CN-122025124-A - Female age prediction system and method integrating facial image and biochemical index

CN122025124ACN 122025124 ACN122025124 ACN 122025124ACN-122025124-A

Abstract

The invention provides a female age prediction system and method for fusing facial images and biochemical indexes, belonging to the field of artificial intelligence and health management, wherein the system comprises a data acquisition and preprocessing module, a dual-channel feature extraction module, a multi-mode feature topology fusion module, an age prediction and health assessment module and a personalized health intervention module, feature fusion is realized through a cross-modal attention interaction network and a topological consistency maintenance mechanism, the system predicts the biological age based on the fusion features, generates a three-dimensional health risk score, combines female physiological cycle data for dynamic calibration, the invention solves the technical contradiction that the traditional detection method has high precision, high cost and obvious error, and provides accurate and reliable female physiological age assessment.

Inventors

YIN XIN
LI LISHA
ZHANG SHUNYAO

Assignees

吉林大学

Dates

Publication Date: 20260512
Application Date: 20260126

Claims (10)

1. A female age prediction system that fuses facial images with biochemical indicators, comprising: the data acquisition and preprocessing module is used for acquiring female biochemical index data and facial image data and preprocessing the biochemical index data and the facial image data; The double-channel feature extraction module is connected with the data acquisition and preprocessing module and is used for extracting biochemical features from the preprocessed biochemical index data, extracting facial features from the preprocessed facial image data and acquiring female physiological cycle data; The multi-modal feature topology fusion module is connected with the two-channel feature extraction module and is used for constructing a biochemical feature topology space and a facial feature topology space, realizing alignment mapping of the biochemical feature topology space and the facial feature topology space, calculating feature fusion weights through a cross-modal attention interaction network and realizing fusion of the biochemical features and the facial features based on topological consistency constraint; the age prediction and health assessment module is connected with the multi-mode feature topology fusion module and used for receiving the fused features, predicting the biological age, calculating the age acceleration, generating a health risk score, and dynamically calibrating the biological age and the health risk score based on the female physiological cycle data.
2. The system of claim 1, wherein the data acquisition and preprocessing module comprises: the biochemical index acquisition unit is used for acquiring biochemical index data of females, wherein the biochemical index data comprise indexes such as estrogen, progesterone, creatinine, alkaline phosphatase and the like; the biochemical index preprocessing unit is connected with the biochemical index acquisition unit and is used for carrying out missing value processing, abnormal value detection and numerical normalization on the biochemical index data; a face image acquisition unit configured to acquire face image data of a female; the image preprocessing unit is connected with the facial image acquisition unit and is used for carrying out face detection, image alignment, illumination normalization and characteristic region segmentation on the facial image data; And the cycle data acquisition unit is used for acquiring information such as the start-stop time, the cycle length, the symptom record and the like of the female menstrual cycle and converting the information into standardized time sequence characteristics.
3. The system of claim 1, wherein the dual channel feature extraction module comprises: the biochemical characteristic extraction unit is used for extracting biochemical characteristics from the pretreated biochemical index data through a multi-layer characteristic extraction network, wherein the multi-layer characteristic extraction network comprises a first characteristic layer, a second characteristic layer and a third characteristic layer which are sequentially connected; a facial feature extraction unit for extracting facial features from the preprocessed facial image data through a hierarchical feature extraction network, the hierarchical feature extraction network including a shallow feature layer for extracting textures and pigments, a middle feature layer for extracting skin elasticity and wrinkle distribution, and a deep feature layer for extracting skeletal structure changes; And the physiological cycle characteristic extraction unit is used for converting the female physiological cycle data into cycle state codes and generating hormone fluctuation characteristics by combining the biochemical characteristics.
4. The system of claim 1, wherein the multi-modal feature topology fusion module comprises: A topology space construction unit for constructing a feature topology space for the biochemical feature and the facial feature, respectively; The space alignment unit is connected with the topological space construction unit and is used for realizing the initial alignment of the biochemical characteristic topological space and the facial characteristic topological space through reference point matching and determining the mapping relation between the biochemical characteristic topological space and the facial characteristic topological space; An attention calculating unit, connected to the space alignment unit, for generating a query vector from the biochemical feature topology space, generating a key vector from the facial feature topology space, and calculating an attention weight based on the similarity of the query vector and the key vector; the topological consistency fusion unit is connected with the attention calculation unit and is used for applying neighborhood keeping constraint and distance keeping constraint in the feature fusion process to ensure that the feature topological structure is not destroyed in the fusion process.
5. The system of claim 4, wherein the attention computing unit comprises a multi-headed attention network including a plurality of independent attention heads, each of the attention heads being responsible for learning a particular type of cross-modal feature association, the outputs of the multi-headed attention network fusing the results of the individual attention heads by means of weighted combination.
6. The system of claim 1, wherein the age prediction and health assessment module comprises: The decoding network unit is used for receiving the fused characteristics and mapping the high-dimensional characteristics to an age prediction space through progressive characteristic conversion; The age prediction unit is connected with the decoding network unit and is used for outputting biological age values, calculating the difference and the ratio of the biological ages to the actual ages and generating probability distribution of different age segments; a risk assessment unit, connected to the decoding network unit, for generating a three-dimensional risk score for metabolic, endocrine, and cardiovascular risk based on the fusion features; And the dynamic calibration unit is connected with the age prediction unit and the risk assessment unit and is used for carrying out cycle stage-related dynamic calibration on the biological age and the three-dimensional risk score based on the female physiological cycle data.
7. The system of claim 1, further comprising a personalized health intervention module coupled to the age prediction and health assessment module, the personalized health intervention module comprising: the knowledge map engine is used for storing health intervention knowledge, wherein the health intervention knowledge comprises a nutrition database and Chinese and Western medicine conditioning knowledge; a query processing unit, configured to query the knowledge graph engine based on the biological age and the health risk score, and obtain related intervention knowledge; The intervention scheme generation unit is connected with the query processing unit and is used for generating personalized health intervention schemes containing Western medicine suggestions and Chinese medicine suggestions; And the effect prediction unit is connected with the intervention scheme generation unit and is used for simulating potential effects of different intervention schemes and recommending optimal intervention scheme combinations.
8. The system of claim 1, further comprising a data storage module for storing historical detection data, user information, and intervention records, the data storage module being coupled to the data acquisition and preprocessing module, the age prediction and health assessment module for providing data persistence storage and historical data analysis functionality.
9. The system of claim 1, wherein the system employs a lightweight B/S architecture, comprising: the front-end module is used for providing a responsive user interface and supporting biochemical index input, facial image uploading and health report display; the back-end service module is connected with the front-end module, and is used for processing the front-end request, coordinating the work of each functional module and providing an API interface to support the integration of a third-party system; The model deployment module is connected with the back-end service module and used for providing model reasoning service and supporting model increment updating.
10. A female age prediction method for fusing facial images with biochemical indicators, using the system of any one of claims 1-9, comprising the steps of: acquiring female biochemical index data and facial image data, and preprocessing the biochemical index data and the facial image data; Extracting biochemical characteristics from the pretreated biochemical index data, extracting facial characteristics from the pretreated facial image data, and acquiring female physiological cycle data; Constructing a biochemical feature topological space and a facial feature topological space, realizing alignment mapping of the biochemical feature topological space and the facial feature topological space, calculating feature fusion weights through a cross-modal attention interaction network, and realizing fusion of the biochemical features and the facial features based on topological consistency constraints; Receiving the fused features, predicting the biological age, calculating the age acceleration, generating a health risk score, and dynamically calibrating the biological age and the health risk score based on the female physiological cycle data.

Description

Female age prediction system and method integrating facial image and biochemical index Technical Field The invention relates to the fields of artificial intelligence, medical health and data fusion, in particular to a female age prediction system and method for fusing facial images and biochemical indexes. Background The biological age is used as a comprehensive index for reflecting the real health state of an individual, is different from the calendar age (actual age), and has important guiding significance for early screening of diseases and health intervention. In recent years, as the population ages and the female health awareness increases, the need for accurate assessment of biological age increases. Existing biological age assessment methods are mainly divided into three categories, epigenetic assessment based on DNA methylation, activity monitoring assessment based on wearable devices, and visual assessment based on facial image analysis. However, there are significant limitations to the prior art. DNA methylation detection, although highly accurate (error about ±0.8 years old), is expensive (single cost exceeding 3000 yuan), and relies on specialized laboratories, which is difficult to popularize in the background. Methods based on wearable devices (such as Fitbit, etc.) rely only on kinetic sleep data, with errors reaching + -3-5 years old, lacking in depth analysis of biochemical indicators. The facial image analysis technology only evaluates the epidermis age, but cannot early warn the metabolism and endocrine risks. In addition, the existing method generally ignores the influence of the specific physiological cycle of females on biological age assessment, so that the assessment results for female populations have large fluctuation. Particularly, in the aspect of multi-mode data fusion, the prior art mostly adopts a simple characteristic splicing or weighted average method, and neglects the inherent structural relationship among different mode data, so that the fusion effect is limited. Meanwhile, traditional health intervention suggestions are often seriously homogenized, and a customized scheme aiming at individual specificity is lacking, so that the requirement of accurate health management is difficult to meet. Therefore, there is a need for a biological age assessment system that can comprehensively utilize biochemical indicators and facial image data, and that takes into account female-specific physiological characteristics, to achieve low-cost, high-precision physiological age prediction and personalized health intervention. Disclosure of Invention The invention aims to provide a female age prediction system and method for fusing facial images and biochemical indexes, which are used for realizing high-precision and low-cost biological age prediction and personalized health intervention by combining female physiological cycle characteristics through an innovative multi-mode characteristic topological fusion architecture. The invention provides a female age prediction system fusing facial images and biochemical indexes, which comprises: the data acquisition and preprocessing module is used for acquiring female biochemical index data and facial image data and preprocessing the biochemical index data and the facial image data; The double-channel feature extraction module is connected with the data acquisition and preprocessing module and is used for extracting biochemical features from the preprocessed biochemical index data, extracting facial features from the preprocessed facial image data and acquiring female physiological cycle data; The multi-modal feature topology fusion module is connected with the two-channel feature extraction module and is used for constructing a biochemical feature topology space and a facial feature topology space, realizing alignment mapping of the biochemical feature topology space and the facial feature topology space, calculating feature fusion weights through a cross-modal attention interaction network and realizing fusion of the biochemical features and the facial features based on topological consistency constraint; the age prediction and health assessment module is connected with the multi-mode feature topology fusion module and used for receiving the fused features, predicting the biological age, calculating the age acceleration, generating a health risk score, and dynamically calibrating the biological age and the health risk score based on the female physiological cycle data. Preferably, the data acquisition and preprocessing module includes: the biochemical index acquisition unit is used for acquiring biochemical index data of females, wherein the biochemical index data comprise indexes such as estrogen, progesterone, creatinine, alkaline phosphatase and the like; the biochemical index preprocessing unit is connected with the biochemical index acquisition unit and is used for carrying out missing value processing, abnormal value detection and numerical