CN-122024276-A - Intelligent and accurate positioning method for Baliao acupoint based on clinical knowledge-multi-mode image data combined driving

Abstract

The invention relates to the technical field of acupuncture point positioning, in particular to an intelligent accurate positioning method for Baliao points based on combined driving of clinical knowledge and multi-mode image data, which is used for collecting multi-mode image data of a sacrum region of a patient, preprocessing the multi-mode image data, extracting multi-mode fusion characteristics through a teacher-student network architecture in a positioning model after training, generating space constraint parameters by combining a clinical knowledge constraint sensing module, obtaining Baliao point coordinates through the teacher-student network architecture in the positioning model after training, and carrying out post-processing on the Baliao point coordinates and integrating the Baliao point coordinates into a mobile terminal APP system. The invention realizes that the positioning error is less than or equal to 10mm through the combined driving of the multi-mode image data and the clinical knowledge of the traditional Chinese medicine.

Inventors

• Dong Aiai
• DONG QI
• XU HANG
• LIU LI

Assignees

• 山西中医药大学

Dates

Publication Date

20260512

Application Date

20251231

Claims (5)

1. 1. The intelligent and accurate positioning method for the Baliao acupoint based on the combined driving of clinical knowledge and multi-mode image data is characterized by comprising the following steps of: s1, acquiring multi-mode image data of a sacrum region of a patient to obtain a multi-mode image database, wherein the multi-mode image database comprises a pelvic cavity CT image and a body surface RGB image; S2, preprocessing the multi-mode image data obtained in the step S1; s3, extracting multi-mode fusion characteristics through a teacher-student network architecture in the trained positioning model, and generating space constraint parameters by combining a clinical knowledge constraint sensing module; S4, integrating the trained teacher-student network model into a mobile terminal APP system.
2. 2. The intelligent accurate positioning method for the Baliao acupoints based on clinical knowledge-multi-mode image data combined driving according to claim 1, wherein the multi-mode acupoint database comprises basic acupoint data and space positioning data, and the basic acupoint data is obtained by the following steps: A1, collecting a plurality of groups of pelvic cavity CT images and body surface RGB images; a2, preprocessing the pelvic cavity CT image and the body surface RGB image obtained in the A1, and marking acupoint information, wherein the marked acupoint information is the standard name, anatomical positioning description, main treatment function and indication of the Baliao acupoint; the process of acquiring the space positioning data comprises the following steps: b1, acquiring three-dimensional data of a sacrum region of a human body through three-dimensional scanning equipment; and B2, labeling three-dimensional coordinate information of the Baliao acupoint and relative distance and angle relation between the Baliao acupoint and the anatomical landmark point on the three-dimensional data.
3. 3. The intelligent accurate positioning method for the Baliao acupoint based on the combined driving of clinical knowledge and multi-mode image data is characterized in that in a teacher-student network architecture, a teacher network is based on a pelvic CT image, resNet-50 backbone is adopted, a CBAM attention mechanism is integrated, the accurate spatial position characteristics of a sacral foramen are focused, output characteristics are synchronously transmitted to an anatomical constraint submodule of a clinical knowledge constraint perception module for space parameter verification, a student network is based on an RGB image, a MobileNetV3 light-weight architecture is adopted, the inference speed is optimized through depth separable convolution, the output characteristics are required to meet the individual same-body constraint of a clinical rule submodule, and the teacher network and the student network realize characteristic fusion through multi-level knowledge distillation.
4. 4. The intelligent accurate positioning method for the Baliao acupoint based on the combined driving of clinical knowledge and multi-mode image data according to claim 1, wherein the constraint mechanism of the clinical knowledge constraint perception module is as follows: (1) Space symmetry and sequence constraint, which maintains the left-right symmetry rule of human anatomy structure and the basic vertical arrangement sequence of upper liao, secondary liao, middle liao and lower liao by average symmetry deviation and sequence integrity product, the mathematical expression is: (1) In the formula, The coefficient of the averaging is represented as such, Representing the summation of the 4 liao points, Representing the coordinate position of the i-th liao point on the left, Representing the coordinate position of the i-th liao point on the right, Representing the sacral midline of the patient, Represents a symmetry-tolerance threshold value, The indication function is represented by a representation of the indication function, Representing the ordinate of the ith liao point, The sequence constraints are represented by a sequence of values, Representing the distance from the ith liao point on the left to the midline, Representing the distance from the ith liao point on the right to the midline, Representing a continuous multiplication operation; (2) The geometric relation constraint of adjacent acupoints, the constraint specifically constrains the spatial relation between adjacent acupoints through the combination of normalized adjacent interval deviation and normalized colinear deviation, promotes the eight liao acupoints to present a reasonable linear arrangement mode, and the adjacent interval basically accords with the anatomical statistical rule, and the mathematical expression is as follows: (2) In the formula, The coefficient of the averaging is represented as such, Representing the summation of 3 adjacent acupoint pairs, Representing the summation of the 4 liao points, Representing the coordinate location of the ith liao point, Representing the coordinate position of the i+1st liao point, The actual euclidean distance between adjacent acupoints, Indicating the reference spacing of the ith adjacent acupoint pair, The pitch ratio is represented by the ratio of the pitches, The standard deviation of the pitch is indicated, A best-fit straight line is shown, Representing the distance from the ith liao point to the fitted straight line, Representing the maximum allowable co-linearity deviation, The co-linearity weight coefficient is used to determine the co-linearity weight coefficient, Representing a neighbor relationship consistency threshold; (3) An individualized isotactical constraint that ensures that the absolute positional relationship of each liao acupoint and the sacral midline conforms to individual anatomical features, expressed mathematically as: (3) In the formula, Representing the normalized importance weight of the i-th liao point, The indication function is represented by a representation of the indication function, Representing the actual distance of the ith liao point to the midline, Representing the reference distance from the ith Liao point to the midline under the standard body type, Representing the same-size scale factor as that of the human body, Representing the individual reference distance of the object, The absolute deviation is indicated and the absolute deviation is indicated, The relative deviation rate is indicated by the relative deviation, The weighted voting score is represented and, Represents a personalized deviation tolerance threshold value, Representing the identity uniformity threshold.
5. 5. The intelligent accurate positioning method for the Baliao acupoints based on the combined driving of clinical knowledge and multi-mode image data is characterized in that a trained positioning model is converted into a format supported by a mobile-end reasoning frame through a model conversion tool and deployed on a mobile-end APP, a real-time RGB image is acquired through a mobile-end camera, the position of the Baliao acupoints is obtained through model reasoning, the position is displayed on the real-time image, and the Baliao acupoint name, positioning information and needling depth prompt are output.

Description

Intelligent and accurate positioning method for Baliao acupoint based on clinical knowledge-multi-mode image data combined driving Technical Field The invention relates to the technical field of acupuncture point positioning, in particular to an intelligent and accurate positioning method for Baliao points based on combined driving of clinical knowledge and multi-mode image data. Background The Chinese medicine acupuncture is used as an important component of traditional medicine, and has wide application in the fields of treating pelvic diseases, urinary system dysfunction, chronic pain and the like, wherein the accurate positioning of the Baliao acupoint is important for curative effect. The traditional Baliao acupoint positioning method mainly relies on palpation of doctors and visual inspection of body surface anatomical marks (such as sacral angle and posterior superior iliac spine), and combines clinical experience to judge. This approach has the following limitations: 1. The operation efficiency is low, the positioning of each acupoint needs repeated touching and measurement, the time consumption is long, and the diagnosis and treatment efficiency is affected; 2. Subjective dependence is high, and inconsistent positioning results can be caused by the experience level of different doctors; 3. Is easy to be interfered by individual difference, and the positioning difficulty is increased by the body type, the obesity degree or the skeletal variation of the patient; 4. environmental factors influence that light conditions or body position changes may further reduce reliability. In recent years, computer vision and deep learning techniques have advanced in medical image analysis, such as human body keypoint detection through convolutional neural networks. However, the direct application of these techniques to Baliao point positioning still faces challenges: 1. the positions of the acupoints lack obvious visual characteristics, and the deep bone structure is difficult to map accurately only by means of a single image mode (such as RGB image); 2. The registration and fusion complexity of multi-mode data (such as CT and RGB images) is high, and errors are easy to introduce; 3. the model is suitable for anatomical variation of different people, but the traditional algorithm has limited generalization capability, and is difficult to integrate clinical knowledge of traditional Chinese medicine (such as the spatial distribution rule of the sacral foramen). Therefore, aiming at the problems of insufficient positioning precision, low efficiency and lack of multi-mode data collaboration in the prior art, an intelligent positioning method integrating clinical knowledge and multi-mode images needs to be developed so as to improve the accuracy and the practicability of the Baliao acupoint positioning. Disclosure of Invention The invention provides an intelligent accurate positioning method and system for Baliao acupoints based on combined driving of clinical knowledge and multi-mode image data, which aims to solve the problems of insufficient positioning precision, low efficiency and lack of multi-mode data cooperation of the existing acupoints. The invention is realized by the following technical scheme that the intelligent and accurate positioning method of Baliao acupoint based on clinical knowledge-multi-mode image data combined driving comprises the following steps: s1, acquiring multi-mode image data of a sacrum region of a patient to obtain a multi-mode image database, wherein the multi-mode image database comprises a pelvic cavity CT image and a body surface RGB image; S2, preprocessing the multi-mode image data obtained in the step S1; s3, extracting multi-mode fusion characteristics through a teacher-student network architecture in the trained positioning model, and generating space constraint parameters by combining a clinical knowledge constraint sensing module; S4, integrating the trained teacher-student network model into a mobile terminal APP system. As a further improvement of the technical scheme of the invention, the multi-mode acupoint database comprises basic acupoint data and space positioning data, and the basic acupoint data is obtained as follows: A1, collecting a plurality of groups of pelvic cavity CT images and body surface RGB images; a2, preprocessing the pelvic cavity CT image and the body surface RGB image obtained in the A1, and marking acupoint information, wherein the marked acupoint information is the standard name, anatomical positioning description, main treatment function and indication of the Baliao acupoint; the process of acquiring the space positioning data comprises the following steps: b1, acquiring three-dimensional data of a sacrum region of a human body through three-dimensional scanning equipment; and B2, labeling three-dimensional coordinate information of the Baliao acupoint and relative distance and angle relation between the Baliao acupoint and the anatomical landmar