without additional testing, effectively avoiding situations where patients need to undergo re-examination due to the absence of a single parameter, and significantly improving the clinical applicability of the method.

[0035] Furthermore, this invention clarifies the construction logic and application method of the fatigue level assessment model, enabling the model to perform efficient reasoning based on existing data relationships. Assessment results can be obtained simply by inputting external representation data or internal examination data, thereby avoiding multiple examinations of patients, simplifying the assessment process, and further increasing its practicality.

[0036] Furthermore, this invention proposes a more scientific sample optimization strategy for patients with multiple serum samples: only the assessment data corresponding to one set of high-quality serum samples is retained, while the remaining sample data are assessed manually using the PBC-PL scale. This avoids the waste of medical resources caused by repeated biomarker testing, while ensuring the sufficiency of the total sample size through supplementary assessment using the scale. It reduces the testing burden on patients while ensuring the labeling quality of non-serum data through a manual review mechanism, reducing the impact of subjective bias from a single assessor on the database quality. The diverse input of high-quality training samples further improves the training accuracy and generalization ability of the subsequent fatigue assessment model.

[0037] In detail, this invention eliminates the influence of patient gender and age differences on the data by standardizing external representation data, making grip strength and muscle circumference data comparable among different patients. Standardized scores can more objectively reflect the patient's physical function status, providing higher-quality data for model training and helping to improve the accuracy of model evaluation.

[0038] Furthermore, patients' fatigue levels are also easily affected by medications. This invention incorporates treatment plan data into model training and focuses on changes in fatigue levels before and after medication adjustments, enabling the model to assess patients' fatigue in conjunction with their treatment status, thereby improving the accuracy of the model's assessment of patient fatigue at different treatment stages. Attached Figure Description

[0039] Figure 1 This is a flowchart illustrating the steps of the fatigue assessment method based on PBC patient medical data according to an embodiment of the present invention.

[0040] Figure 2 This is a structural block diagram of the fatigue assessment system based on PBC patient medical data, as described in an embodiment of the present invention. Detailed Implementation

[0041] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0042] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0043] Please see Figure 1 The diagram shows a flowchart of the fatigue assessment method based on PBC patient medical data according to an embodiment of the present invention, including:

[0044] Step S1: Collect patient medical data through a multimodal terminal and group the medical data of a single patient's single examination into a data group.

[0045] Step S2: Filter the data group according to the preset filtering conditions and store the filtered data group in the database.

[0046] Step S3: Obtain fatigue-related biomarkers from serum samples;

[0047] Step S4: Assess the level of fatigue based on biomarkers and match the assessment results with the corresponding data sets;

[0048] Step S5: Train the model based on the data set and evaluation results to obtain a fatigue assessment model;

[0049] Step S6: Assess the patient's fatigue level based on the fatigue assessment model.

[0050] Furthermore, in step S2, the preset screening conditions include:

[0051] Combined with chronic HBV, HCV, CMV, EBV infection or hepatitis or cirrhosis caused by alcohol, drugs or other reasons;

[0052] Liver failure.

[0053] Combined with primary liver cancer or liver metastases;

[0054] Combined with fatty liver or severe heart failure, or organic kidney disease;

[0055] Pregnant women.

[0056] If any of the above conditions are met, the data set will be filtered out.

[0057] In detail, during patient medical examinations, data such as grip strength, facial expression, and liver function are collected simultaneously, and all data from a single examination are integrated into a single dataset. For multiple tests on a single patient, a serum sample is randomly selected for biomarker testing, and data from the remaining tests are assessed by having the patient complete the official version of the PBC Fatigue Scale (PBC-PL), and then matched with the corresponding dataset. Using a large number of matched datasets and assessment results, a random forest algorithm is employed for model training.

[0058] Specifically, this invention accurately determines a patient's fatigue level by obtaining biomarkers from serum samples, avoiding the limitations of inaccurate evaluation results caused by simply using the official version of the PBC Fatigue Scale (PBC-PL) manually to assess fatigue level. This ensures the accuracy of subsequent training of the fatigue level assessment model. At the same time, by collecting only a single serum sample, the amount of blood sample collected is reduced, and the number of testing steps in a single medical visit is reduced, thereby improving the patient's medical experience.

[0059] Furthermore, by acquiring data divided into external characterization data and internal detection data, this invention allows those skilled in the art to accumulate the following in actual medical practice: for the same patient at the same time, the external characterization parameters, internal detection data, and fatigue level should have a one-to-one correspondence. That is, the external characterization parameters correspond to a specific fatigue level, or the internal detection parameters correspond to a specific fatigue level. In other words, the existing data can be any set of data included in either the external characterization data or the internal examination data, and the patient's fatigue level can be output. This avoids the situation where a single parameter is required for fatigue assessment but has not been tested, thus avoiding the need for the patient to be re-examined, thereby further increasing practicality.

[0060] In detail, medical data includes external characterization data and internal examination data, among which:

[0061] External characterization data, including: grip strength data and muscle circumference data, or, facial expression data;

[0062] Internal examination data, including: abdominal ultrasound data, or serum liver function test data, or treatment plan data.

[0063] In detail, the patients' grip strength and muscle circumference data were calculated as the average of three grip strength tests and muscle circumference measurements. Abdominal ultrasound data included: liver morphology, spleen length and thickness, portal vein diameter, and splenic vein diameter. Serum liver function tests included: complete blood count, albumin (Alb), globulin (Glb), total protein, total bilirubin (TBil), aspartate aminotransferase (AST), alanine aminotransferase (ALT), alkaline phosphatase (ALP), total bile acids, gamma-glutamyl transferase (GGT), cholinesterase, total cholesterol, prothrombin activity, international normalized ratio, and lactate. Treatment data did not involve intervention in patient treatment, but the study should record the patients' use of hepatoprotective and anti-fibrotic medications, including: patient laboratory and imaging results.

[0064] Furthermore, the fatigue assessment model is built by using the correspondence between external characterization data or internal examination data and assessment results representing fatigue level as input data. If any set of data is input, including grip strength data, muscle circumference, facial expression data, abdominal ultrasound examination data, liver function serum index examination data, and treatment plan data, the assessment result will be output.

[0065] In response to the existence of multiple sets of input data, each set of data is substituted into the fatigue assessment model, and the average of the output assessment results is selected as the final assessment result.

[0066] In implementation, the fatigue assessment model is a hybrid model with two input channels. Channel 1 processes external representation data (such as grip strength and facial features), using CNN to extract image features and a fully connected layer to process numerical features. Channel 2 processes internal examination data (such as ultrasound and liver function indicators), using decision trees to extract structured features. The features from the two channels are concatenated in a fusion layer (fully connected layer) to output a fatigue score (0-60 points, corresponding to the PBC-PL scale range). The specific model construction is based on existing technology and will not be elaborated here.

[0067] Furthermore, this invention clarifies the construction logic and application method of the fatigue level assessment model, enabling the model to perform efficient reasoning based on existing data relationships. Assessment results can be obtained simply by inputting external representation data or internal examination data, thereby avoiding multiple examinations of patients, simplifying the assessment process, and further increasing its practicality.

[0068] Specifically, obtaining fatigue-related biomarkers from serum samples includes:

[0069] Since a single patient may have multiple serum samples, only one data set with corresponding serum samples is generated. The fatigue levels of the remaining data sets are assessed manually using the official version of the PBC Fatigue Scale (PBC-PL).

[0070] In practice, serum sample screening: when the system detects that the same patient has ≥2 serum test records within 3 months, it automatically compares the sample collection time (preferably the most recent one) and the completeness of the test items (must include core markers such as IL-6 and TNF-α), selects the best sample group to generate the corresponding data group, and marks the remaining sample data as "non-serum group".

[0071] Non-serum group assessment: Two attending physicians independently assessed the patients using the PBC-PL scale (blind assessment, without referring to other data). If the difference between the two scores was ≤4 points, the average score was used as the label; if the difference was >4 points, the third attending physician reviewed the assessment, and the review result was taken as the standard to ensure that the label consistency was >90%.

[0072] Furthermore, this invention retains only one set of corresponding data for patients with multiple serum samples, avoiding the repetitive costs of biomarker testing. At the same time, the PBC-PL scale is used to supplement the evaluation of other data, ensuring the sample size while reducing the burden on patients. The manual assessment combined with the scale ensures the labeling quality of non-serum data, reduces the impact of a single person on the data quality in the database, and provides diverse training samples for the model, thereby further improving the accuracy of the subsequent training fatigue assessment model.

[0073] Furthermore, before screening the data set, a data preprocessing step is included: standardizing the external representation data and converting grip strength data and muscle circumference data into standardized scores that match the patient's gender and age.

[0074] In practice, this invention provides a method for standardizing external characterization data, including:

[0075] A reference range was established based on grip strength and muscle circumference data from 1,000 healthy individuals (divided by gender and age group of 5 years) (e.g., grip strength reference range for women aged 20-25 is 18-28 kg).

[0076] Calculation formula: Standardized score = (patient measured value - minimum value in the same group) / (maximum value in the same group - minimum value in the same group) × 100, with a score range of 0 to 100 (above 60 is normal, below 40 is abnormal).

[0077] In detail, this invention eliminates the influence of patient gender and age differences on the data by standardizing external representation data, making grip strength and muscle circumference data comparable among different patients. Standardized scores can more objectively reflect the patient's physical function status, providing higher-quality data for model training and helping to improve the accuracy of model evaluation.

[0078] After the data is entered into the system, it automatically matches the patient's gender and age group, calls the preset reference range to calculate the score, and replaces the original measured value with the data in the database.

[0079] Furthermore, fatigue-related biomarkers include at least one of interleukin-6 (IL-6), tumor necrosis factor-α (TNF-α), cortisol, and serotonin.

[0080] In this embodiment of the invention, serum samples were detected using electrochemiluminescence immunoassay (ECLIA), with detection limits of 0.5 pg / mL for IL-6, 1.0 pg / mL for TNF-α, 0.1 μg / dL for cortisol, and 5-hydroxytryptamine (5 ng / mL).

[0081] Furthermore, the treatment plan data includes the types of drugs used, dosages, and duration of medication, and the changes in fatigue levels before and after drug adjustments are used as incremental features for model training.

[0082] Furthermore, patients' fatigue levels are also easily affected by medications. This invention incorporates treatment plan data into model training and focuses on changes in fatigue levels before and after medication adjustments, enabling the model to assess patients' fatigue in conjunction with their treatment status, thereby improving the accuracy of the model's assessment of patient fatigue at different treatment stages.

[0083] Specifically, the features extracted from facial expression data include the activity level of the muscles around the eyes, the angle of drooping corners of the mouth, and the duration of facial micro-expressions. The rate of change in muscle activity intensity across consecutive frames is used as a fatigue-related feature. In practice, high-definition cameras are used to capture videos of patients' facial expressions at a frame rate of at least 30 frames per second. Facial landmark detection technology from computer vision algorithms is used to locate muscles in key areas such as around the eyes and corners of the mouth. The activity level of these muscles, the angle of drooping corners of the mouth, and the duration of micro-expressions in consecutive frames are calculated. By analyzing the rate of change of these parameters over time, they are quantified into specific numerical features and input into the evaluation model as fatigue-related features.

[0084] Furthermore, this invention clearly defines the features extracted from facial expression data and employs computer vision algorithms for quantitative analysis, enabling the facial expression data to be transformed into objective data that can be used for model training. By analyzing the rate of change in muscle activity intensity, the fatigue state of patients can be captured more accurately, providing another possibility for non-invasive and rapid detection of fatigue levels, thereby further improving assessment efficiency and facilitating rapid application in clinical and daily scenarios.

[0085] Further, please refer to Figure 2 As shown, it is a structural block diagram of the fatigue assessment system based on PBC patient medical data according to an embodiment of the present invention, including:

[0086] The data collection module is used to collect patient medical data through a multimodal terminal, and to group the medical data of a single patient's single examination into a data set.

[0087] The data cleaning module is used to filter data groups according to preset filtering conditions and store the filtered data groups into the database.

[0088] The serum processing module is used to obtain fatigue-related biomarkers from serum samples.

[0089] The assessment module is used to evaluate fatigue levels based on biomarkers and match the assessment results with the corresponding datasets.

[0090] The modeling module trains the model based on the data set and evaluation results to obtain a fatigue assessment model.

[0091] The output module is used to assess the patient's fatigue level based on the fatigue assessment model.

[0092] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A fatigue assessment method based on PBC patient medical data, characterized in that, include: Patient medical data is collected through multimodal terminals, and the medical data of a single patient's single examination is treated as a data group. The data groups are filtered according to preset filtering conditions, and the filtered data groups are stored in the database. To obtain fatigue-related biomarkers from serum samples; The degree of fatigue is assessed based on the biomarkers, and the assessment results are matched with the corresponding data sets. A fatigue assessment model is obtained by training based on the data set and the evaluation results. The patient's fatigue level is assessed based on the fatigue assessment model described above.

2. The fatigue assessment method based on PBC patient medical data according to claim 1, characterized in that, The medical data includes external characterization data and internal examination data, wherein: External characterization data, including: grip strength data and muscle circumference data, or, facial expression data; Internal examination data, including: abdominal ultrasound data, or serum liver function test data, or treatment plan data.

3. The fatigue assessment method based on PBC patient medical data according to claim 2, characterized in that, The fatigue assessment model is built using the correspondence between the external characterization data or internal inspection data and the assessment results representing the fatigue level as input data. If external characterization data or internal inspection data is input, the assessment result is output.

4. The fatigue assessment method based on PBC patient medical data according to claim 1, characterized in that, The acquisition of fatigue-related biomarkers from serum samples includes: In response to the fact that a single patient has multiple serum samples, only one data set corresponding to the serum sample is generated, and the fatigue level of the remaining data sets is assessed manually using the PBC fatigue scale.

5. The fatigue assessment method based on PBC patient medical data according to claim 4, characterized in that, Before screening the data set, a data preprocessing step is also included: standardizing the external representation data and converting grip strength data and muscle circumference data into standardized scores that match the patient's gender and age.

6. The fatigue assessment method based on PBC patient medical data according to claim 5, characterized in that, The fatigue-related biomarkers include at least one of interleukin-6, tumor necrosis factor-α, cortisol, and serotonin.

7. The fatigue assessment method based on PBC patient medical data according to claim 2, characterized in that, The treatment plan data includes the types of drugs used, dosages, and duration of medication, and the changes in fatigue levels before and after drug adjustments are used as incremental features for model training.

8. The fatigue assessment method based on PBC patient medical data according to claim 2, characterized in that, The features extracted from the facial expression data include the degree of activity of the muscles around the eyes, the angle of drooping corners of the mouth, and the duration of facial micro-expressions. The rate of change of muscle activity intensity in consecutive frames is used as a fatigue-related feature.

9. A fatigue assessment system based on PBC patient medical data, used to implement the fatigue assessment method based on PBC patient medical data as described in any one of claims 1 to 8, characterized in that, include: The data collection module is used to collect patient medical data through a multimodal terminal, and to group the medical data of a single patient's single examination into a data set. The data cleaning module is used to filter data groups according to preset filtering conditions and store the filtered data groups into the database. The serum processing module is used to obtain fatigue-related biomarkers from serum samples. An assessment module is used to assess the degree of fatigue based on the biomarkers and match the assessment results with the corresponding data set; The modeling module trains a fatigue assessment model based on the data set and the evaluation results. The output module is used to assess the patient's fatigue level based on the fatigue assessment model.

Images