CN-122004932-A - Full-automatic three-dimensional (3D) ultrasonic inspection system for brain under guidance of robot and LLM model

CN122004932ACN 122004932 ACN122004932 ACN 122004932ACN-122004932-A

Abstract

The invention discloses a full-automatic brain 3D ultrasonic inspection system guided by robots and LLM models, which constructs a closed loop system integrating intelligent semantic understanding, accurate robot control, high-quality image acquisition and intelligent diagnosis report generation through software and hardware deep integration and intelligent algorithms, realizes the full-process from clinical request to diagnosis report and end-to-end automation, remarkably improves the standardization degree, repeatability, efficiency and accuracy of brain ultrasonic inspection, reduces the workload of doctors, and provides a high-efficiency and reliable technical tool for early screening, accurate diagnosis and treatment planning of brain diseases.

Inventors

YING TAO
HUA CHEN
XU XINRU
TAN JIALU
SU YIXUAN
CUI XIAOYU

Assignees

上海市第六人民医院

Dates

Publication Date: 20260512
Application Date: 20260112

Claims (5)

1. A robot and LLM model guided full-automatic 3D ultrasound inspection system for the brain, comprising: The central intelligent control module is used for receiving a structured and/or natural language scanning request submitted by a doctor through a hospital information network, automatically analyzing clinical intention based on a built-in LLM model, identifying a scanning target area in the clinical intention, generating a structured task list, inquiring a brain ultrasound examination-available area knowledge base according to a target area in the task list, judging whether the request is within a technical feasible range, judging that the request is invalid if the request is aimed at a bone deep area which cannot be penetrated by ultrasound, returning rejection reasons and recommending alternative examination schemes to an applicant doctor through an HIS, and generating guiding patient positioning data and scanning requirements if the request is effectively based on a standard anatomical map; The patient scheduling and managing system is used for acquiring positioning data of a guided patient, displaying patient queue information on a display screen of a waiting area according to the queue sequence, guiding the patient to lie on an adjustable examination bed in a supine position through voice and animation on a wall display screen after the patient is detected to enter an examination room, and placing the head in a head support groove; The adjustable examining table is used for synchronously adjusting according to the positioning data of the guiding patient, and transmitting the position information of the patient to the central intelligent control module in real time until the position of the patient meets the requirement of the guiding patient on the data; The high-precision 3D structure optical vision sensor and the high-precision six-degree-of-freedom mechanical arm are used for acquiring head and neck pose information of a patient and face and head and neck point clouds of the patient, transmitting the head and neck pose information of the patient and the face and head and neck point clouds of the patient to the central intelligent control module, operating a rapid point cloud registration algorithm to register the acquired point clouds with a pre-stored standard supine scanning head model, calculating to obtain position and pose deviation between the acquired point clouds, combining patient position information, scanning target area and scanning requirements based on the deviation, generating an accurate adjustable inspection bed adjustment instruction sequence, performing iterative fine adjustment after the adjustable inspection bed is used for executing adjustment, ensuring that the errors of the actual position and the standard model of the head of the patient are within a preset threshold, obtaining a head 3D surface model of the patient after the body position calibration of the patient, combining a structured task list, geometric and acoustic parameters of an ultrasonic probe, prohibiting access area related safety constraint, calculating key parameter information of scanning control through multi-mode information fusion and self-adaptive path algorithm, reading the current coordinate information of the high-mode mechanical arm and the scanning target area, performing initial coordinate adjustment on the scanning arm, and performing the motion track adaptation to the mechanical arm, and finally performing the motion planning based on the high-precision mechanical arm and the motion track of the head by using the high-precision mechanical arm; The special ultrasonic scanning device is used for driving the special ultrasonic scanning device to perform stable and uniform contact scanning along the scanning track of the mechanical arm by the high-precision six-degree-of-freedom mechanical arm under the force-position mixed control, and simultaneously monitoring and adjusting the contact pressure of the probe in real time to ensure the image quality, and continuous 2D ultrasonic sequence images acquired in the scanning process are transmitted to the central intelligent control module in real time; The diagnosis report display and output module is used for screening high-quality frames from continuous 2D ultrasonic sequence images by the central intelligent control module based on a real-time image quality evaluation algorithm of deep learning, rapidly synthesizing high-resolution isotropic brain 3D ultrasonic images by using an improved sparse view three-dimensional reconstruction algorithm (such as an iterative reconstruction algorithm based on a deep learning priori), carrying out automatic quantitative analysis on the brain 3D ultrasonic images, combining quantitative analysis data with a medical knowledge base to generate a structured preliminary diagnosis report, acquiring the preliminary diagnosis report transmitted by the central intelligent control module, acquiring an auditing result of an ultrasonic doctor, returning the preliminary diagnosis report to the central intelligent control module if the auditing result of the ultrasonic doctor passes the auditing, and returning the preliminary diagnosis report to a doctor applying workstation by the central intelligent control module through a hospital information network.
2. The robotic and LLM model guided brain full-automatic 3D ultrasound examination system of claim 1, wherein the scan target region comprises a specific brain lobe, vascular region, or lesion suspected site.
3. The fully automated 3D ultrasound system for brain under the guidance of a robot and LLM model according to claim 1, wherein the automated quantitative analysis includes ventricle volume measurement, centerline offset calculation, critical vessel flow spectrum extraction and parameter calculation, automatic segmentation and feature extraction of abnormal echo regions.
4. The robotic and LLM model directed fully automated 3D ultrasound examination system for brain as claimed in claim 1, wherein the quantitative analysis data is a structured data summary of screenshots with quantitative results, key parameters.
5. The robotic and LLM model directed fully automated 3D ultrasound examination system of brain as claimed in claim 1, wherein the preliminary diagnostic report includes examination findings, measurement data, abnormality cues, diagnostic impressions, and follow-up examination suggestions.

Description

Full-automatic three-dimensional (3D) ultrasonic inspection system for brain under guidance of robot and LLM model Technical Field The invention belongs to the technical field of medical imaging technology and intelligent robots, and particularly relates to a full-automatic three-dimensional (3D) ultrasonic inspection system for a brain under the guidance of a robot and an LLM model. Background Traditional ultrasound scanning of the brain is largely dependent on experience and skill of the operator, and is low in standardization, limited in efficiency and split in diagnostic procedures, which not only results in variability in the inspection results, but also affects the inspection efficiency. Despite the existence of the partially semi-automatic auxiliary scanning equipment, manual intervention positioning and path planning are still needed, intelligent analytic capability directly interfacing with clinical semantic requirements is lacking, and full-flow closed-loop automation from examination instructions to diagnostic reports is difficult to achieve. Disclosure of Invention In order to solve the technical problems, the technical scheme of the invention provides a full-automatic 3D ultrasonic inspection system for brain under the guidance of a robot and a LLM model, which comprises the following components: The central intelligent control module is used for receiving a structured and/or natural language scanning request submitted by a doctor through a hospital information network, automatically analyzing clinical intention based on a built-in LLM model, identifying a scanning target area in the clinical intention, generating a structured task list, inquiring a brain ultrasound examination-available area knowledge base according to a target area in the task list, judging whether the request is within a technical feasible range, judging that the request is invalid if the request is aimed at a bone deep area which cannot be penetrated by ultrasound, returning rejection reasons and recommending alternative examination schemes to an applicant doctor through an HIS, and generating guiding patient positioning data and scanning requirements if the request is effectively based on a standard anatomical map; The patient scheduling and managing system is used for acquiring positioning data of a guided patient, displaying patient queue information on a display screen of a waiting area according to the queue sequence, guiding the patient to lie on an adjustable examination bed in a supine position through voice and animation on a wall display screen after the patient is detected to enter an examination room, and placing the head in a head support groove; The adjustable examining table is used for synchronously adjusting according to the positioning data of the guiding patient, and transmitting the position information of the patient to the central intelligent control module in real time until the position of the patient meets the requirement of the guiding patient on the data; The high-precision 3D structure optical vision sensor and the high-precision six-degree-of-freedom mechanical arm are used for acquiring head and neck pose information of a patient and face and head and neck point clouds of the patient, transmitting the head and neck pose information of the patient and the face and head and neck point clouds of the patient to the central intelligent control module, operating a rapid point cloud registration algorithm to register the acquired point clouds with a pre-stored standard supine scanning head model, calculating to obtain position and pose deviation between the acquired point clouds, combining patient position information, scanning target area and scanning requirements based on the deviation, generating an accurate adjustable inspection bed adjustment instruction sequence, performing iterative fine adjustment after the adjustable inspection bed is used for executing adjustment, ensuring that the errors of the actual position and the standard model of the head of the patient are within a preset threshold, obtaining a head 3D surface model of the patient after the body position calibration of the patient, combining a structured task list, geometric and acoustic parameters of an ultrasonic probe, prohibiting access area related safety constraint, calculating key parameter information of scanning control through multi-mode information fusion and self-adaptive path algorithm, reading the current coordinate information of the high-mode mechanical arm and the scanning target area, performing initial coordinate adjustment on the scanning arm, and performing the motion track adaptation to the mechanical arm, and finally performing the motion planning based on the high-precision mechanical arm and the motion track of the head by using the high-precision mechanical arm; The special ultrasonic scanning device is used for driving the special ultrasonic scanning device to perform stable and uniform contact scanning alo