A patient self-service medical report interpretation and triage method for a public hall scene
By integrating a local multimodal big data model and knowledge base into an independent integrated self-service terminal, the problem of interpreting and triaging medical reports in a lobby environment has been solved, achieving privacy security, easy interpretation, and highly reliable service, thus protecting patient data and life safety.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 聂凌虎
- Filing Date
- 2026-04-15
- Publication Date
- 2026-07-10
AI Technical Summary
In the public lobby of outpatient and emergency departments of large medical institutions, existing medical self-service equipment cannot provide interpretation of medical reports and triage guidance, and there are issues such as data privacy leakage, spying risks, and network instability.
It employs a local multimodal large model and a composite local medical knowledge base, and interprets and triages medical reports through physically isolated independent integrated self-service terminals. It utilizes transient sandboxes and one-way hardware design to ensure data security, and provides easy-to-understand interpretations and departmental suggestions in offline environments.
It achieves absolute privacy and security in a public lobby environment, provides easy-to-understand report interpretation, prevents data leakage and spying, and ensures highly reliable service stability and patient safety.
Smart Images
Figure CN122369772A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of smart healthcare and medical self-service terminal technology, and more specifically, to a method, system, electronic device, and computer-readable storage medium for patient self-service medical report interpretation and triage in public hall scenarios. Background Technology
[0002] In the current public lobby settings of large medical institutions' outpatient and emergency departments, patients often face the pain points of "not being able to understand the reports" and "long waiting times for follow-up appointments" after obtaining various medical reports from tests and examinations. Most existing medical self-service devices only provide report printing functions and cannot provide interpretation or triage guidance.
[0003] While some hospitals have attempted to introduce cloud-based AI-powered triage systems, these face numerous challenges in public lobby settings:
[0004] First, medical reports contain a large amount of highly sensitive patient personal health privacy data, and uploading such data to the cloud poses a risk of being intercepted by hackers or leaked by the cloud.
[0005] Secondly, the public hall is noisy and has a huge flow of people, with dense queues, making it easy for conventional interactive devices to be spied on by others.
[0006] Finally, in the event of network fluctuations or interruptions, the cloud-based large-scale model service will be directly paralyzed, and service stability under high concurrency cannot be guaranteed.
[0007] Therefore, there is an urgent need for a technical solution that can ensure data security in both the physical environment and the digital infrastructure, and provide easy-to-understand report interpretations and triage suggestions in real time in offline environments. Summary of the Invention
[0008] Purpose of the invention: In order to solve the above-mentioned technical problems, the present invention provides a method, system, equipment and medium for patient self-service medical report interpretation and triage in public hall scenarios.
[0009] Invention Method: This invention provides a self-service method for interpreting and triaging medical reports for patients in public hall settings. It is applied to a physically isolated, independent, integrated self-service terminal without an external network connection module. The method includes:
[0010] Step S1: Deploy a local multimodal large model and a composite local medical knowledge base within the self-service terminal.
[0011] Step S2: In response to the patient's physical interaction command, generate a single-session token valid only for the current operation cycle, and open a transient isolation sandbox bound to the token in local memory to reject any concurrent access requests.
[0012] Step S3: Obtain heterogeneous medical report data through physical peripherals and store it in the transient isolation sandbox for processing.
[0013] Step S4: Call the local multimodal large model to extract abnormal indicators and image features, and search in the composite local medical knowledge base.
[0014] Step S5: Perform parallel inference to generate simplified explanations of complex medical mechanisms and generate offline department recommendations for the hospital based on static routing rules.
[0015] Step S6: Display the interpreted text and suggested departments in one direction; when the current patient's session ends, immediately trigger a hardware-level clear command to completely erase all data and cache in the sandbox.
[0016] Invention System Solution: This invention also provides a system integrated into a single physical form of upright self-service equipment, comprising:
[0017] a. Offline computing power chassis module: The motherboard physically removes all Ethernet interfaces and wireless RF communication modules, and includes independent power supply and local AI acceleration chipset.
[0018] b. Single-player transient sandbox module: Creates a single-session token and dedicated memory space to ensure one-to-one privacy isolation.
[0019] c. Plain Language Dimensionality Reduction and Triage Reasoning Module: Integrates a large medical model and knowledge base, translates professional test results into popular science text and images, and outputs registration suggestions.
[0020] d. Person leaving and instant overwrite module: Links with environmental sensors to force a low-level memory overwrite when a person leaves or is replaced.
[0021] Beneficial effects:
[0022] 1. Absolute privacy and security protection: By physically isolating the system without external network modules, and combining the highest priority memory overwrite instructions and read-only file system, zero data residue is achieved in the public hall environment, fundamentally eliminating the risk of medical privacy being leaked by uploading to the cloud or disk.
[0023] 2. Easy-to-understand cognitive dimensionality reduction: The first knowledge base with a dimensionality reduction structure for patients' cognition is created, which transforms obscure medical standard codes and professional terms into lifestyle intervention guidelines and popular pathological metaphors, greatly reducing the reading threshold for patients.
[0024] 3. Highly reliable critical value circuit breaker mechanism: The unique critical value prevention and forced intervention mechanism forcibly suspends normal rendering when a fatal abnormal indicator is identified, provides a full-screen warning and gives the highest level of emergency advice, and protects the patient's life safety.
[0025] 4. Physical security design to prevent peeping: Equipped with anti-peeping barriers, directional anti-peeping touch screens and other hardware, it prevents visual theft while queuing, forming a comprehensive security loop from software to hardware. Attached Figure Description
[0026] Figure 1 : This is a schematic diagram of the overall process provided by the present invention.
[0027] Figure 2 : This is a block diagram of the module structure provided in an embodiment of the present invention. Detailed Implementation
[0028] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments.
[0029] Example 1: Interpretation and Triage Method Workflow
[0030] This embodiment demonstrates a self-service method for interpreting and triaging medical reports for patients in public lobby settings, deployed in a standalone integrated self-service terminal in the outpatient lobby. This terminal completely eliminates Wi-Fi, Bluetooth, and Ethernet interfaces at the hardware motherboard level, making it a pure "information island."
[0031] When a patient inserts a paper test report into the scanner slot, a change in the voltage level of the paper feed sensor triggers a single-session token. The system uses this token to define a transient isolation sandbox in memory, during which the screen does not respond to any other external requests. After the device scans the report (heterogeneous medical report data), it stores it in the sandbox, and the multimodal big data model quickly identifies abnormal indicators such as "triglycerides 5.8 mmol / L".
[0032] During the processing phase, the system invokes a comprehensive local medical knowledge base. This knowledge base includes not only standard code but also dimensionality reduction transformation models and a library of pathological mechanism analogies. Instead of outputting complex lipid metabolism disorder mechanisms, the system generates simplified text: "Your blood has too much 'oil,' like a thick layer of grease in your pipes. We recommend you consult a cardiologist or endocrinologist for further examination. Also, please pay attention to a light diet." This information is displayed to the patient via a targeted screen.
[0033] Example 2: Prevention of Missed Diagnoses and Secure Session Destruction Mechanism
[0034] To ensure patient safety, the method incorporates a mandatory intervention mechanism to prevent missed diagnoses of critical values. For example, if a patient scans a report related to myocardial infarction, the local large-scale model identifies "extremely high troponin levels," a critical value that meets the threshold for extremely abnormal physiological characteristics. At this point, the system immediately displays the aforementioned simplified explanation, the screen turns red, and a full-screen pop-up window displays an emergency alert, directly directing the user to the emergency department. Simultaneously, other menus on the screen are locked until the current operation is completed.
[0035] Once the patient has finished reading, the session end event may be triggered by three conditions: first, the patient actively clicks "End"; second, there is no operation on the screen within a preset time (e.g., 30 seconds); third, the person detection sensor at the top determines that the patient has moved out of the safe distance or has been replaced by another person. At this point, the system sends a Memory Overwrite command to the AI chip with the highest priority, clearing the image and KV cache. Because the underlying configuration is in Read-Only mode, no medical data can be written to the solid-state drive, completely achieving "data cleared when the person leaves."
[0036] Example 3: System and Hardware Devices
[0037] This invention also discloses a system and an electronic device. The system is encapsulated in a single, physically-formed vertical cabinet. The hardware features specially designed physical privacy peripherals, including a report scanning slot with a physical privacy shutter and a directional privacy touchscreen visible only directly in front of the operator (within ±30 degrees of viewing angle) to prevent peeping from the side or rear. The device internally includes a processor, memory, and computer program, enabling the machine to independently execute the aforementioned method features. Furthermore, this invention also covers a non-transitory computer-readable storage medium storing the method.
Claims
1. A method for patient self-service medical report interpretation and triage in public lobby scenarios, characterized in that, The method for a stand-alone, integrated self-service terminal that is physically isolated and does not have an external network connection module includes the following steps: Step S1: Deploy a local multimodal large model and a composite local medical knowledge base within the self-service terminal; Step S2: In response to the patient's physical interaction command, the self-service terminal generates a single-session token that is valid only for the current operation cycle, and opens a transient isolation sandbox bound to the token in local memory to reject any concurrent access requests; Step S3: Obtain the heterogeneous medical report data of the current patient through physical peripherals and store it in the transient isolation sandbox for processing; Step S4: Call the local multimodal large model to extract abnormal indicators and image features from the heterogeneous medical report data, and perform matching retrieval in the composite local medical knowledge base; Step S5, perform parallel reasoning: generate simplified explanations of complex medical mechanisms based on the search results, and generate offline consultation department suggestions for the hospital based on static routing rules; Step S6: The interpreted text and suggested departments are displayed unidirectionally through the interactive screen; when the current patient's session ends, a hardware-level clear command is immediately triggered to completely erase all session data and inference cache in the transient isolation sandbox.
2. The method according to claim 1, characterized in that, The "patient session end event" in step S6 is triggered by a joint hardware and software monitoring mechanism and includes any of the following conditions: Condition A: A read-to-end instruction is received from the interactive screen; Condition B: The time without operation of peripheral input or screen touch exceeds the preset timeout threshold; Condition C: The human detection sensor deployed on the self-service terminal determines that the current operator has left the preset physical safety distance in front of the machine, or that the operator's facial features are alternating.
3. The method according to claim 2, characterized in that, After the session end event is triggered, the specific execution logic of the hardware-level clear instruction is as follows: the operating system of the self-service terminal sends a memory overwrite instruction to the artificial intelligence acceleration chipset with the highest priority to clear the key-value cache (KV cache) of the multimodal input image and the inference process; at the same time, based on the underlying file system configured as read-only mode, it blocks any physical path for writing medical privacy data to the local solid-state drive, thereby achieving zero data residue in the public environment of the lobby.
4. The method according to claim 1, characterized in that, The composite local medical knowledge base in step S1 adopts a dimensionality reduction structure oriented towards patient cognition, specifically including: the bottom layer contains standard medical terminology and laboratory standard codes; the middle layer contains a dimensionality reduction transformation model that maps the standard medical terminology to natural language; and the top layer contains lifestyle intervention guidelines for specific abnormal indicators and a popularized metaphor library of pathological mechanisms.
5. The method according to claim 1, characterized in that, The trigger instruction for the single session token in step S2 originates from the level signal change of the physical peripheral component, including the trigger signal of the paper report scanning paper feed sensor or the reading signal of the physical slot, thereby ensuring that the terminal is strictly locked in a single-person, single-thread service mode in the physical space dimension.
6. The method according to claim 1, characterized in that, The method further includes a mandatory intervention mechanism to prevent missed diagnoses of critical values between steps S5 and S6: when the local multimodal large model identifies an extremely abnormal physiological indicator (such as a critical value of a myocardial ischemia marker) that meets a preset fatal characteristic threshold, the terminal operating system suspends normal popular interpretation text rendering, forcibly outputs an emergency status reminder with a full-screen eye-catching visual sign on the interactive screen, and directly generates the highest-level triage suggestion pointing to the emergency department or resuscitation room, while blocking all other interactive functions of the terminal until the session end event is triggered.
7. A self-service medical report interpretation and triage system for patients in public lobby settings, characterized in that, Integrated into a single physical form of upright self-service equipment, including: a. Offline computing power chassis module: includes independent power supply components and local artificial intelligence acceleration chipset, and its motherboard is physically stripped of all Ethernet interfaces and wireless radio frequency communication modules; b. Single-person transient sandbox module: used to create a single session token and dedicated memory space when a patient scans a report, shielding concurrent requests and ensuring one-to-one privacy isolation; c. Plain Language Reduction and Triage Reasoning Module: Integrates a medical language model and a local medical knowledge base, specifically designed to translate professional medical test results into easy-to-read popular science text and graphics, and outputs registration suggestions for specific departments based on local rules; d. Person leaving monitoring and instant overwrite module: In conjunction with the environmental perception sensor, when the signal of the patient leaving or being replaced is captured, the underlying memory overwrite operation is forcibly executed.
8. The system according to claim 7, characterized in that, The system is also equipped with physical privacy peripheral components adapted to the lobby environment; the components include a report scanning slot with a physical privacy shield and a paper feed sensor, and a directional privacy touch screen whose viewing angle is strictly limited to the front of the operator to prevent visual theft of sensitive medical information by people queuing behind or to the side.
9. A standalone integrated electronic device, characterized in that, include: a. At least one processor; b. A memory communicatively connected to the at least one processor; c. Display components for human-computer interaction, image acquisition peripherals for reading reports, and environmental perception sensors; The memory stores a computer program that can be executed by the at least one processor, which enables the standalone integrated electronic device to independently execute the method as described in any one of claims 1 to 6.
10. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions for causing a computer to execute the patient self-service medical report interpretation and triage method in a public hall setting as described in any one of claims 1 to 6.