Puncture point bleeding real-time monitoring system and method based on multi-modal sensing

Abstract

The invention provides a puncture point bleeding real-time monitoring system and a puncture point bleeding real-time monitoring method based on multi-modal sensing, which are used for specially building a set of novel puncture point bleeding real-time monitoring architecture, and compared with the limitation of single signal detection, the high-performance puncture point bleeding real-time monitoring effect is realized by fusing multi-dimensional physical parameters, so that the puncture point bleeding real-time monitoring system and the puncture point bleeding real-time monitoring method based on multi-modal sensing are provided. And the monitoring performance is continuously enhanced by matching corresponding optimization settings in many details, so that a real-time, high-precision, high-processing-efficiency and light-weight high-quality puncture point bleeding real-time monitoring service is provided, better tool support is realized for puncture point bleeding real-time monitoring work, and a good decision-making assisting effect is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical treatment, in particular to a puncture point bleeding real-time monitoring system and method based on multi-modal sensing. BACKGROUND

[0002] Current interventional operations, such as percutaneous coronary intervention (PCI), transjugular intrahepatic portosystemic shun (TIPS), etc., involve puncture operations, and postoperative bleeding monitoring of the puncture point relies on traditional gauze compression bandaging combined with manual observation method.

[0003] However, the present inventors have found that the traditional gauze compression bandaging combined with manual observation method has corresponding defects. When medical staff observe that the gauze is penetrated by blood with the naked eye, the actual amount of bleeding is often more than 50-100 ml. At this time, the patient may have been in a decompensation state, and even serious complications may occur. In a clinical study on bleeding after PCI, the active discovery rate of bleeding in the control group without using intelligent monitoring equipment was only 30%, and the average amount of bleeding was as high as 118-156 ml. SUMMARY

[0004] The present application provides a puncture point bleeding real-time monitoring system and method based on multi-modal sensing, which is used to specially build a novel puncture point bleeding real-time monitoring architecture. Compared with the limitations of single signal detection, high-performance puncture point bleeding real-time monitoring effect is achieved by fusing multi-dimensional physical parameters, and the monitoring performance is further enhanced through corresponding optimization settings in many details. Thus, real-time, high-precision, high-processing-efficiency and lightweight high-quality puncture point bleeding real-time monitoring services are provided, which realizes better tool support for puncture point bleeding real-time monitoring work and plays a good decision-making assistance role.

[0005] In a first aspect, the present application provides a puncture point bleeding real-time monitoring system based on multi-modal sensing. The puncture point bleeding real-time monitoring system based on multi-modal sensing includes a monitoring terminal and a data processing platform. The monitoring terminal with a bandage structure is configured at the puncture point area on the patient's skin. The multi-modal sensors in the monitoring terminal include conductivity sensors, temperature sensors and humidity sensors. The monitoring terminal adopts a detachable battery. The multi-modal sensors are arranged in multiple groups along the radial direction of the puncture point. In the working process of the puncture point bleeding real-time monitoring system based on multi-modal sensing, the following processing contents are included: The monitoring terminal collects conductivity data, temperature data and humidity data through the conductivity sensor, the temperature sensor and the humidity sensor respectively, and transmits the conductivity data, the temperature data and the humidity data to the data processing center through the communication module. The data processing center performs corresponding real-time bleeding risk assessment processing based on the multi-modal data formed by the conductivity data, the temperature data and the humidity data, and performs hierarchical early warning output based on the real-time bleeding risk assessment result.

[0006] In a second aspect, the application provides a puncture point bleeding real-time monitoring method based on multi-modal sensing. The puncture point bleeding real-time monitoring method based on multi-modal sensing is applied to a puncture point bleeding real-time monitoring system based on multi-modal sensing. The puncture point bleeding real-time monitoring system based on multi-modal sensing includes a monitoring terminal and a data processing center. The monitoring terminal with a bandage structure is configured at a puncture point area on a patient's skin. The multi-modal sensors in the monitoring terminal include a conductivity sensor, a temperature sensor and a humidity sensor. The monitoring terminal uses a detachable battery. The multi-modal sensors are arranged in multiple groups along the radial direction of the puncture point. The method includes the following steps: The monitoring terminal collects conductivity data, temperature data and humidity data through the conductivity sensor, the temperature sensor and the humidity sensor respectively. The monitoring terminal transmits the conductivity data, the temperature data and the humidity data to the data processing center through the communication module. The data processing center performs corresponding real-time bleeding risk assessment processing based on the multi-modal data formed by the conductivity data, the temperature data and the humidity data. The data processing center performs hierarchical early warning output based on the real-time bleeding risk assessment result.

[0007] In a third aspect, the application provides a computer readable storage medium. The computer readable storage medium stores a plurality of instructions. The instructions are suitable for being loaded by a processor to execute the method provided in the second aspect of the application.

[0008] From the above, the application has the following beneficial effects: For the puncture point bleeding real-time monitoring target, the application specially builds a novel puncture point bleeding real-time monitoring architecture. Compared with the limitation of single signal detection, the application realizes high-performance puncture point bleeding real-time monitoring effect by fusing multi-dimensional physical parameters, and further enhances the monitoring performance through corresponding optimization settings in many details. Thus, the application provides real-time, high-precision, high-processing-efficiency and lightweight high-quality puncture point bleeding real-time monitoring service, and realizes better tool support for puncture point bleeding real-time monitoring work, and plays a good decision-making assistance role. BRIEF DESCRIPTION OF DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments description. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.

[0010] Figure 1 A system architecture schematic diagram of a puncture point bleeding real-time monitoring system based on multi-modal sensing of the present application; Figure 2 A flowchart schematic diagram of a puncture point bleeding real-time monitoring method based on multi-modal sensing of the present application. DETAILED DESCRIPTION

[0011] The technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings of the embodiments of the present application. Obviously, the described embodiments only constitute some embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0012] The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or modules does not necessarily have to be limited to those steps or modules clearly listed, but can include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices. The naming or numbering of the steps appearing in the present application does not mean that the steps in the method flow must be executed in the time / logical order indicated by the naming or numbering, and the flow steps that have been named or numbered can change the execution order according to the technical purpose to be achieved, as long as the same or similar technical effects can be achieved.

[0013] The division of the modules appearing in the present application is a logical division, and in actual application, there can be another division manner, for example, multiple modules can be combined or integrated in another system, or some features can be ignored or not executed, in addition, the coupling or direct coupling or communication connection between the modules shown or discussed can be through some interfaces, the indirect coupling or communication connection between the modules can be electrical or other similar forms, which are not limited in the present application. Moreover, the modules or sub-modules described as separate components can or can not be physically separated, can or can not be physical modules, or can be distributed in multiple circuit modules, and part or all of the modules can be selected according to actual needs to achieve the purpose of the present application scheme.

[0014] First, referring to Figure 1 The present application shows a system architecture schematic diagram of the puncture point bleeding real-time monitoring system based on multi-modal sensing, from Figure 1 It can be clearly seen from the above that the puncture point bleeding real-time monitoring system based on multi-modal sensing designed for the puncture point bleeding real-time monitoring target can specifically include two parts of a monitoring terminal and a data processing platform.

[0015] As an example, the puncture point bleeding real-time monitoring system based on multi-modal sensing can be configured with the standard nursing equipment for minimally invasive surgery involving puncture operation in clinical work in actual application, which is suitable for all arterial and venous puncture interventional procedures, and can specifically involve percutaneous coronary intervention, transjugular intrahepatic portosystemic shunt, hemodialysis, cardiovascular intervention, etc.

[0016] It should be noted that the monitoring terminal and the data processing platform are distinguished in terms of whether they are configured on the patient's body, not that there are two devices.

[0017] In the hospital application scenario, puncture point bleeding real-time monitoring is often required for multiple patients, so multiple monitoring terminals that can simultaneously provide puncture point bleeding real-time monitoring for multiple patients need to be configured, in addition, for the same patient, there can be a need for bleeding real-time monitoring at different positions, i.e., different puncture points, which can also be accompanied by the configuration of multiple monitoring terminals on the same patient.

[0018] In terms of specific implementation, it can be understood that the monitoring terminal is used to configure the area where the puncture point on the skin of the patient is located, and the monitoring terminal adopts a bandage structure design, so that the monitoring terminal with the bandage structure design can be conveniently configured on the skin area where the puncture point is located. The body / effective sensing part of the monitoring terminal can cover the puncture point or be fixed beside the puncture point, that is, within a certain distance from the puncture point, which can be adaptively configured according to actual needs.

[0019] The multi-modal sensor in the monitoring terminal can specifically include an electrical conductivity sensor, a temperature sensor, and a humidity sensor, which corresponds to the case that subsequent data processing involves three types of sensing data / signals, i.e., electrical conductivity data (conductivity and impedance are inversely proportional to each other), temperature data, and humidity data.

[0020] The monitoring terminal specifically adopts a detachable battery, which is convenient to replace and avoids the trouble of charging.

[0021] As an example, the monitoring terminal can be configured to have a size of ≤3×3 cm and a weight of <10 g, be fixed beside the puncture point by a medical adhesive, use an STM32L4 low-power MCU + LoRa radio frequency module (transmission distance 1 km, full coverage of the ward) as the core chip, and use a detachable button battery (such as CR2032) with a battery life of ≥14 days.

[0022] As can be seen from the example, in the wireless transmission scheme other than the wired transmission scheme, the LoRa communication module can be specifically selected, of course, the Bluetooth communication module, the WIFI communication module, and other types of wireless communication modules can also be used. Wireless communication has more flexible characteristics than wired communication, and the LoRa radio frequency module can be more suitable for large-scale hospital field conditions due to its relatively large signal coverage range.

[0023] In addition, it can be understood that in actual situations, the monitoring terminal can also continue to be configured with other components in terms of hardware, such as reinforcement components and reminder components, according to work requirements. Since this is not the focus of the present application, it will not be described in more detail.

[0024] In the preferred sensor arrangement, the present application specifically configures: the multi-modal sensors are arranged in multiple groups along the radial direction of the puncture point.

[0025] It can be understood that compared with the single-sensor deployment scheme or the multi-modal sensor deployment scheme uniformly arranged in space, the present application introduces a scheme arrangement of arranging multi-modal sensors in layers along the radial direction of the puncture point. In this way, the sensing signal characteristics along the radial direction of the puncture point can be captured to more accurately perform multi-modal fusion and more accurately determine the bleeding situation of the puncture point.

[0026] The data processing platform is responsible for core data processing tasks and can also be called processing equipment or processing hub. It can also be configured with equipment clusters more flexibly according to different considerations of data processing tasks. For example, an equipment cluster consists of gateway nodes deployed in rooms / wards / hospitals and server equipment in the back-end laboratory. In this case, the gateway node usually plays the role of aggregating data from the on-site monitoring terminals and transmitting the data back to the server equipment for further processing.

[0027] In the specific data processing process, the data processing platform can involve both the invocation of corresponding data models and the application of corresponding artificial intelligence (AI) models. For the latter, machine learning models, especially deep learning models, are usually used. Understandably, AI models can achieve efficient and high-precision data processing tasks. Furthermore, due to the powerful learning capabilities of AI models, they can also help capture the mapping relationship between subtle signal features that are difficult to capture manually and the bleeding situation at the puncture point (which usually exists in a "black box" manner), thus achieving better prediction accuracy.

[0028] Next, in conjunction with further explanations of the monitoring terminal and data processing platform, we will provide a detailed description of the real-time monitoring of puncture site bleeding in this application's real-time puncture site bleeding monitoring system (which is usually carried out as a manual task in practice).

[0029] Specifically, the real-time monitoring system for puncture site bleeding based on multimodal sensing may include the following processing steps during operation: 1) The monitoring terminal collects conductivity data, temperature data, and humidity data through conductivity sensors, temperature sensors, and humidity sensors, respectively, and transmits them to the data processing platform through the communication module; Understandably, during normal operation, the monitoring terminal can trigger the conductivity sensor, temperature sensor, and humidity sensor to perform normal data / signal acquisition. The data collected from the three modes can then be transmitted to the data processing platform in real time or through a pre-built communication channel for further data processing to determine the current bleeding status at the puncture site.

[0030] In practice, the monitoring terminal can either enter normal working state by default after powering on, switch from sleep state (non-normal working state) to normal working state under manual triggering, or switch from sleep state to normal working state under the trigger signal of the data processing platform. This can be flexibly configured according to actual needs.

[0031] 2) The multi-modal data formed by the data processing middle platform based on the conductivity data, temperature data and humidity data is used for real-time bleeding risk assessment processing, and the real-time bleeding risk assessment result is used for grading early warning output.

[0032] It can be understood that in the specific real-time bleeding risk assessment processing, the application also carries out grading processing for specific multi-modal data and bleeding risk assessment, so that in actual operation, appropriate early warning and intervention can be conveniently and effectively promoted, and high-quality postoperative puncture point bleeding management of patients can be realized.

[0033] Thus, under the above setting, it can be seen that for the puncture point bleeding real-time monitoring target, the application specially builds a novel puncture point bleeding real-time monitoring architecture. Compared with the limitation of single signal detection, high-performance puncture point bleeding real-time monitoring effect is realized by fusing multi-dimensional physical parameters to provide real-time, high-precision, high-processing-efficiency and lightweight high-quality puncture point bleeding real-time monitoring service, which realizes better tool support for puncture point bleeding real-time monitoring work and plays a good decision-making auxiliary role.

[0034] Further, in details, the application also makes corresponding optimization settings in many details to continue to enhance the real-time monitoring performance of the system for puncture point bleeding.

[0035] Specifically, in an exemplary embodiment, the application introduces a sandwich structure flexible sensor mechanism to realize a good bionic sensing network.

[0036] Specifically, the multi-modal sensor deployed on the monitoring terminal introduced in the foregoing can adopt a three-layer flexible sensor structure design, including a bottom layer, a middle layer and a top layer. The bottom layer corresponding to the patient's skin side (that is, the inner side) is a hydrophobic and breathable film, mainly used to prevent blood cells from blocking, and has good skin friendliness, which is conducive to improving user comfort; The middle layer is implanted with a cross-finger type electrode and coated with a diamond-like carbon coating to reduce platelet adhesion, mainly used for data acquisition tasks (corresponding to conductivity, temperature and humidity) through the electrode array structure; The top layer corresponding to the outer side is a nano-porous PVDF film, mainly used for specific adsorption of plasma proteins to enhance signal sensitivity.

[0037] Among them, PVDF is Polyvinylidene Fluoride, diamond-like carbon coating is DLC, and the cross-finger type electrode can also be embedded with conductive fibers such as carbon fibers and metal fibers to enhance flexibility and stretchability, and also help to have high sensitivity sensing performance.

[0038] It should be noted that the setting of the multi-modal sensor in the monitoring terminal is specifically set so that the blood penetrating from the puncture point can penetrate the sensing area of the sensor, and at a subtle level, it is not said that the blood needs to penetrate the sensor to be able to sense the corresponding signal. After the blood penetrates from the puncture point, it will bring corresponding effects, which will be reflected from the corresponding signal characteristics that the multi-modal sensor can sense.

[0039] In addition, it can be understood that, corresponding to the safety consideration of biological compatibility and reuse, in addition to the above-mentioned hydrophobic and breathable film, diamond-like carbon coating setting, disposable degradable sensor module (cost < ￥50) or the introduction of low-temperature plasma sterilization and other settings / operations can also be used to ensure better safety and avoid the deposition of sensor surface proteins caused by blood contact, which in turn brings adverse effects on detection accuracy for multiple times of use.

[0040] In addition, in the design of the specific sensor module of the monitoring terminal, the present application considers that multi-module integration may enhance the problem of device volume, and in details, it can also involve modular design of separating the core processing unit from the sensing unit (the latter can be configured as a disposable patch), and the use of flexible printed electronic technology to compress the thickness of the corresponding circuit to 0.1mm.

[0041] Further, the present application also configures a three-level early warning mechanism. Correspondingly, as another exemplary embodiment, the real-time bleeding risk assessment processing carried out by the data processing center can specifically follow the following strategy content: Low-risk warning level (also can be recorded as Level 1), the bleeding amount calculated by the multi-modal data is <5ml / h, the humidity data rising rate is ≤0.5% / min, the warning mode of the mobile phone App warning, and the clinical processing suggestion of corresponding intensive observation + 30 minutes recheck; Medium-risk warning level (also can be recorded as Level 2), the bleeding amount calculated by the multi-modal data is 5-50ml / h, the conductivity data rising rate is >15% (corresponding to impedance drop rate >15%), the warning mode of corresponding nurse station sound and light alarm, and the clinical processing suggestion of corresponding adjustment of compression band + checking coagulation index; High-risk warning level (also can be recorded as Level 3), the bleeding amount calculated by the multi-modal data is >50ml / h, the temperature data rising rate is >1°C / min, the warning mode of corresponding monitoring terminal high-frequency buzzer warning + compression air bag automatic inflation, and the clinical processing suggestion of corresponding emergency bedside treatment + preparation of hemostatic drugs or preparation of hemostatic surgery.

[0042] It can be seen that the specific puncture point bleeding condition grading setting in the above strategy content also corresponds to the related human-computer interaction design, which can involve the patient end and the medical staff end.

[0043] Among them, at the patient end, the mobile phone App (Application, application) early warning is based on the related application program (client, web service, public number, applet, etc.) of the patient himself or the patient's family, such as vibration reminder + voice pacification (or voice reminder).

[0044] At the medical staff end, the nurse station can specifically involve different hardware devices such as all-in-one machine, desktop computer, personal digital assistant (PDA), sound and light alarm lamp that can do human-computer interaction or reminder output, and the doctor side is the same, in addition, in some positions of the hospital area, such as corridor, toilet, etc., the corresponding devices for human-computer interaction or reminder output can also be deployed to output reminders.

[0045] Among them, it should be noted that in the above embodiment, the early warning method and the clinical treatment suggestion are involved, both of which can be included in the real-time bleeding risk assessment result, or the subsequent graded early warning output corresponding to the real-time bleeding risk assessment result, which is more flexible.

[0046] From the embodiment here, it can be more intuitively seen that under the puncture point bleeding real-time monitoring architecture designed based on the multi-modal in the present application, the effective identification of slow oozing (<5ml / h), active bleeding (>50ml / h) and puncture point bleeding between the two can be realized, while the traditional method cannot issue an early warning in the initial stage of bleeding (<10ml), when the oozing is visible to the naked eye, the actual bleeding volume often exceeds 50ml, which delays the best treatment opportunity. Compared with the lag problem of the traditional monitoring, the present application can effectively realize the detection of low-bleeding latent bleeding (sensitivity of 1-3ml), advance the early warning time to the golden intervention window (within 5 minutes after bleeding), and then realize precise intervention guidance.

[0047] And this situation, for the original dependence on artificial night patrol work, can play a better guarantee effect, it can be understood that on the basis of the limitation of nurse experience level and patrol interval in artificial observation, the bleeding discovery delay rate of night nursing manpower weak period is as high as 70%, and the present application can realize unattended monitoring, and through multi-level alarm (mobile phone App→nurse station→sound and light alarm), covering the whole period of guarantee, avoiding the dependence on manpower and subjective bias.

[0048] In addition, on the medical staff side, the puncture point bleeding real-time monitoring situation of each monitoring terminal side can also be involved under normal circumstances, and the puncture point state heat map of the whole hospital, ward, etc. granularity displayed by the electronic display board can be used to reflect it, among which the normal state, low risk state and medium-high risk state can be identified by green, yellow and red, so as to play a good visualization on the whole level, and combined with the above-mentioned early warning output, to ensure automatic, real-time and accurate safety protection.

[0049] Further, as another exemplary embodiment, the real-time bleeding risk assessment processing made by the data processing platform can also follow the following strategy content: The blood conductivity ranges from 3-5 S / m, the tissue fluid conductivity ranges from 1-2 S / m, the fresh blood temperature is higher than the body surface by 2-3℃, and the sweat conductivity ranges from >10 S / m.

[0050] It can be understood that in specific operation, the present application introduces a conductivity-temperature cooperative mechanism, combines bioimpedance spectrum (10 kHz-100 kHz) and thin film thermocouple, uses the conductivity difference (blood conductivity 3-5 S / m, tissue fluid 1-2 S / m) and temperature change (fresh blood temperature is higher than body surface by 2-3℃) of blood and tissue fluid, to construct a double-parameter verification mechanism, so as to significantly reduce the false alarm rate and better make bleeding condition and bleeding amount estimation processing.

[0051] And in the case that the interference caused by sweat, disinfectant and other factors can be excluded by temperature and humidity, the present application also makes more delicate sweat-blood discrimination based on the strategy of following the sweat conductivity range >10 S / m.

[0052] Further, as another exemplary embodiment, the real-time bleeding risk assessment processing made by the data processing platform can also follow the following strategy content: On the basis of the multi-modal sensors arranged in multiple groups along the radial direction of the puncture point, the specific type of the current condition belonging to the local small amount of bleeding type and the diffuse active bleeding type is subdivided by analyzing the humidity diffusion direction and speed.

[0053] Among them, as an example, under the special design of the present application, 4 groups of multi-modal sensors can be arranged along the radial direction of the puncture point to realize a gradient processing mechanism with good balance in all aspects.

[0054] It can be understood that for the above-mentioned sensing signal features along the radial direction of the puncture point which are helpful for accurate multi-modal fusion, in this embodiment, the humidity diffusion direction and humidity diffusion speed are specifically involved, so as to further subdivide the bleeding conditions of the local small amount of bleeding type and the diffuse active bleeding type.

[0055] Meanwhile, corresponding to the multi-modal sensors arranged in multiple groups along the radial direction of the puncture point, the present application also considers different user parts where the puncture points are located, and can also consider different adaptive spacing mechanisms, that is, although they are arranged along the radial direction of the puncture point, the spacing between each group of multi-modal sensors is not the same, and it needs to be understood that the difference in spacing is not due to the difference in the user part where the puncture point is located, but the present application believes that the spacing between the multi-modal sensors of the puncture point of different user parts can be relatively close or even difficult to detect with the naked eye, and the purpose of the present application is to make subtle spacing adjustments of the multi-group multi-modal sensors involved in the puncture points of different user parts, so as to capture more delicate radial direction sensing signal characteristics along the puncture point.

[0056] Under this concept, the optimal multi-modal sensor deployment scheme adapted to the puncture points of different user parts can be searched in advance by a corresponding search algorithm, such as simulated annealing algorithm, whale optimization algorithm, etc., and the search target can be mainly based on monitoring accuracy, and other factors such as monitoring stability and monitoring efficiency.

[0057] Further, for the puncture points that can be involved in clinical work, the present application can also make corresponding classification processing, and for the same type of puncture point, the same optimal multi-modal sensor deployment scheme (mainly for sensor spacing) can be used.

[0058] Meanwhile, in specific operations, the present application can also consider some special cases to configure special optimal multi-modal sensor deployment schemes, for example, due to factors such as the current physical condition, the current diagnosis scheme, the current treatment scheme, the current regimen scheme, personal limb habits, etc., some patients may have puncture operations (corresponding to clinical processing involving puncture operations such as interventional surgery) different from the conventional / general puncture operation, and in this case, the present application believes that the general version of the puncture point bleeding real-time monitoring may also be disturbed, and for this, the present application can add samples or constraints with the above special cases to configure the corresponding optimal multi-modal sensor deployment scheme and its type, so as to better cope with rare awkward problems in actual situations and better play the monitoring performance of the present application for puncture point bleeding.

[0059] As an example, in the case of central area humidity rising sharply + peripheral area rising slowly, corresponding to the situation / type of arterial rupture bleeding; in the case of global synchronous rising, corresponding to the situation / type of coagulation dysfunction.

[0060] After the specific bleeding situation is determined, it is obvious that more delicate reminder output can be performed.

[0061] Further, as mentioned before, the data processing platform can involve the application of machine learning models, and in yet another exemplary embodiment, the real-time bleeding risk assessment processing performed by the data processing platform can be specifically processed by a processing model configured on the gateway node, and the processing model is specifically a machine learning model, which is trained in advance by a sample bleeding sample configured with corresponding labels.

[0062] In actual application, it is easy to understand that not all data processing involved in real-time bleeding risk assessment processing needs to be processed by machine learning models. The introduction of the embodiment machine learning model is to better perform complex or large data processing tasks. For some relatively simple data processing, such as mapping table, the corresponding real-time bleeding risk assessment result can be accurately obtained, and the machine learning model does not need to be called.

[0063] As an example, the three warning levels involved in the previous embodiment can be completed based on a mapping table, and the data processing input is bleeding volume, temperature rise rate, conductivity, and temperature rise rate. For complex bleeding volume data processing, a machine learning model can be used for corresponding prediction processing.

[0064] As an example, the MobileNetV3 model or its modified / variant model can be used.

[0065] In addition, if it is necessary to introduce a time sequence feature in the model prediction logic, that is, on the basis of the special attention paid to the sensing signal features along the radial direction of the puncture point in the present application, the time sequence feature is combined to make more accurate puncture point bleeding volume prediction work from the time dimension. As an example, the Long Short-Term Memory (LSTM) or its modified / variant network can also be considered.

[0066] In this way, through the intelligent AI model, the traditional postoperative puncture point bleeding volume blind area of the postoperative patient is effectively and high-qualityly overcome, and in combination with the accurate real-time bleeding risk assessment processing introduced before, the patient is better avoided from being in the complications of radial nerve injury, skin necrosis, and decompensation state in actual situations.

[0067] For the pre-training of the AI model, it can be understood that corresponding training samples need to be configured and labeled (labeled predicted true value) to expand the specific model training processing.

[0068] As an example, in actual operation, the hemodynamic parameters in the CHAMPION-HF test can be integrated with the post-PCI bleeding database to configure a specific training sample.

[0069] At the same time, it can be understood that the specific model type that can be adopted can be flexibly configured in a similar manner to the case, and the specific model training scheme adopted in the model training process and the specific loss function can also be flexibly configured according to actual needs, that is, both existing schemes can be adopted, and further optimization and improvement can be made on the basis of existing schemes, or even novel self-developed schemes can be adopted.

[0070] In addition, corresponding to the special cases mentioned earlier that may be encountered in actual application involving current physical conditions, current diagnosis schemes, current treatment schemes, current maintenance schemes, personal limb habits, etc., corresponding model inputs can also be configured in the training process of the model here, so that when the model training is completed, the model can also capture the impact of these special cases on the puncture point bleeding amount prediction work, to achieve more delicate and accurate prediction results.

[0071] In addition, it can be noted in the embodiments herein that the complex data processing tasks in the data processing platform side, such as the data processing tasks to be performed by the processing model, are specifically performed at the gateway node.

[0072] As an example, 80% of the signal analysis work can be completed at the gateway end, which can effectively reduce the cloud load.

[0073] In the design of the present application scheme, the gateway node is not necessarily a related device node (such as a professional gateway device that can be adopted) that is specifically responsible for the gateway service, but can also be an existing device with communication capability that is additionally configured with the gateway service required by the present application scheme, forming a gateway node configuration mechanism that is diversified and highly adapted to the on-site situation.

[0074] For example, the gateway node can specifically be a hardware device with corresponding data processing capability at the medical staff end, such as a nurse station, or can also be a hardware device of the patient himself or his family members, and install the corresponding application program of the present application scheme.

[0075] Compared with the former, the latter has obvious flexibility in networking, and can select or switch gateway nodes according to actual needs. It can be understood that in actual situations, it is also helpful for patients with stronger medical service quality needs / willingness to provide better equipment performance to promote better system monitoring performance. In the case of manual patrol, it plays a good monitoring resource supplement, and also makes home monitoring in special cases feasible.

[0076] Or, for the latter, it can better meet the special needs of some patients for data visualization or information security. It can be understood that the patient side can flexibly configure the device providing a friendly visualization interface according to real-time needs, and can also store relevant data, which can effectively promote a better medical environment and doctor-patient relationship in today's medical service work, and to some extent, it is also helpful to help the puncture point bleeding real-time monitoring work to be more smoothly and high quality.

[0077] As an example, the gateway node can also involve the application of a natural language model, so that in addition to the corresponding content of the basic puncture point bleeding real-time monitoring result, the processing logic (more colloquially, the reasoning process) in the model processing process can also be given, thereby having better visualization, allowing the patient side to more intuitively understand the specific situation and meet more delicate and high-quality user visualization needs.

[0078] In addition, as another exemplary embodiment, for the index threshold involved in the real-time bleeding risk assessment processing, it can also be adaptively adjusted according to the patient's body obesity condition and the patient's posture condition during monitoring, wherein the patient's body obesity condition is specifically modeled by pre-CT data or MRI data, and the patient's posture condition is specifically monitored by a posture sensor; The sensor sensitivity is also adaptively adjusted according to the patient's body obesity condition.

[0079] It should be noted that the index threshold that can be adaptively adjusted here does not involve the model parameters that are autonomously iteratively optimized in the model training process, for example, it can act on the related index thresholds involved in the three-level early warning mechanism in the previous embodiment.

[0080] It can be understood that in terms of details, the present application believes that the subcutaneous tissue thickness of obese patients may also affect signal transmission, and sudden changes in the patient's posture may cause the sensor to shift and detect abnormally. Therefore, through the pre-modeling of the patient's body obesity condition and the real-time monitoring of the patient's posture condition, personalized calibration can be performed to improve the anti-interference ability.

[0081] The posture sensor can specifically refer to a three-axis accelerometer, a gyroscope, or the like. For the same obesity problem, different index thresholds can be adapted to different obesity conditions, for example, the body mass index (BMI) can be directly divided, and the index threshold of the standard mode is adopted when BMI < 25, and the index threshold of the low sensitivity mode is adopted when BMI ≥ 25.

[0082] In addition, corresponding to the obesity condition of the patient's body, the sensitivity of the corresponding sensor can also be adjusted in stages, so as to more accurately collect signals.

[0083] At the same time, the present application also considers the problem of compression complication prevention. It can be understood that excessive compression can cause limb ischemia or nerve damage, and insufficient compression can also cause hemostasis failure. Therefore, a distributed pressure sensor array can be integrated to promote dynamic optimization of compression force and avoid iatrogenic injury.

[0084] Correspondingly, as another exemplary embodiment, the multi-modal sensor can also include a distributed pressure sensor array. When the local pressure exceeds the ischemia threshold or is lower than the hemostasis failure threshold, the data processing center generates a compression force heat map based on the pressure data, and the data processing center automatically adjusts the tightness and prompts.

[0085] The pressure sensor array can specifically use a piezoresistive array sensor (16-point The automatic adjustment prompt can use the hierarchical warning output mentioned earlier. The pressure sensor array can be configured inside the bandage and does not necessarily need to be configured at the same position as the conductivity sensor, the temperature sensor, and the humidity sensor mentioned earlier.

[0086] As an example, the ischemia threshold can be 40 mmHg, and the hemostasis failure threshold can be 20 mmHg.

[0087] In actual cases, it can be understood that accurate pressure monitoring and adjustment can reduce compression-related complications by 40%.

[0088] Further, it can be understood that pressure monitoring is usually used as a measure to adjust the compression degree of the puncture point and its surrounding area to reasonably avoid ischemia and hemostasis failure. When considering how to better monitor the real-time bleeding situation of the puncture point, the present application also believes that, in some cases, the pressure situation can also be an effective decision factor for real-time bleeding risk or bleeding amount.

[0089] For the former, it can be added to the real-time bleeding risk assessment processing strategy based on the estimated bleeding volume and a series of decision factors. For the latter, it can be added to the processing logic of the processing model for making bleeding volume prediction. Thus, in actual cases, the pressure monitoring situation can be combined to make more delicate and accurate bleeding volume prediction.

[0090] In particular, the present application believes that in some cases, regardless of the specific situation of the tightness of the bandage, the shaking and posture of some patients may directly cause compression at the puncture point and its vicinity, thereby causing further bleeding. Alternatively, patients in an unconscious / unaware state such as sleep may also be subjected to external compression, thereby causing further bleeding. Alternatively, some special situations such as touching the patient or the bed may also cause abnormal fluctuations in bleeding.

[0091] Thus, under the condition of introducing pressure monitoring and participating in bleeding volume prediction, more accurate condition determination can be assisted, false positives can be reduced, and bleeding volume prediction accuracy can be improved.

[0092] In addition, based on the accurate monitoring of puncture point bleeding, the system can also be linked to online work systems that can be involved in hospital scenarios such as hospital information systems (HIS), to update the patient's condition in real time, and also as a warning output channel.

[0093] In some cases, the puncture point bleeding situation monitored in real time, i.e., the real-time bleeding risk assessment result, can be combined with patient anticoagulant medication, coagulation function data (which can also involve an artificial confirmation link), to ultimately achieve a full closed-loop management of "bleeding discovery-accurate compression-drug intervention", which can help reduce severe bleeding events by more than 60% and reduce the average bleeding volume to less than 10ml.

[0094] In addition, it can be understood that some of the hospital online work systems involved here can be involved in external calls through docking / linkage, or the puncture point bleeding real-time monitoring system can be directly implanted into the existing hospital online work system to apply the scheme. Obviously, in the specific application scheme, this can be flexibly configured.

[0095] Finally, for the above scheme content, in general, for the real-time monitoring target of puncture point bleeding, the application specially builds a novel puncture point bleeding real-time monitoring architecture. Compared with the limitation of single signal detection, high-performance puncture point bleeding real-time monitoring effect is realized by fusing multi-dimensional physical parameters, and the monitoring performance is further enhanced through corresponding optimization settings in many details. Thus, real-time, high-precision, high-processing efficiency and light-weight high-quality puncture point bleeding real-time monitoring services are provided, which realizes better tool support for puncture point bleeding real-time monitoring work and plays a good decision-making assistance role.

[0096] The above is the introduction of the puncture point bleeding real-time monitoring system based on multi-modal sensing provided by the application. Correspondingly, the application also provides a puncture point bleeding real-time monitoring method based on multi-modal sensing from the perspective of work flow. The puncture point bleeding real-time monitoring method based on multi-modal sensing is applied to the puncture point bleeding real-time monitoring system based on multi-modal sensing. The puncture point bleeding real-time monitoring system based on multi-modal sensing briefly includes a monitoring terminal and a data processing platform. The monitoring terminal with a bandage structure design is used to be configured on the puncture point area of the patient's skin. The multi-modal sensors in the monitoring terminal include conductivity sensors, temperature sensors and humidity sensors. The monitoring terminal adopts a detachable battery, and the multi-modal sensors are arranged in multiple groups along the radial direction of the puncture point.

[0097] Based on the above brief system structure content, with reference to Figure 2 The application provides a puncture point bleeding real-time monitoring method based on multi-modal sensing. The puncture point bleeding real-time monitoring method based on multi-modal sensing provided by the application can specifically include the following steps S201 to S204: Step S201, the monitoring terminal respectively collects conductivity data, temperature data and humidity data through the conductivity sensor, the temperature sensor and the humidity sensor; Step S202, the monitoring terminal transmits the conductivity data, the temperature data and the humidity data to the data processing platform through the communication module; Step S203, the data processing platform performs corresponding real-time bleeding risk assessment processing based on the multi-modal data formed by the conductivity data, the temperature data and the humidity data; Step S204, the data processing platform performs hierarchical warning output based on the real-time bleeding risk assessment result.

[0098] As an exemplary embodiment, the multi-modal sensor adopts a three-layer flexible sensor structure design, including a bottom layer, a middle layer and a top layer; The bottom layer corresponding to the patient's skin side is a hydrophobic breathable film, mainly used for preventing blood cells from blocking; The middle layer is implanted with interdigital electrodes and coated with a diamond-like coating, mainly used for data acquisition tasks through the electrode array structure; The top layer corresponding to the outside is a nano-porous PVDF film, mainly used for specific adsorption of plasma proteins to enhance signal sensitivity.

[0099] As another exemplary embodiment, the real-time bleeding risk assessment process follows the following strategy content: Low-risk warning level, multi-modal data estimated bleeding volume <5ml / h, humidity data rise rate ≤0.5% / min, corresponding to the warning mode of the mobile phone App warning, and the clinical treatment suggestion of corresponding intensive observation + 30 minutes of re-examination; Medium-risk warning level, multi-modal data estimated bleeding volume 5-50ml / h, conductivity data rise rate >15%, corresponding to the warning mode of the nurse station sound and light alarm, and the clinical treatment suggestion of corresponding adjustment of compression band + examination of coagulation indicators; High-risk warning level, multi-modal data estimated bleeding volume >50ml / h, temperature data rise rate >1°C / min, corresponding to the warning mode of the monitoring terminal high-frequency buzzer warning + compression airbag automatic inflation, and the clinical treatment suggestion of corresponding emergency bedside treatment + preparation of hemostatic drugs or preparation of hemostatic surgery.

[0100] As another exemplary embodiment, the real-time bleeding risk assessment process also follows the following strategy content: Blood conductivity range is 3-5S / m, interstitial fluid conductivity range is 1-2S / m, fresh blood temperature is higher than body surface by 2-3℃, and sweat conductivity range is >10S / m.

[0101] As another exemplary embodiment, the real-time bleeding risk assessment process also follows the following strategy content: On the basis of multi-modal sensors arranged in multiple groups along the radial direction of the puncture point, the specific type of the current condition is subdivided by analyzing the humidity diffusion direction and speed, and the specific type of the current condition is subdivided into both local small amount of bleeding type and diffuse active bleeding type.

[0102] As another exemplary embodiment, the real-time bleeding risk assessment process is specifically processed by a processing model configured on the gateway node, the processing model is a machine learning model, and the processing model is trained by a sample bleeding sample with corresponding annotations in advance.

[0103] As another exemplary embodiment, the index threshold involved in the real-time bleeding risk assessment process is also adaptively adjusted according to the patient's body obesity condition and the patient's posture condition during monitoring, wherein the patient's body obesity condition is specifically modeled by pre-existing CT data or MRI data, and the patient's posture condition is specifically monitored by a posture sensor; The sensor sensitivity is also adaptively adjusted according to the obesity of the patient's body.

[0104] As another exemplary embodiment, the multi-modal sensor further comprises a distributed pressure sensor array, and the data processing center generates a compression force heat map based on the pressure data, and if the local pressure exceeds the ischemia threshold or is lower than the hemostasis failure threshold, the data processing center automatically adjusts the tightness and prompts.

[0105] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-mentioned multi-modal sensor-based real-time monitoring method for puncture point bleeding can be referred to as Figure 1 The corresponding embodiment of the multi-modal sensor-based real-time monitoring system for puncture point bleeding is described, and the specific operation is not described here.

[0106] Those skilled in the art can understand that all or part of the steps in the various methods of the above-mentioned embodiments can be completed by instructions, or by related hardware controlled by instructions, which can be stored in a computer readable storage medium and loaded and executed by a processor.

[0107] Therefore, the present application provides a computer readable storage medium, which stores a plurality of instructions capable of being loaded by a processor to execute the present application as Figure 2 The steps of the multi-modal sensor-based real-time monitoring method for puncture point bleeding in the corresponding embodiment are described, and the specific operation can be referred to as Figure 2 The multi-modal sensor-based real-time monitoring method for puncture point bleeding in the corresponding embodiment is described, and the specific operation is not described here.

[0108] The computer readable storage medium can include a read only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0109] Due to the instructions stored in the computer readable storage medium, the present application as Figure 2 The steps of the multi-modal sensor-based real-time monitoring method for puncture point bleeding in the corresponding embodiment, therefore, can achieve the present application as Figure 2 The beneficial effects of the multi-modal sensor-based real-time monitoring method for puncture point bleeding in the corresponding embodiment can be achieved, as described above, and are not described here.

[0110] The above provides a multi-modal sensing-based puncture point bleeding real-time monitoring system, a multi-modal sensing-based puncture point bleeding real-time monitoring method and a computer readable storage medium provided by the present application. The principles and implementation modes of the present application are described in this paper by applying specific examples. The above examples are only used to help understand the core idea of the present application. At the same time, for those skilled in the art, according to the idea of the present application, the specific implementation mode and application range will be changed. In view of the above, the content of the specification should not be understood as a limitation of the present application.

Claims

1. A real-time monitoring system for puncture site bleeding based on multimodal sensing, characterized in that, The puncture point bleeding real-time monitoring system based on multi-modal sensing includes a monitoring terminal and a data processing center. The monitoring terminal is designed to be attached to the skin of the patient at the puncture point. The multi-modal sensors in the monitoring terminal include conductivity sensors, temperature sensors, and humidity sensors. The monitoring terminal uses a detachable battery. The multi-modal sensors are arranged in multiple groups radially along the puncture point. During operation, the system includes the following processing content: The monitoring terminal collects conductivity data, temperature data, and humidity data through the conductivity sensors, temperature sensors, and humidity sensors, respectively, and transmits them to the data processing center through a communication module. The data processing center performs real-time bleeding risk assessment based on the multi-modal data formed by the conductivity data, temperature data, and humidity data, and outputs a graded warning based on the real-time bleeding risk assessment results.

2. The multi-modal sensor based real-time monitoring system for puncture site bleeding according to claim 1, wherein, The multi-modal sensors use a three-layer flexible sensor structure, including a bottom layer, a middle layer, and a top layer. The bottom layer corresponding to the patient's skin side is a hydrophobic and breathable film, mainly used to prevent blood cells from blocking. The middle layer is implanted with cross-finger electrodes and coated with a diamond-like coating, mainly used for data collection through the electrode array structure. The top layer corresponding to the outside is a nano-porous PVDF film, mainly used for specific adsorption of plasma proteins to enhance signal sensitivity.

3. The multi-modal sensor based real-time monitoring system for puncture site bleeding according to claim 1, wherein, The real-time bleeding risk assessment process follows the following strategies: Low-risk warning level: the bleeding volume calculated by the multi-modal data is <5ml / h, the humidity data rise rate is ≤0.5% / min, the warning mode of the mobile phone App, and the clinical treatment suggestion of 30 minutes of intensive observation and re-examination. Medium-risk warning level: the bleeding volume calculated by the multi-modal data is 5-50ml / h, the conductivity data rise rate is >15%, the warning mode of the nurse station sound and light alarm, and the clinical treatment suggestion of adjusting the compression band and checking the blood coagulation indicators. High-risk warning level: the bleeding volume calculated by the multi-modal data is >50ml / h, the temperature data rise rate is >1°C / min, the warning mode of the monitoring terminal high-frequency buzzer warning and compression airbag automatic inflation, and the clinical treatment suggestion of emergency bedside treatment and preparation of hemostatic drugs or hemostatic surgery.

4. The multi-modal sensor based real-time monitoring system for puncture site bleeding according to claim 3, wherein, The real-time bleeding risk assessment process also follows the following strategies: The blood conductivity range is 3-5S / m, the tissue fluid conductivity range is 1-2S / m, the fresh blood temperature is 2-3°C higher than the body surface, and the sweat conductivity range is >10S / m.

5. The multi-modal sensor based real-time monitoring system for puncture site bleeding as claimed in claim 3, wherein, Based on the multi-modal sensors arranged in multiple groups radially along the puncture point, the specific type of the current condition is subdivided into local small amount of bleeding type and diffuse active bleeding type by analyzing the humidity diffusion direction and speed. ​ 6. The multi-modal sensor based real-time monitoring system for puncture site bleeding as claimed in claim 3, wherein, The real-time bleeding risk assessment processing is specifically processed by a processing model configured on the gateway node, the processing model is a machine learning model, and the processing model is trained in advance by a sample bleeding sample configured with a corresponding label.

7. The multi-modal sensor based real-time monitoring of puncture site bleeding system of claim 1, wherein, The index threshold involved in the real-time bleeding risk assessment processing is also adaptively adjusted according to a patient body obesity condition and a patient posture condition during monitoring, wherein the patient body obesity condition is specifically modeled in advance by pre-existing CT data or MRI data, and the patient posture condition is specifically monitored by a posture sensor; The sensor sensitivity is also adaptively adjusted according to the patient body obesity condition.

8. The multi-modal sensor based real-time monitoring of puncture site bleeding system of claim 1, wherein, The multi-modal sensor further includes a distributed pressure sensor array, and the data processing center generates a compression force heat map based on pressure data, and if the local pressure exceeds an ischemia threshold or is lower than a hemostasis failure threshold, the data processing center automatically adjusts and prompts the tightness condition.

9. A method for real-time monitoring of puncture site bleeding based on multimodal sensing, characterized in that, The puncture point bleeding real-time monitoring method based on multi-modal sensing is applied to a puncture point bleeding real-time monitoring system based on multi-modal sensing, the puncture point bleeding real-time monitoring system based on multi-modal sensing includes a monitoring terminal and a data processing center, the monitoring terminal with a bandage structure is used to be arranged on the area of the puncture point on the patient's skin, the multi-modal sensor in the monitoring terminal includes an electrical conductivity sensor, a temperature sensor and a humidity sensor, the monitoring terminal adopts a detachable battery, the multi-modal sensor is arranged in multiple groups along the radial direction of the puncture point, and the method includes: The monitoring terminal collects electrical conductivity data, temperature data and humidity data through the electrical conductivity sensor, the temperature sensor and the humidity sensor respectively; The monitoring terminal transmits the electrical conductivity data, the temperature data and the humidity data to the data processing center through a communication module; The data processing center performs corresponding real-time bleeding risk assessment processing based on multi-modal data formed by the electrical conductivity data, the temperature data and the humidity data; The data processing center performs hierarchical early warning output based on the real-time bleeding risk assessment result.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by the processor to execute the method of claim 9.

Images