CN-121746139-B - Neuroimmune disease immersive virtual patient diagnosis and treatment thinking training system

CN121746139BCN 121746139 BCN121746139 BCN 121746139BCN-121746139-B

Abstract

The invention provides a nerve immune disease immersive virtual patient diagnosis and treatment thinking training system which comprises a multi-dimensional case holographic modeling module, a diagnosis and treatment thinking anchor point dynamic capture module, a double-AI collaborative diagnosis and treatment deduction module, a evidence-based feedback and thinking deviation correction module and a diagnosis and treatment capacity quantitative assessment module which are sequentially connected in data and form closed loop optimization. The invention aims to solve the problems that the scene is static, thinking evaluation is absent, and an AI scheme is separated from clinical pain points without closed-loop optimization in the conventional neural immune disease diagnosis and treatment training.

Inventors

YAO XIAOYING
ZHI NAN
Pan Yuanmei

Assignees

上海交通大学医学院附属仁济医院

Dates

Publication Date: 20260508
Application Date: 20260225

Claims (4)

1. The nerve immune disease immersive virtual patient diagnosis and treatment thinking training system is characterized by comprising a multi-dimensional case holographic modeling module, a diagnosis and treatment thinking anchor point dynamic capture module, a double AI collaborative diagnosis and treatment deduction module, a evidence-based feedback and thinking deviation correction module and a diagnosis and treatment capacity quantitative assessment module which are sequentially connected in data and form closed loop optimization; The multi-dimensional case holographic modeling module inputs authoritative clinical case data and evidence-based medical guideline content, a case core anchor point data set is generated through medical ontology modeling, a holographic VR case scene file with pathological physiological dynamic evolution rules is built by combining a medical image reconstruction technology, and the case core anchor point data set and the holographic VR case scene file are used as input of a diagnosis and treatment thinking anchor point dynamic capturing module together; The diagnosis and treatment thinking anchor point dynamic capture module collects multi-mode interaction data of students through VR equipment based on input content, adopts a two-algorithm bidirectional fusion thinking anchor point quantization method to generate a time sequence anchor point sequence vector, and the time sequence anchor point sequence vector is used as input of the double AI collaborative diagnosis and treatment deduction module; The double AI collaborative diagnosis and treatment deduction module adopts a double AI collaborative deduction method to process the input time sequence anchor point sequence vector, and generates a virtual checking result, a differential diagnosis key point primary screening result, a virtual treatment scheme and scheme demonstration parameters which are used as the input of the evidence-based feedback and thinking deviation correction module together; The evidence-based feedback and thinking deviation correction module compares the input content with the sequence vector of the time sequence anchor point, completes VR treatment effect demonstration by combining scheme demonstration parameters, generates an evidence-based feedback report and thinking deviation data, and is used as the input of the diagnosis and treatment capability quantitative evaluation module together; The diagnosis and treatment capacity quantization evaluation module finishes the diagnosis and treatment capacity quantization scoring of the trainee based on the input content and generates a difficulty adjustment instruction, and the difficulty adjustment instruction is reversely input to the multidimensional case holographic modeling module for optimizing the scene of the subsequent case; The double-algorithm bidirectional fusion thinking anchor quantization method specifically comprises an anchor point association reasoning algorithm based on a cognitive map and a time sequence attention weighted anchor point sequence generation algorithm, wherein the association reasoning weight output by the anchor point association reasoning algorithm based on the cognitive map is used as a calculation parameter of the time sequence attention weighted anchor point sequence generation algorithm, and the time sequence attention weight output by the time sequence attention weighted anchor point sequence generation algorithm is reversely fed back to the anchor point association reasoning algorithm based on the cognitive map and is used for correcting the operation-anchor point semantic association degree; the anchor point association reasoning algorithm based on the cognitive map constructs a cognitive map comprising a neural immune disease diagnosis and treatment whole-flow anchor point node, a clinical association edge and an association intensity weight matrix, and based on the semantic association degree of the learner multi-mode interaction data and the anchor point node, the association reasoning weight of the operation and the anchor point is calculated by combining the weight matrix; The time sequence attention weighted anchor point sequence generation algorithm calculates a time sequence score according to the deviation between the operation moment of a student and the recommended operation time sequence of the evidence-based medical guideline, obtains time sequence attention weight based on the time sequence score, and then fuses the associated reasoning weight and the anchor point feature vector to generate a time sequence anchor point sequence vector; The double AI collaborative deduction method specifically comprises a multi-scale pathological feature fusion checking result judging algorithm and a evidence-based constraint generating type anti-network treatment scheme generating algorithm, wherein a virtual checking result output by the multi-scale pathological feature fusion checking result judging algorithm is used as a core constraint condition of the evidence-based constraint generating type anti-network treatment scheme generating algorithm, and a scheme rationality judging value output by the evidence-based constraint generating type anti-network treatment scheme generating algorithm is reversely fed back to the multi-scale pathological feature fusion checking result judging algorithm to adjust feature fusion weights; The inspection result judgment algorithm of the multi-scale pathological feature fusion adopts convolution kernels of different sizes to extract fine granularity, middle granularity and coarse granularity pathological features of sequential anchor sequence vectors, the multi-scale feature fusion is completed through preset weights, and a positive or negative judgment conclusion of a virtual inspection result is output based on the fusion feature vectors; The evidence-based constraint generating type countermeasure network treatment scheme generating algorithm is internally provided with evidence-based constraint coefficients calculated by matching degree of the virtual examination result and the evidence-based medical guideline scheme, treatment scheme vectors are generated through the generator, the generated schemes are judged by the discriminator in combination with the guideline scheme, and parameters of the generator are optimized through countermeasure training, so that the output virtual treatment scheme meets clinical guideline requirements.
2. The neuroimmune disease immersive virtual patient diagnosis and treatment thinking training system according to claim 1, wherein the multi-modal interaction data comprises inquiry interaction data acquired through a voice acquisition device, query interaction data acquired through a motion capture device and examination selection data acquired through a VR interaction device, and the multi-modal interaction data is input into a dual-algorithm bidirectional fusion thinking anchor quantification method after standardized analysis.
3. The neuroimmune disease immersive virtual patient diagnosis and treatment thinking training system according to claim 1, wherein the evidence-based feedback and thinking correction module performs multi-dimensional comparison on virtual examination results, virtual treatment schemes and time sequence anchor point sequence vectors, positions anchor point deletion, time sequence errors and scheme deviation of the trainee diagnosis and treatment thinking, generates deviation correction basis in combination with evidence-based medical guideline content, and synchronously supplements guideline-based differential diagnosis strengthening gist.
4. The neuroimmune disease immersive virtual patient diagnosis and treatment thinking training system according to claim 1, wherein the diagnosis and treatment ability quantization evaluation module completes quantization scoring by adopting a three-dimensional evaluation model of thinking integrity, thinking logic and evidence-based compliance, and generates four-level difficulty adjustment instructions according to three-dimensional scoring results, wherein the difficulty adjustment instructions are specifically used for adjusting the number of case core anchor points, the complexity of pathophysiological dynamic evolution rules and the addition number of complications in a multi-dimensional case holographic modeling module.

Description

Neuroimmune disease immersive virtual patient diagnosis and treatment thinking training system Technical Field The invention relates to the crossing technical field of medical education technology, virtual Reality (VR) technology and Artificial Intelligence (AI) technology, in particular to a diagnosis and treatment thinking training system for a neuroimmune disease immersive virtual patient. Background The diagnosis and treatment process of the neuroimmune diseases (such as neuromyelitis optica and myasthenia gravis) is complex, and the clinical thinking and evidence-based decision making capability of doctors is extremely high. The existing diagnosis and treatment skill training technology has obvious limitations: the traditional training relies on static cases or entity simulators, only basic operation scenes can be simulated, the dynamic progress process of the illness state cannot be restored, and students cannot easily understand the association between diagnosis and treatment operation and illness state change; The training and evaluation of focusing is whether the operation is finished, quantitative evaluation of deep diagnosis and treatment thinking of students is lacked, deviation of the thinking level is difficult to find, and the training is heavy operation and light thinking; The existing AI auxiliary training system mostly adopts a single algorithm to generate an inspection result or a treatment scheme, does not combine with a evidence-based medical guideline to form strong constraint, and has the advantages that the generated content is easy to deviate from clinical practice and a bidirectional feedback optimization mechanism is not provided; the training process is not optimized in a closed loop, the case difficulty is fixed, and the training process cannot be dynamically adjusted according to the capability of students, so that the training efficiency is low, and the requirements of students with different levels are difficult to adapt. The problems cause poor clinical suitability of the existing training system, and students still have difficulty in rapidly adapting to real clinical diagnosis and treatment scenes after training, so that the diagnosis and treatment capability of the neuroimmune diseases can not be effectively improved. Disclosure of Invention The invention provides a nerve immune disease immersive virtual patient diagnosis and treatment thinking training system, which aims to solve the problems that 'scene statization, thinking assessment lack, AI scheme is separated from clinical pain points without closed-loop optimization' in the existing nerve immune disease diagnosis and treatment training, and specifically comprises the following steps: Breaking through the scene limitation of the traditional training, constructing a dynamic VR diagnosis and treatment environment close to the real clinic, and enabling students to intuitively feel the association of diagnosis and treatment operation and disease progress; the accurate capturing and quantitative evaluation of the deep diagnosis and treatment thinking of the students are realized, and the students are helped to find and correct the thinking deviation instead of only focusing on the operation surface; The evidence-based compliance of the AI generation content is improved, the judgment of the inspection result is ensured, the clinical guideline is met by the treatment scheme generation, and the clinical suitability of training is improved; a dynamic closed-loop optimization mechanism is constructed, training difficulty is adjusted according to the capability level of a student, the demands of the student in different stages are adapted, the diagnosis and treatment thinking and practical operation capability of the nerve immune disease is effectively improved, and the clinical qualified nerve immune disease diagnosis and treatment talents are provided. In order to achieve the above purpose, the invention adopts the following technical scheme: The neuroimmune disease immersive virtual patient diagnosis and treatment thinking training system comprises a multi-dimensional case holographic modeling module, a diagnosis and treatment thinking anchor point dynamic capturing module, a double AI collaborative diagnosis and treatment deduction module, a evidence-based feedback and thinking deviation correcting module and a diagnosis and treatment capacity quantification assessment module which are sequentially connected in data and form closed loop optimization; The multi-dimensional case holographic modeling module inputs authoritative clinical case data and evidence-based medical guideline content, a case core anchor point data set is generated through medical ontology modeling, a holographic VR case scene file with pathological physiological dynamic evolution rules is built by combining a medical image reconstruction technology, and the case core anchor point data set and the holographic VR case scene file are used as input of a di