Search

CN-121983251-A - Cobb angle intelligent detection system based on AES encrypted communication and two-way authentication

CN121983251ACN 121983251 ACN121983251 ACN 121983251ACN-121983251-A

Abstract

The invention belongs to the field of medical intelligent detection and information security intersection, and provides a Cobb angle intelligent detection system based on AES (advanced encryption standard) encryption communication and two-way authentication, which is used for solving the problems of low detection efficiency, large error and insufficient data security in the prior art. The method comprises the steps of S1, data acquisition and encryption uploading, obtaining a spine image through uploading to a Yuxiu intelligent bone applet by a user, meanwhile, adopting AES-256 encryption and two-way authentication to ensure data safety, S2, key point detection, identifying the position of a spine key centrum based on an improved Keypoint R-CNN model introducing a homodyne uncertainty loss weight and CBAM attention module, S3, cobb angle calculation, automatic calculation of each section Cobb angle through a vector dot product and angle solving algorithm, S4, side bending degree evaluation, grading and generating a diagnosis report according to clinical standards, and S5, doctor-patient interaction and data management, and supporting remote interpretation and history record inquiry.

Inventors

  • HAN YU
  • HAN YUNBO
  • LIU GUOCHAO
  • LIU JIAXUAN
  • SUN ZHONGCHUAN
  • WANG YOUWEI
  • TIAN KE

Assignees

  • 郑州大学第一附属医院

Dates

Publication Date
20260505
Application Date
20260124

Claims (9)

  1. 1. An AES-based encrypted communication and two-way authentication Cobb angle intelligent detection system is characterized by comprising the following steps: (1) The data acquisition and encryption uploading, namely receiving spine image or hospital clinical image data uploaded by a user through a Yuxiu intelligent bone WeChat applet, completing identity verification of equipment and a cloud platform by adopting an AES-256 encryption algorithm and a two-way authentication mechanism, and uploading the encrypted data through a TLS 1.3 channel; (2) Improving Keypoint R-CNN model construction and training, namely taking ResNet-50+FPN as a backbone network, embedding CBAM attention module, introducing a homodyne uncertainty loss weight optimization loss function, and completing model training based on a clinical image dataset; (3) Inputting decrypted data into a trained model, identifying 12 spinal centrum key point coordinates, and calculating Cobb angles of all the sections by a vector dot product method; (4) Generating a lateral bending evaluation and report, namely dividing the lateral bending grade according to the Cobb angle, and automatically generating a structured report containing detection data, grade and advice; (5) Doctor-patient interaction and data management, namely a Yuxiu intelligent bone WeChat applet supports the remote interpretation of a doctor and the inquiry history report of a patient, and data is stored in an encrypted manner and meets the requirement of the standard of 2.0 level compliance.
  2. 2. The Cobb angle intelligent detection system according to claim 1, wherein step (1) supports JPG/JPEG, PNG, BMP and other format image uploading, single Zhang MB, 5 batch upper limit, automatic conversion of non-JPG format and quality detection, camera shooting supports flash lamp control, automatic focusing and cutting rotation, image denoising through bilateral filtering and contrast adjustment is integrally converted into RGB mode JPG and compressed, and uploading is performed by checking format and size, 512KB block uploading and breakpoint continuous transmission are adopted, and prompt and solution are given to abnormal conditions.
  3. 3. The Cobb angle intelligent detection system of claim 1 is characterized in that in the step (1), image data and a diagnosis result are encrypted by adopting AES-256, a front end and a rear end complete two-way authentication and key negotiation through SHA-256 hash identity information, a key is generated by a server and derived by HMAC-SHA256, is encrypted by adopting an AES-256-CBC mode and is bound with a user session, a rear end receives and decrypts NIfTI files by adopting an MQTT protocol, a model is input to detect a spine and calculate a Cobb angle (the precision is less than or equal to 1.5 degrees), report and label image encryption is returned to the front end and is backed up to a cloud, and the cloud adopts a segmented encryption strategy of independent IV of every 5 MB.
  4. 4. The Cobb corner intelligent detection system of claim 1 wherein in step (2) the improved Keypoint R-CNN model extracts infrastructure with FPN+ ResNet-50 as characteristic, after generating candidate region and RoI Align layer alignment feature map by RPN, outputting heat map positioning key point by key point detection head, and the mask generating module outlines target outline, two core improvements are as follows: s1, introducing uncertainty of the same variance to learn a loss weight, discarding a traditional fixed weight mode, automatically adjusting the weight by learning variance of each loss term, and designing a model loss function as follows Wherein the method comprises the steps of And The penalty is detected and the mask generated for the keypoint, And The variance of the corresponding task is used as an uncertainty parameter for learning the weight of each task; S2, embedding CBAM attention modules among residual blocks of the feature extractor BackBone, extracting global information through GAP and GMP in a channel attention stage, learning channel dependence by means of MLP, strengthening core region features in a space attention stage, and finally outputting through a sigmoid activation function to accurately focus on a key region of an image so as to improve detection precision.
  5. 5. The intelligent Cobb angle detection system according to claim 1, wherein the calculation process of the Cobb angle in the step (3) mainly comprises the following steps: s1, determining an end cone, namely finding out a cone with the biggest inclination of the curved head side and the tail side of the spine on an X-ray film, wherein the cone is an upper cone and a lower cone respectively; s2, drawing a straight line along the upper end plate of the upper end cone and drawing a straight line along the lower end plate of the lower end cone; S3, calculating an included angle of the two straight lines or an intersection angle of the perpendicular straight lines, wherein the included angle is the Cobb angle. The calculation formula is as follows:
  6. 6. The intelligent Cobb angle detection system according to claim 1, wherein the data visualization module of the micro-letter applet of the intelligent bone platform in the step (4) supports multi-mode viewing of X-ray films, wherein the data visualization module displays Cobb angles to present PT, MT and TL values, displays mark points to present blue mark points and lines, supports adjustment of mark point sizes and types, displays a detection frame with labels and confidence, and displays label supplementary information in combination with the detection frame, generates a report after system analysis, displays detection time and side bending level, visualizes PT, MT and TL angles through a histogram, and provides accurate measurement values and curve types.
  7. 7. The Cobb corner intelligent detection system of claim 1, wherein step (5) uses SQLite to encrypt and store user data, sets an automatic cleaning mechanism to clean out expiration data for 3 months regularly, supports cloud storage and multi-terminal data synchronization, and provides report screening and inquiring functions of time, risk level, angle range and keywords.
  8. 8. The intelligent Cobb angle detection system of claim 1, wherein the intelligent bone-in-doctor program in step (5) is divided into doctor-side and user-side, and the doctor-side supports instant return of consultation, provides diagnosis and treatment guidance, checks detailed diagnosis report of patient and makes treatment plan, and supports comparison of historical data to track change of illness state.
  9. 9. Use of the Cobb angle intelligent detection system of any one of claims 1 to 8 in mass screening, clinical auxiliary diagnosis and telemedicine of juvenile scoliosis.

Description

Cobb angle intelligent detection system based on AES encrypted communication and two-way authentication Technical Field The invention belongs to the technical field of medical intelligent detection and data safety, and relates to a scoliosis automatic detection system integrating deep learning and encrypted communication, which is suitable for large-scale screening, clinical auxiliary diagnosis and remote medical scenes of teenager scoliosis. Background Modern scoliosis, especially idiopathic scoliosis of teenagers, has developed a potential of hidden spreading, known as "chronic disease in posture". Currently, the number of teenager scoliosis patients in China is more than 500 ten thousand, and about 30 ten thousand cases are newly increased each year. The causes are not single, but multidimensional symphony of biomechanics, genetic development and environmental behaviors, and the physical appearance and bone health of a patient are seriously affected. The Cobb angle is a core imaging quantitative index for scoliosis diagnosis, evaluation and follow-up, and has a clinical significance of being basic. However, the current Cobb angle measurement mainly depends on manual operation and professional equipment, and has multiple bottlenecks such as high subjective error rate, unbalanced efficiency and cost, low accessibility of equipment, lack of interaction between patients and doctors and the like. In addition, the conventional spine detection involves image data such as X-ray and MRI, which contains sensitive contents such as personal identity information and physical condition of a patient. In the past, the problems of poor storage management, data transmission loopholes and the like are solved, and mental pressure and privacy invasion risks are easily brought to patients, so that an accurate and safe intelligent detection system is constructed and is an important means for early screening, auxiliary diagnosis and data safety management of scoliosis. Currently, the proposal of the spine detection is mainly a manual measuring tool or single-function equipment, for example, chinese patent application CN202320567890.X discloses a scoliosis measuring ruler. The commonly used intelligent detection scheme mainly comprises an image recognition model and a simple screening tool, for example, chinese patent application CN202211345678.9 discloses a spine posture screening method based on a common camera, and Chinese patent application CN202110876543.2 discloses a spine X-ray film key point manual labeling system. However, in the existing spine detection scheme, the combination of the precision optimization of the deep learning model and the medical data encryption is not realized, and the combination is not directly related to the actual requirements of the accurate measurement of the clinical Cobb angle and the data privacy protection, so that the problems of large detection error, low efficiency, high risk of data leakage and the like cannot be solved at the same time. Therefore, the above-mentioned related schemes for spine detection are difficult to meet the requirements of large-scale screening and clinical auxiliary diagnosis. Disclosure of Invention In order to solve the problems that scoliosis detection accuracy is low, data safety is insufficient and large-scale screening and clinical requirements are difficult to meet in the prior art, the invention constructs a Cobb angle intelligent detection system based on AES (advanced encryption standard) encryption communication and two-way authentication by combining an encryption technology through improving a deep learning model, and realizes integration of accurate detection and data safety management. In order to achieve the purpose, the invention adopts the following technical scheme that the Cobb angle intelligent detection system based on AES encrypted communication and two-way authentication comprises the following steps: (1) The data acquisition and encryption uploading, namely receiving spine image or hospital clinical image data uploaded by a user through a Yuxiu intelligent bone WeChat applet, and uploading the encrypted data through a channel after the identity verification of equipment and a cloud is completed by adopting an AES-256 encryption algorithm and a two-way authentication mechanism; (2) Improving Keypoint R-CNN model construction and training, namely embedding CBAM attention module into backbone network, introducing homodyne uncertainty loss weight optimization loss function, and completing model training based on clinical image dataset; (3) Inputting decrypted data into a trained model, identifying 12 spinal centrum key point coordinates, and calculating Cobb angles of all the sections by a vector dot product method; (4) Generating a lateral bending evaluation and report, namely dividing the lateral bending grade according to the Cobb angle, and automatically generating a structured report containing detection data, grade and advice; (5) Docto