CN-122000046-A - Abnormal physical sign AI early warning system and method for helicopter airborne medical subsystem

CN122000046ACN 122000046 ACN122000046 ACN 122000046ACN-122000046-A

Abstract

The application discloses an AI early warning system and method for abnormal signs of an airborne medical subsystem of a helicopter, wherein the method comprises the steps of receiving vital sign data from an airborne multisource physiological sensor in real time; the method comprises the steps of carrying out real-time signal quality evaluation and self-adaptive enhancement processing on vital sign data, inputting the processed data into a pre-trained lightweight anomaly detection AI model, carrying out multi-parameter fusion analysis to obtain anomaly identification and classification results, generating structured collaborative early warning information when the critical grade indicated by the anomaly identification and classification results exceeds a threshold value, wherein the collaborative early warning information at least comprises anomaly types, grades and data evidences, and simultaneously sending the collaborative early warning information to an auxiliary agent of medical staff and an agent of a cloud command center through a communication interface. From the discovery of the abnormality to the completion of the multiple pre-warning for <1 second, precious treatment time is striven for medical staff. The application greatly improves the early warning accuracy, multiplies the cooperative efficiency of treatment and has extremely strong system robustness.

Inventors

HUANG YUHONG
YU XIAOPIN
HUANG JINYAO
TANG JUNJIE
ZHONG JIN
YAN HONGCHANG

Assignees

中船海神医疗科技有限公司

Dates

Publication Date: 20260508
Application Date: 20251229

Claims (10)

1. An abnormal sign AI warning system for an onboard medical subsystem of a helicopter, comprising: The multisource physiological sensor is used for collecting vital sign data of wounded persons, at least comprising electrocardio, blood pressure and blood oxygen, in real time; The edge AI early warning agent is disposed on the helicopter, and the communication connection multisource physiological sensor, the edge AI early warning agent includes: The self-adaptive signal processing module is used for carrying out enhancement processing based on signal quality evaluation on the received vital sign data; The lightweight anomaly detection AI model is used for carrying out multi-parameter fusion analysis on the processed vital sign data, identifying and grading anomaly types and generating an early warning result; the local alarm is used for triggering a local audible and visual alarm according to the early warning result; the multi-agent cooperative network is connected with the edge AI early warning agent through a cooperative communication interface and at least comprises: The cloud command center intelligent agent is used for receiving and fusing early warning information from at least one edge AI early warning intelligent agent and carrying out global situation analysis and secondary verification; The medical staff auxiliary intelligent body is deployed in the terminal equipment and is used for receiving and displaying the early warning result and the associated treatment advice; The edge AI early warning agent is configured to synchronously trigger the local alarm of the local alarm within 1 second after the lightweight abnormality detection AI model identifies an abnormality reaching a preset level, and send cooperative early warning information to at least the medical staff auxiliary agent and the cloud command center agent in the multi-agent cooperative network through the cooperative communication interface.
2. The system of claim 1, wherein the edge AI-warning agent further comprises a local collaborative decision module configured to: According to the type, the level and the confidence coefficient of the early warning result, combining a preset rule base or strategy to dynamically generate the content and the sending target of the cooperative early warning information; When the communication interruption with the cloud command center agent is detected, switching to a degradation autonomous mode, enhancing decision logic based on a local knowledge base, and maintaining cooperation with the medical staff auxiliary agent.
3. The system of claim 1 or 2, wherein the adaptive signal processing module is configured to: Calculating a signal quality index of input vital sign data; Dynamically selecting or adjusting a signal filtering and enhancing algorithm according to the signal quality index; when the signal quality index is below a first threshold, an adjustment of an inference parameter or sensitivity of the lightweight anomaly detection AI model is triggered.
4. The system of claim 1, wherein the lightweight anomaly detection AI model is a knowledge-distilled lightweight visual transducer model, input as a time-spectrum two-dimensional image representation of multi-parameter vital sign time-series data, output as structured data comprising at least one anomaly type tag and its probability, and comprehensive criticality level.
5. The system of claim 1, wherein the collaborative early warning information is a structured data message including at least an anomaly type, a critical grade, a timestamp, a key vital sign data segment that triggers the early warning, and a preliminary treatment recommendation generated by the edge AI early warning agent or the cloud command center agent.
6. The system of claim 1, wherein the multi-agent collaboration network further comprises a hospital receiving agent deployed at a target hospital, the cloud command center agent further configured to send a resource preparation instruction to the hospital receiving agent based on the abnormal type, level, and expected arrival time of the wounded after receiving the collaboration pre-warning information.
7. The system of claim 1, wherein the healthcare worker auxiliary agent is configured to, upon receiving the collaborative pre-alarm information, hierarchically display the pre-alarm results, associated vital sign trend graphs, and dynamic treatment advice listings in an interactive interface and receive actual treatment operation records fed back by the healthcare worker.
8. An edge AI-warning agent, comprising: an input interface for receiving real-time vital sign flow data from a multi-source physiological sensor; a processor configured to perform: performing adaptive signal enhancement and quality assessment on the vital sign stream data; Performing anomaly detection and classification on the enhanced data by utilizing a pre-trained lightweight AI model; if an abnormality reaching a predetermined critical level is detected, generating a control instruction to trigger a local alarm within 1 second; generating a cooperative early warning message containing abnormal details and evidence; and the output interface is used for synchronously outputting the control signal of the local alarm and sending the cooperative early warning message.
9. An AI early warning method for physical sign abnormality of an onboard medical subsystem of a helicopter is characterized in that, The method is executed by an edge AI early warning agent and comprises the steps of receiving vital sign data from an airborne multi-source physiological sensor in real time, and carrying out real-time signal quality evaluation and self-adaptive enhancement processing on the vital sign data; inputting the processed data into a pre-trained lightweight anomaly detection AI model, and performing multi-parameter fusion analysis to obtain anomaly identification and classification results; When the critical level indicated by the abnormal identification and grading result exceeds a threshold value, the following parallel steps are executed: triggering a local audible and visual alarm device to finish local alarm; and the cooperative early warning information is simultaneously sent to the auxiliary intelligent agent of medical staff and the intelligent agent of the cloud command center through a communication interface.
10. The method as recited in claim 9, further comprising: Monitoring the state of a communication link with the intelligent agent of the cloud command center; when the communication link is interrupted, a local degradation decision strategy is started, the abnormality identification and grading result is subjected to complementary analysis and treatment suggestion generation based on a local clinical knowledge base, and information interaction with the medical staff auxiliary agent is maintained.

Description

Abnormal physical sign AI early warning system and method for helicopter airborne medical subsystem Technical Field The invention relates to the technical field of aviation medical first aid and artificial intelligence intersection, in particular to an airborne medical subsystem which is deployed on a helicopter platform and can monitor vital signs of wounded in real time, intelligently analyze, quickly early warn and cooperatively treat based on a multi-agent cooperative architecture. Background In helicopter medical rescue (HEMS), continuous monitoring and early warning of vital signs of critically ill individuals are key to reducing mortality during transit. Existing on-board medical equipment (such as monitors) usually operate independently and only have a threshold alarming function. The alarm information stays in the equipment local or needs visual discovery of medical staff, the early warning delay often exceeds 5-10 seconds, and the early warning delay cannot be synchronized with a remote command platform in real time. Conventional threshold alarms fail to identify complex, hidden trends in physiological deterioration (e.g., subtle ST-segment changes in the heart's electricity, compensatory changes in blood pressure). Early and complex anomaly recognition capabilities based on pathophysiological models for multiparameter fusion are lacking. After the alarm is triggered, the response is completely dependent on the personal experience and immediate judgment of the on-site medical staff. Lack of automated, structured information collaboration and decision support with other systems on board the aircraft (e.g., navigation, communication), ground command centers, and targeted hospital rescue teams, valuable "gold rescue time" is consumed in communication coordination. The helicopter cabin has strong vibration, large noise and complex electromagnetic environment, the signal quality of the sensor is easy to be interfered, the false alarm rate of the existing algorithm is high, and the alarm fatigue of medical staff is easy to be caused. Disclosure of Invention The technical problems to be solved by the invention mainly include how to solve the main technical problems of the medical rescue of the existing helicopter. In a first aspect, an embodiment of the present invention provides an AI early warning system for abnormal signs of an onboard medical subsystem of a helicopter, including: The multisource physiological sensor is used for collecting vital sign data of wounded persons, at least comprising electrocardio, blood pressure and blood oxygen, in real time; The edge AI early warning agent is disposed on the helicopter, and the communication connection multisource physiological sensor, the edge AI early warning agent includes: The self-adaptive signal processing module is used for carrying out enhancement processing based on signal quality evaluation on the received vital sign data; The lightweight anomaly detection AI model is used for carrying out multi-parameter fusion analysis on the processed vital sign data, identifying and grading anomaly types and generating an early warning result; the local alarm is used for triggering a local audible and visual alarm according to the early warning result; the multi-agent cooperative network is connected with the edge AI early warning agent through a cooperative communication interface and at least comprises: The cloud command center intelligent agent is used for receiving and fusing early warning information from at least one edge AI early warning intelligent agent and carrying out global situation analysis and secondary verification; The medical staff auxiliary intelligent body is deployed in the terminal equipment and is used for receiving and displaying the early warning result and the associated treatment advice; The edge AI early warning agent is configured to synchronously trigger the local alarm of the local alarm within 1 second after the lightweight abnormality detection AI model identifies an abnormality reaching a preset level, and send cooperative early warning information to at least the medical staff auxiliary agent and the cloud command center agent in the multi-agent cooperative network through the cooperative communication interface. In a specific embodiment of the present invention, the edge AI-warning agent further includes a local collaborative decision module configured to: According to the type, the level and the confidence coefficient of the early warning result, combining a preset rule base or strategy to dynamically generate the content and the sending target of the cooperative early warning information; When the communication interruption with the cloud command center agent is detected, switching to a degradation autonomous mode, enhancing decision logic based on a local knowledge base, and maintaining cooperation with the medical staff auxiliary agent. In a specific embodiment of the present invention, the adaptive signal processing