System for providing health information by measuring finger morphology and application method thereof

Abstract

The invention provides a system for providing health information by measuring finger morphology and an application method thereof, which are used for detecting finger morphology and comprise a system framework and a mathematical model, the system framework comprises a data input module of a finger tip and a core function module; the data input module comprises three-dimensional form data (such as CT or MRI scanning), two-dimensional images (such as X-rays and photos) and basic information (such as age and gender) of a patient; the core function module comprises an image preprocessing module, a feature extraction module and a model analysis module; the mathematical model includes a nail corner model, a distal knuckle ratio and a volume measurement model. According to the invention, system analysis is utilized to realize automatic operation and reduce human errors, and an automatic data acquisition and processing system is established to replace traditional manual measurement and recording; result deviation caused by different skill levels of detectors is reduced by using a standardized process; the method supports large-scale sample analysis and is suitable for large-scale screening. Through system analysis, the detection time can be shortened to 1 / 5 of that of a traditional method, and only several seconds are needed for single measurement. Through automatic analysis, more than 90% of manual intervention is reduced, and the operation complexity is greatly reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of finger clubbing prediction and diagnosis, and in particular to a health information system and an application method thereof that provides health information by measuring finger morphology. Background Art

[0002] Finger morphology is associated with a variety of gene regulation, mutations, and prediction of disease progression. For example, the finger length ratio, which is the ratio of the length of the index finger to the ring finger (2D:4D), is related to the Hox gene.

[0003] Hox genes are a family of highly conserved transcription factors that are common in primates and some other mammals. These genes control the differentiation of the urogenital system and limb skeletal system of vertebrates. Hox genes (especially HOXA and HOXD) determine the length ratio of individual fingers and toes during embryogenesis and no longer change during the growth of the individual. This control of Hox genes on limb bones is affected by sperm production and prenatal hormone concentrations.

[0004] For example, the formation of clubbing is closely related to ABCA8, HPGD and SLCO2A1 gene mutations, and the molecular mechanisms involved may include abnormal prostaglandin metabolism and the release of angiogenic factors. The HPG axis (hypothalamus-pituitary-gonadal axis) and the HPA axis (hypothalamus-pituitary-adrenal axis) are two important regulatory axes in the human endocrine system, and there is interaction and influence between them. Prenatal hormones and prostaglandins in the HPG axis interact with each other to jointly regulate the release and synthesis of reproductive hormones, and affect reproduction, sexual development and many other physiological functions. The HPA axis is mainly responsible for coping with stress and stress response. Gonadal hormones in the HPG axis play a key role in the establishment, activation and regulation of the HPA axis. HPG axis imbalance may affect mental health and stress response disorders, thereby showing symptoms of neuropsychiatric diseases. Monitoring of finger morphology can help to detect HPA and HPG axis disorders early and provide a new perspective for clinical diagnosis and treatment.

[0005] With the improvement of living standards, simply measuring height and weight can no longer meet the needs of monitoring physical health functions in health examinations and measurements. There is a consensus in existing research that the finger length ratio can indicate the level of sex hormones in the human body. The common measurement method is to manually use a vernier caliper to record the length of each finger ( Figure 1), and then obtain the corresponding data through calculation. Clubbing is a common sign of certain diseases, or an early manifestation of some serious diseases, such as bronchiectasis, pulmonary fibrosis, tuberculosis and other respiratory diseases, as well as congenital heart disease. The measurement of clubbing can be obtained by nail angle measurement or finger thickness ratio measurement. Although there are operating standards, manual measurement is highly subjective, the system error is large, errors may occur in the manual recording process, and the steps between importing calculations and measurements are cumbersome, resulting in poor applicability of this indicator, which should have been widely used and had a suggestive effect on sex hormone levels, and the effect of technical promotion is not ideal. If the measurement steps can be standardized through unified measurement tools and unified algorithms, the measurement efficiency will be greatly improved, and a technical basis will be provided for the widespread application of finger morphology measurement to indicate health. Summary of the invention

[0006] The purpose of the present invention is to provide a health information system and its application method by measuring finger morphology, which can reduce human errors, improve diagnostic efficiency, provide quantitative analysis, assist clinical decision-making and improve accuracy through automated operation.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] An application method for providing a health information system by measuring finger morphology comprises the following steps:

[0009] Step S1: collecting data from the user's fingertip through a data input module;

[0010] Step S2: image processing and feature extraction;

[0011] Step S21: performing image segmentation through an image preprocessing module, a feature extraction module and a model analysis module;

[0012] Image segmentation: Based on the convolutional neural network (CNN) segmentation model, key areas including nails, nail beds, and fingertips are segmented and edge detection is performed;

[0013] Step S22: feature extraction: construct the three-dimensional shape of the fingertip using the point cloud data, and extract geometric features and texture features;

[0014] Step S3: classification and prediction;

[0015] Step S31: Classification model: Use the labeled data to train the classification model to determine whether it is clubbing and the classification level; use the classification model for initial screening, and the regression model for detailed disease course assessment;

[0016] Step S32: Regression model: used to predict continuous variables and study the relationship between features and target values; the regression model is used to predict the clubbing score, and then the patients are classified based on the score threshold;

[0017] Step S4: Deep learning: Build a deep convolutional network based on ResNet or EfficientNet to automatically learn features from images;

[0018] Step S5: Information display

[0019] Step S51: Presenting a three-dimensional model and feature comparison chart through a visual structure allows doctors to intuitively display and analyze the morphological changes of the fingertips, as well as display changes in key indicators, including the angle of the finger and the volume of the fingertips.

[0020] Preferably, extracting geometric features in step S22 includes angle measurement, calculation of curvature radius and thickness ratio;

[0021] The angle measurement: using a geometric formula to measure the angle between the nail and the finger base;

[0022]

[0023] Among them, (x1, y1) and (x2, y2) are the coordinates of two points on the nail surface;

[0024] The radius of curvature (R): The curvature of the nail is obtained by using a curve fitting method:

[0025]

[0026] Where, f(x) is the fitted nail contour curve;

[0027] The thickness ratio (T): The thickness ratio of the finger tip is defined as:

[0028]

[0029] Preferably, the texture features extracted in step S22 include grayscale features, morphological features, directional features and spatial frequency features;

[0030] The grayscale features include average grayscale value, grayscale variance and grayscale co-occurrence matrix, which are used to reflect the overall intensity of blood vessel distribution and describe the inhomogeneity of grayscale distribution, respectively. Proliferating blood vessels usually lead to higher variance and reflect the difference in texture and blood vessel indicators in the image.

[0031] The morphological characteristics include vascular density, vascular diameter, number of branch points and fractal dimension, which are used to evaluate and count the proliferation and complexity indicators of blood vessels;

[0032] The directional features include directional gradient histogram and Gabor filter, which are used to extract local directional change features and detect blood vessel texture of cloth direction and frequency;

[0033] The spatial frequency characteristics include Fourier transform and wavelet transform, which are used to analyze the periodicity and spatial frequency characteristics of blood vessel texture and multi-scale feature analysis to capture blood vessel proliferation patterns at different scales.

[0034] Preferably, the classification model in step S31 includes auxiliary diagnosis, disease screening and clubbing classification;

[0035] The auxiliary diagnosis includes determining whether the patient has clubbing; identifying the cause of clubbing, including chronic lung disease or heart disease;

[0036] The disease screening utilizes the extracted nail features to automatically classify patients into “normal” and “high-risk” groups; rapidly screening potential cases in a large population;

[0037] The clubbing classification includes classification by severity, including mild, moderate and severe; and combining other features including blood oxygen levels to classify clubbing in different disease contexts.

[0038] Preferably, the regression model in step S32 includes quantifying characteristic changes, establishing a relationship model of physiological parameters and evaluating disease progression;

[0039] The quantitative feature changes include predicting the severity of clubbing and scoring based on feature extraction; analyzing and studying the contribution of nail curvature, interdigital angle, thickness ratio and other features to the degree of clubbing;

[0040] The establishment of the relationship model of physiological parameters includes analyzing the association between clubbing characteristics and continuous medical indicators such as oxygen saturation and cardiopulmonary function; constructing a multi-factor linear or nonlinear model to explain how different biological characteristics jointly affect the formation of clubbing;

[0041] The disease progression assessment predicts the development trend or change rate of clubbing based on patient follow-up data.

[0042] Preferably, to determine whether a patient has clubbing, the formula of the logistic regression classification model is as follows:

[0043] P (positive clubbing) = σ (w0 + w1·AIA + w2·CR + w3·TR + w4·(100-SpO2) + w5·PH);

[0045] σ(x): Sigmoid function, defined as:

[0046]

[0047] Its output is between 0 and 1, indicating the probability of being classified as positive;

[0048] P(positive clubbing): the probability of predicting positive clubbing;

[0049] Among them, AIA stands for interphalangeal angle, which is defined as the angle between the nail and the knuckle, in degrees (°);

[0050] CR stands for curvature radius, which is defined as the radius of curvature of the fingertip in millimeters (mm). In patients with clubbing, the fingertips are more rounded and the radius of curvature is smaller.

[0051] TR stands for thickness ratio, which is defined as the ratio of the thickness at the finger tip to the thickness at the base of the finger;

[0052] SpO2 stands for blood oxygen saturation, definition: the degree of hypoxia in blood oxygen saturation (unit: %);

[0053] PH stands for medical history score, which is defined as a score that quantifies medical history-related characteristics, including smoking index or chronic disease score;

[0054] Model parameters (w i )

[0055] w0: intercept of the model (bias term);

[0056] When all feature variables are 0, the initial value of the model output;

[0057] w1,w2,...,w n : The weight corresponding to each feature;

[0058] Indicates the degree of influence of each feature on the probability of positive clubbing;

[0059] w i >0: The increase of eigenvalue will increase the probability of positive clubbing;

[0060] w i <0: an increase in the eigenvalue will reduce the probability of positive clubbing;

[0061] The absolute value of the weight indicates the strength of the influence;

[0062] Based on multiple linear regression, predict the clubbing score for a patient:

[0063] S=β0+β1·AIA+β2·CR+β3·TR+β4·(100-SpO2)+β5·PH+∈

[0064] S: Clubbing score (predicted target).

[0065] β0, β1, ..., β5: model parameters (fitted from training data).

[0066] AIA: Interdigital angle (unit: degree).

[0067] CR: radius of curvature (unit: mm).

[0068] TR: Thickness ratio (the ratio of fingertip thickness to finger base thickness, unitless).

[0069] SpO2: blood oxygen saturation (unit: %).

[0070] PH: Patient-related characteristics, quantified by numerical values ​​(such as long-term smoking index or specific patient score).

[0071] ∈: Random error term. .

[0072] Preferably, the method further includes step S6: model optimization and performance index evaluation, which specifically includes the following steps:

[0073] Step S61: Model optimization: by expanding the sample size, improving the generalization ability of the model, and enhancing the data; using grid search or Bayesian optimization to select the best model parameters to tune the hyperparameters;

[0074] Step S62: Performance index evaluation: verify the effect through sensitivity, specificity, and ROC curve evaluation algorithm; or ensure the reliability of the automation system through measurement error analysis.

[0075] The present invention also provides a system for providing health information by measuring finger morphology, which is used to detect finger morphology, including a system framework and a mathematical model;

[0076] The system framework includes a data input module and a core function module at the fingertips;

[0077] The data input module includes three-dimensional morphological data (such as CT or MRI scans), two-dimensional images (such as X-rays, photos), and basic patient information (such as age, gender).

[0078] The core functional modules include an image preprocessing module, a feature extraction module and a model analysis module;

[0079] The mathematical model includes a nail angle model, a distal phalanx ratio, and a volume measurement model;

[0080] Preferably, the image preprocessing module includes denoising, normalization and contrast enhancement processing on the data;

[0081] The feature extraction module includes extracting anatomical features related to clubbing, including nail angle and distal phalanx thickness.

[0082] The model analysis module includes analyzing the extracted features using a mathematical model or a machine learning method to determine whether clubbing exists.

[0083] Preferably, the nail angle model includes establishing an angle describing the nail and the nail bed, wherein the normal angle is preset to be less than 160°, and when the angle is greater than 160°, it is judged as clubbing;

[0084] The distal phalanx ratio includes calculating the width ratio of the distal phalanx to the middle phalanx. It is preset that the distal phalanx is smaller than the middle phalanx. When the distal phalanx is larger than the middle phalanx, it is determined to be clubbing.

[0085] The volume measurement model includes obtaining the volume change of the fingertip through three-dimensional modeling.

[0086] Compared with the prior art, the present invention has the following beneficial effects:

[0087] (1) The present invention utilizes system analysis to realize automated operation and reduce human errors. It replaces traditional manual measurement and recording by establishing an automated data collection and processing system. It utilizes standardized processes to reduce result deviations caused by different skill levels of testers. The present invention supports large-scale sample analysis and is suitable for large-scale screening.

[0088] (2) Through system analysis, the detection time can be shortened to 1 / 5 of that of traditional methods, and a single measurement takes only a few seconds. Automated analysis reduces more than 90% of manual intervention and greatly reduces operational complexity.

[0089] (3) The present invention uses algorithms to automatically extract key parameters (such as interdigital angle, curvature radius, and thickness ratio), avoiding errors caused by human subjective judgment. Mathematical models and statistical algorithms are used to ensure the scientificity and consistency of the measurement results. Image processing algorithms can accurately locate key points (such as the base and end of the nail), and the calculation results are more accurate.

[0090] The measurement accuracy is improved to sub-millimeter level, and the error of key parameters (such as curvature and angle) is controlled within ±1°. In the detection of subtle symptoms in the early stage of the disease, the accuracy is improved by about 30%, reducing the missed diagnosis rate.

[0091] (4) The present invention combines machine learning classification algorithms to quickly determine the presence and severity of clubbing. Multidimensional feature extraction and modeling enable the system to simultaneously consider multiple pathogenic factors (such as lung function, blood oxygen level, etc.), thereby achieving a more comprehensive diagnosis.

[0092] The average diagnosis time was shortened from several minutes to seconds. The classification model accuracy exceeded 95%, and the misdiagnosis rate was reduced by about 40%.

[0093] (5) The present invention can provide quantitative analysis to assist clinical decision-making. The system generates detailed quantitative reports, including key parameter values, scoring results, classification labels, etc. It can track changes in the course of the disease and help doctors evaluate the efficacy or adjust treatment plans. It can provide accurate data support for disease research (such as the correlation analysis between the morphology of clubbing and specific diseases).

[0094] (6) The system of the present invention reduces the cost of repeated examinations caused by human errors or missed diagnoses. By improving screening efficiency, diseases can be detected earlier, thereby reducing subsequent treatment costs. The unit test cost is reduced by more than 50%, and by detecting diseases early, the patient's long-term treatment costs are reduced by about 20%.

[0095] (7) The present invention transforms the complex clubbing detection problem into a quantitative and computable problem through an algorithm, setting a precedent for the combination of medical imaging and mathematical models.

[0096] Improve utilization of medical resources, especially in resource-limited areas, through low-cost solutions for early diagnosis.

[0097] It can reduce the missed diagnosis rate of clubbing-related diseases, help improve patients' survival rate and quality of life. At the same time, it can improve the diagnostic efficiency of the overall medical system and relieve the workload of doctors. BRIEF DESCRIPTION OF THE DRAWINGS

[0098] Figure 1 A schematic diagram of the workflow of an application method of a health information system for providing health information by measuring finger morphology provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0099] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0100] This embodiment provides a system for providing health information by measuring finger morphology, which is used to detect finger morphology, including a system framework and a mathematical model;

[0101] The system framework includes the fingertip data input module and the core function module;

[0102] The data input module includes three-dimensional morphological data (such as CT or MRI scans), two-dimensional images (such as X-rays, photos), and basic patient information (such as age, gender)

[0103] The core functional modules include image preprocessing module, feature extraction module and model analysis module; the image preprocessing module includes denoising, normalization and contrast enhancement of data. The feature extraction module includes extracting anatomical features related to clubbing, including nail angle and distal phalanx thickness. The model analysis module includes using mathematical models or machine learning methods to analyze the extracted features and determine whether clubbing exists.

[0104] The mathematical model includes a nail angle model, a distal phalanx ratio and a volume measurement model; the nail angle model includes establishing a description of the angle between the nail and the nail bed, with the normal angle preset to be below 160°, and when the angle is greater than 160°, it is judged as clubbing; the distal phalanx ratio includes calculating the ratio of the width of the distal phalanx to the middle phalanx, with the normal preset being that the distal phalanx is smaller than the middle phalanx, and when the distal phalanx is larger than the middle phalanx, it is judged as clubbing.

[0105] The volume measurement model includes obtaining the volume changes of the fingertips through three-dimensional modeling.

[0106] like Figure 1 As shown, an application method of providing a health information system by measuring finger morphology provided by this embodiment includes the following steps:

[0107] Step S1: collecting data from the user's fingertip through a data input module;

[0108] Step S2: image processing and feature extraction;

[0109] Step S21: performing image segmentation through an image preprocessing module, a feature extraction module and a model analysis module;

[0110] Image segmentation: Convolutional neural network (CNN)-based segmentation models, such as U-Net, are used to segment key areas including nails, nail beds, and fingertips, and perform edge detection: Canny or Sobel algorithms are used to detect nail corners and tissue boundaries.

[0111] Step S22: Feature extraction: Use point cloud data to construct the three-dimensional shape of the fingertip, and extract geometric features including angle, thickness and texture features including image features of vascular hyperplasia.

[0112] Extracting geometric features includes angle measurement, calculating curvature radius and thickness ratio;

[0113] Angle measurement: Use geometric formulas to measure the angle between the nail and the base of the finger;

[0114]

[0115] Among them, (x1, y1) and (x2, y2) are the coordinates of two points on the nail surface;

[0116] Curvature radius (R): The curvature of the nail is obtained using the curve fitting method:

[0117]

[0118] Where, f(x) is the fitted nail contour curve;

[0119] Thickness ratio (T): The thickness ratio of the finger tip is defined as:

[0120]

[0121] This embodiment uses an algorithm to automatically extract key parameters (such as interdigital angle, curvature radius, and thickness ratio), avoiding errors caused by human subjective judgment. Mathematical models and statistical algorithms are used to ensure the scientificity and consistency of the measurement results. Image processing algorithms can accurately locate key points (such as the base and end of the nail), and the calculation results are more accurate.

[0122] The measurement accuracy is improved to sub-millimeter level, and the error of key parameters (such as curvature and angle) is controlled within ±1°. In the detection of subtle symptoms in the early stage of the disease, the accuracy is improved by about 30%, reducing the missed diagnosis rate.

[0123] The extracted texture features include grayscale features, morphological features, directional features and spatial frequency features;

[0124] Grayscale features include mean grayscale value, grayscale variance and grayscale co-occurrence matrix, which are used to reflect the overall intensity of blood vessel distribution and describe the inhomogeneity of grayscale distribution, respectively. Proliferating blood vessels usually lead to higher variance and reflect the difference in texture and blood vessel indicators in the image.

[0125] Morphological characteristics include vascular density, vascular diameter, number of branch points, and fractal dimension, which are used to evaluate and count the proliferation and complexity of blood vessels;

[0126] Directional features include oriented gradient histogram and Gabor filter, which are used to extract local directional change features and detect vascular texture of cloth direction and frequency;

[0127] Spatial frequency features include Fourier transform and wavelet transform, which are used to analyze the periodicity and spatial frequency characteristics of vascular texture and multi-scale feature analysis to capture vascular proliferation patterns at different scales.

[0128] Step S3: classification and prediction;

[0129] Step S31: Classification model: Use the labeled data to train a classification model (such as support vector machine SVM, random forest, etc.) to determine whether it is clubbing and the classification level; use the classification model for initial screening, and the regression model for detailed disease course assessment;

[0130] Classification model, including auxiliary diagnosis, disease screening and clubbing classification. Auxiliary diagnosis includes determining whether the patient has clubbing; distinguishing the cause of clubbing, including chronic lung disease or heart disease;

[0131] Disease screening uses extracted nail features to automatically classify patients into "normal" and "high-risk" groups; quickly screen potential cases in large populations;

[0132] The classification of finger clubbing includes grading by severity, including mild, moderate and severe; and combining other characteristics, including blood oxygen levels, to classify finger clubbing in different disease contexts.

[0133] Furthermore, in this auxiliary diagnosis, to determine whether a patient has clubbing, the formula of the logistic regression classification model is as follows:

[0134] P (positive clubbing) = σ (w0 + w1·AIA + w2·CR + w3·TR + w4·(100-SpO2) + w5·PH);

[0136] σ(x): Sigmoid function, defined as:

[0137] Its output is between 0 and 1, indicating the probability of being classified as positive;

[0138] P(positive clubbing): the probability of predicting positive clubbing;

[0139] in:

[0140] AIA (Interphalangeal Angle)

[0141] Definition: The angle between the nail and the knuckle, in degrees (°)

[0142] The interdigital angle in patients with clubbing is usually less than 160°. AIA values ​​greater than 160° or higher indicate a higher risk of clubbing.

[0143] CR (radius of curvature)

[0144] Definition: The radius of curvature of the fingertip, in millimeters (mm).

[0145] Patients with clubbing have rounder fingertips with a smaller radius of curvature. A lower CR value may indicate features of clubbing.

[0146] TR (Thickness Ratio)

[0147] Definition: The ratio of fingertip thickness to finger base thickness, unitless.

[0148] Patients with clubbing have significantly increased fingertip thickness. Higher TRTRTR values ​​indicate thicker fingertips, which may be a characteristic of clubbing.

[0149] 100-SpO2 (blood oxygen saturation)

[0150] Definition: The degree of hypoxia of blood oxygen saturation (unit: %).

[0151] Chronic hypoxia is an important cause of clubbing. A higher 100-SpO2 value indicates a lower blood oxygen level and a higher risk of clubbing.

[0152] PH (medical history score)

[0153] Definition: A score that quantifies characteristics related to medical history (such as a smoking index or chronic disease score).

[0154] Certain medical histories, such as long-term smoking or chronic obstructive pulmonary disease, are strongly associated with clubbing. Higher medical history scores generally increase the risk of clubbing.

[0155] Model parameters (w i )

[0156] w0: intercept (bias term) of the model.

[0157] The initial value of the model output when all feature variables are 0.

[0158] w1,w2,...,w n : The weight corresponding to each feature.

[0159] Indicates the degree of influence of each feature on the probability of positive clubbing.

[0160] w i >0: An increase in the eigenvalue will increase the probability of positive clubbing.

[0161] w i <0: An increase in the eigenvalue will reduce the probability of positive clubbing.

[0162] The absolute value of the weight indicates the strength of the influence.

[0163] In this embodiment, a machine learning classification algorithm is combined to quickly determine the presence and severity of clubbing. Multidimensional feature extraction and modeling enable the system to simultaneously consider multiple pathogenic factors (such as lung function, blood oxygen level, etc.), thereby achieving a more comprehensive diagnosis.

[0164] The average diagnosis time was shortened from several minutes to seconds. The classification model accuracy exceeded 95%, and the misdiagnosis rate was reduced by about 40%.

[0165] Step S32: Regression model: used to predict continuous variables and study the relationship between features and target values; the regression model is used to predict the clubbing score, and then the patients are classified based on the score threshold.

[0166] Regression model, including quantitative feature changes, establishing relationship models of physiological parameters and disease progression assessment. Quantitative feature changes include feature extraction to predict the severity of clubbing and score; analyze and study the contribution of nail curvature, interdigital angle, thickness ratio and other features to the degree of clubbing;

[0167] Establishing relationship models of physiological parameters includes analyzing the association between clubbing characteristics and continuous medical indicators such as oxygen saturation and cardiopulmonary function; constructing multi-factor linear or nonlinear models to explain how different biological characteristics jointly affect clubbing;

[0168] Disease progression assessment predicts the development trend or rate of change of clubbing based on patient follow-up data.

[0169] Based on multiple linear regression, predict the clubbing score for a patient:

[0170] S=β0+β1·AIA+β2·CR+β3·TR+β4·(100-Sp(O2)+β5·PH∈

[0171] S: Clubbing score (predicted target).

[0172] β0, 1, ..., β5: model parameters (fitted by training data).

[0173] AIA: Interdigital angle (unit: degree).

[0174] CR: radius of curvature (unit: mm).

[0175] TR: Thickness ratio (the ratio of fingertip thickness to finger base thickness, unitless).

[0176] SpO2: blood oxygen saturation (unit: %).

[0177] PH: Characteristics related to the patient's medical history, quantified by numerical values ​​(such as long-term smoking index or specific patient score).

[0178] ∈: Random error term.

[0179] Suppose a patient is suspected of having a disease associated with clubbing, such as chronic obstructive pulmonary disease (COPD) or lung cancer. Through physical examination and imaging data, the following feature data are obtained:

[0180] Angle of Interphalangeal Arc (AIA): 170°

[0181] Curvature Radius (CR): 4.5mm

[0182] Thickness Ratio (TR): 0.85

[0183] Blood oxygen saturation (SpO2): 92%

[0184] Patient History (PH): Long-term smoking (index 30).

[0185] The clubbing score (0-10, higher scores indicate more severe clubbing) was predicted by regression model.

[0186] The classification model is used to determine whether it is "positive clubbing".

[0187] Assume that the training data is fitted with the following parameters:

[0188] Assume that the training data is fitted with the following parameters:

[0189] β0 = -5.0 (intercept)

[0190] β1=0.1(influence of inter-finger angle)

[0191] β2 = 0.5 (influence of curvature radius)

[0192] β3=8.0(Influence of thickness ratio)

[0193] β4 = 0.3 (influenced by blood oxygen level)

[0194] β5 = 0.02 (influence of medical history)

[0195] Use the regression model to predict the clubbing score and substitute the patient data:

[0196] S=-5.0+0.1*170+0.5*4.5+8.0*0.85+0.3*(100-92)+0.02*30;

[0197] Calculation: S = -5.0 + 17.0 + 2.25 + 6.8 + 2.4 + 0.6 = 24.05.

[0198] Results: The predicted score of the patient's clubbing was 24.05. Assuming the score threshold was 10, the value was significantly higher, indicating that the patient might have severe clubbing.

[0199] The classification model is used to judge the positive clubbing of the finger. Assume that the classification model is fitted to obtain the following weights:

[0200] w0 = -20.0 (bias)

[0201] w1=0.15

[0202] w2=0.8

[0203] w3=12.0

[0204] w4 = 0.5

[0205] w5 = 0.05

[0206] Substitute patient data:

[0207] x=-20.0+0.15*170+0.8*4.5+12.0*0.85+0.5*(100-92)+0.05*30;

[0208] Calculation: x = -20.0 + 25.5 + 3.6 + 10.2 + 4.0 + 1.5 = 24.8.

[0209] Classification probability:

[0210]

[0211] result

[0212] The classification model judged that the probability was close to 1 (100%), that is, the patient was likely to be positive for clubbing.

[0213] Step S4: Deep learning: Build a deep convolutional network based on ResNet or EfficientNet to automatically learn features from images.

[0214] Step S5: Information display

[0215] Step S51: Presenting a three-dimensional model and feature comparison chart through a visual structure allows doctors to intuitively display and analyze the morphological changes of the fingertips, as well as display changes in key indicators, including the angle of the finger and the volume of the fingertips.

[0216] Step S6: Model optimization and performance index evaluation, specifically including the following steps:

[0217] Step S61: Model optimization: by expanding the sample size, improving the generalization ability of the model, and enhancing the data; using grid search or Bayesian optimization to select the best model parameters to tune the hyperparameters;

[0218] Step S62: Performance index evaluation: verify the effect through sensitivity, specificity, and ROC curve evaluation algorithm; or ensure the reliability of the automation system through measurement error analysis.

[0219] This embodiment can provide quantitative analysis to assist clinical decision-making. The system generates detailed quantitative reports, including key parameter values, scoring results, classification labels, etc. It can track changes in the course of the disease and help doctors evaluate the efficacy or adjust treatment plans. It can provide accurate data support for disease research (such as the association analysis between the morphology of clubbing and specific diseases).

[0220] This embodiment uses system analysis to achieve automated operation and reduce human errors. It replaces traditional manual measurement and recording by establishing an automated data collection and processing system. It uses standardized processes to reduce result deviations caused by different skill levels of testers. The present invention supports large-scale sample analysis and is suitable for large-scale screening. Through system analysis, the detection time can be shortened to 1 / 5 of the traditional method, and a single measurement takes only a few seconds. Automated analysis reduces more than 90% of manual intervention and greatly reduces the complexity of operation.

[0221] The embodiment reduces the cost of repeated examinations caused by manual errors or missed diagnoses. By improving screening efficiency, diseases can be detected earlier, thereby reducing subsequent treatment costs. The unit detection cost is reduced by more than 50%, and by detecting the disease early, the patient's long-term treatment costs are reduced by about 20%. Through algorithms, the complex clubbing detection problem is transformed into a quantitative and computable problem, creating a precedent for the combination of medical imaging and mathematical models.

[0222] Improve the utilization of medical resources, especially in resource-limited areas, and make early diagnosis through low-cost solutions. It can reduce the missed diagnosis rate of clubbing-related diseases, help improve the survival rate and quality of life of patients. At the same time, it can improve the diagnostic efficiency of the overall medical system and relieve the workload of doctors.

[0223] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. An application method for providing a health information system by measuring finger morphology, characterized in that: The steps include: Step S1: collecting data from the user's fingertip through a data input module; Step S2: image processing and feature extraction; Step S21: performing image segmentation through an image preprocessing module, a feature extraction module and a model analysis module; Image segmentation: Based on the convolutional neural network segmentation model, key areas including nails, nail beds, and fingertips are segmented and edge detection is performed; Step S22: feature extraction: construct the three-dimensional shape of the fingertip using the point cloud data, and extract geometric features and texture features; Step S3: classification and prediction; Step S31: Classification model: Use the labeled data to train the classification model to determine whether it is clubbing and the classification level; use the classification model for initial screening, and the regression model for detailed disease course assessment; Step S32: Regression model: used to predict continuous variables and study the relationship between features and target values; the regression model is used to predict the clubbing score, and then the patients are classified based on the score threshold; Step S4: Deep learning: Build a deep convolutional network based on ResNet or EfficientNet to automatically learn features from images; Step S5: Information display Step S51: Presenting a three-dimensional model and feature comparison chart through a visual structure allows doctors to intuitively display and analyze the morphological changes of the fingertips, as well as display changes in key indicators, including the angle of the finger and the volume of the fingertips.

2. A system for providing health information by measuring finger morphology according to claim 1, characterized in that: Extracting geometric features in step S22 includes angle measurement, calculation of curvature radius and thickness ratio; The angle measurement: using a geometric formula to measure the angle between the nail and the finger base; Among them, (x1, y1) and (x2, y2) are the coordinates of two points on the nail surface; The radius of curvature (R): The curvature of the nail is obtained by using a curve fitting method: Where, f(x) is the fitted nail contour curve; The thickness ratio (T): The thickness ratio of the finger tip is defined as:

3. The system for providing health information by measuring finger morphology according to claim 1, characterized in that: The texture features extracted in step S22 include grayscale features, morphological features, directional features and spatial frequency features; The grayscale features include average grayscale value, grayscale variance and grayscale co-occurrence matrix, which are used to reflect the overall intensity of blood vessel distribution and describe the inhomogeneity of grayscale distribution, respectively. Proliferating blood vessels usually lead to higher variance and reflect the difference in texture and blood vessel indicators in the image. The morphological characteristics include vascular density, vascular diameter, number of branch points and fractal dimension, which are used to evaluate and count the proliferation and complexity indicators of blood vessels; The directional features include directional gradient histogram and Gabor filter, which are used to extract local directional change features and detect blood vessel texture of cloth direction and frequency; The spatial frequency characteristics include Fourier transform and wavelet transform, which are used to analyze the periodicity and spatial frequency characteristics of blood vessel texture and multi-scale feature analysis to capture blood vessel proliferation patterns at different scales.

4. The system for providing health information by measuring finger morphology according to claim 1, characterized in that: The classification model in step S31 includes auxiliary diagnosis, disease screening and clubbing classification; The auxiliary diagnosis includes determining whether the patient has clubbing; identifying the cause of clubbing, including chronic lung disease or heart disease; The disease screening utilizes the extracted nail features to automatically classify patients into "normal" and "high-risk" groups; it can quickly screen potential cases in a large population; The clubbing classification includes classification by severity, including mild, moderate and severe; and combining other features including blood oxygen levels to classify clubbing in different disease contexts.

5. A system for providing health information by measuring finger morphology according to claim 4, characterized in that: The step S32 regression model includes quantifying characteristic changes, establishing a relationship model of physiological parameters and evaluating the course of disease progression; The quantitative feature changes include predicting the severity of clubbing and scoring based on feature extraction; analyzing and studying the contribution of nail curvature, interdigital angle, and thickness ratio features to the degree of clubbing; The establishment of the relationship model of physiological parameters includes analyzing the association between the clubbing characteristics and the continuous medical indicators of oxygen saturation and cardiopulmonary function; constructing a multi-factor linear or nonlinear model to explain how different biological characteristics jointly affect the formation of clubbing; The disease progression assessment predicts the development trend or change rate of clubbing based on patient follow-up data.

6. The application system for providing health information by measuring finger morphology according to claim 5, characterized in that: To determine whether a patient has clubbing, the formula of the logistic regression classification model is as follows: P (positive clubbing) = σ (w0 + w1·AIA + w2·CR + w3·TR + w4·(100-SpO2) + w5·PH); σ(x): Sigmoid function, defined as: Its output is between 0 and 1, indicating the probability of being classified as positive; P(positive clubbing): the probability of predicting positive clubbing; Among them, AIA stands for interphalangeal angle, which is defined as the angle between the nail and the knuckle, in degrees (°); CR stands for curvature radius, which is defined as the radius of curvature of the fingertip in millimeters (mm). In patients with clubbing, the fingertips are more rounded and the radius of curvature is smaller. TR stands for thickness ratio, which is defined as the ratio of the thickness at the finger tip to the thickness at the base of the finger; SpO2 stands for blood oxygen saturation, definition: the degree of hypoxia in blood oxygen saturation (unit: %); PH stands for medical history score, which is defined as a score that quantifies medical history-related characteristics, including smoking index or chronic disease score; Model parameters (w i ) w0: intercept of the model (bias term); When all feature variables are 0, the initial value of the model output; w1,w2,...,w n : The weight corresponding to each feature; Indicates the degree of influence of each feature on the probability of positive clubbing; w i >0: The increase of eigenvalue will increase the probability of positive clubbing; w i <0: an increase in the eigenvalue will reduce the probability of positive clubbing; The absolute value of the weight indicates the strength of the influence; Based on multiple linear regression, predict the clubbing score for a patient: S=β0+β1·AIA+β2·CR+β3·TR+β4·(100-SpO2)+β5·PH+∈ S: Clubbing score (predicted target). β0, β1, ..., β5: model parameters (fitted by training data). AIA: Interdigital angle (unit: degree). CR: radius of curvature (unit: mm). TR: Thickness ratio (the ratio of fingertip thickness to finger base thickness, unitless). SpO2: blood oxygen saturation (unit: %). PH: Characteristics related to the patient's medical history, quantified by numerical values ​​(such as long-term smoking index or specific patient score). ∈: Random error term. .

7. The application system for providing health information by measuring finger morphology according to claim 1, characterized in that: The step S6 is also included: model optimization and performance index evaluation, which specifically includes the following steps: Step S61: Model optimization: by expanding the sample size, improving the generalization ability of the model, and enhancing the data; using grid search or Bayesian optimization to select the best model parameters to tune the hyperparameters; Step S62: Performance index evaluation: verify the effect through sensitivity, specificity, and ROC curve evaluation algorithm; or ensure the reliability of the automation system through measurement error analysis.

8. A system for providing health information by measuring finger morphology according to any one of claims 1 to 7, for detecting finger morphology, comprising a system framework and a mathematical model; The system framework includes a data input module and a core function module at the fingertips; The data input module includes three-dimensional morphological data (such as CT or MRI scans), two-dimensional images (such as X-rays, photos), and basic patient information (such as age, gender). The core functional modules include an image preprocessing module, a feature extraction module and a model analysis module; The mathematical model includes a nail angle model, a distal phalanx ratio, and a volumetric measurement model.

9. The system for providing health information by measuring finger morphology according to claim 7, characterized in that: The image preprocessing module includes denoising, normalization and contrast enhancement processing on the data; The feature extraction module includes extracting anatomical features related to clubbing, including nail angle and distal phalanx thickness. The model analysis module includes analyzing the extracted features using a mathematical model or a machine learning method to determine whether clubbing exists.

10. The system for providing health information by measuring finger morphology according to claim 7, characterized in that: The nail angle model includes establishing an angle describing the nail and the nail bed, wherein the normal angle is preset to be less than 160°, and when the angle is greater than 160°, it is judged as clubbing; The distal phalanx ratio includes calculating the width ratio of the distal phalanx to the middle phalanx. It is preset that the distal phalanx is smaller than the middle phalanx. When the distal phalanx is larger than the middle phalanx, it is determined to be clubbing. The volume measurement model includes obtaining the volume change of the fingertip through three-dimensional modeling.

Images