Multi-mode flexible intra-abdominal pressure measuring device and method based on AI

Abstract

The invention discloses a multi-mode flexible intra-abdominal pressure measuring device and method based on AI, and belongs to the technical field of medical monitoring. The device comprises a flexible wearable carrier, a multi-mode sensor module, a signal conditioning and intelligent wireless transmission module and a data processing module, the flexible wearable carrier is provided with an adjustable fixing assembly and a positioning identifier; the multi-modal sensor module integrates various sensors of pressure, stress strain, sound, acceleration and oxyhemoglobin saturation infrared light, and synchronously collects multi-dimensional physiological signals; the data processing module is provided with an AI algorithm, dynamic scene recognition, self-adaptive weight fusion and individual difference compensation can be achieved, and an intra-abdominal pressure measurement value is obtained through calculation. Non-invasive, real-time and high-precision continuous monitoring of the intra-abdominal pressure is achieved, the clinical measurement error is small, the system is suitable for ICU, postoperative rehabilitation, home management and other scenes, infants and old people are covered, and the comfort, accuracy and clinical efficiency of intra-abdominal pressure monitoring are greatly improved.

Description

Technical Field

[0001] This application relates to the field of medical monitoring equipment and intelligent sensing technology, specifically to an AI-based multimodal flexible intra-abdominal pressure measurement device and method. Background Technology

[0002] Intra-abdominal pressure is a key physiological parameter for clinically assessing abdominal physiological function, predicting disease progression, and guiding treatment. Its accurate measurement is irreplaceable for early warning of intra-abdominal hypertension, prevention and treatment of abdominal compartment syndrome, and reducing mortality in critically ill patients. Clinical statistics show that the incidence and mortality rates of intra-abdominal hypertension and abdominal compartment syndrome are both high among intensive care unit (ICU) patients. Real-time and accurate monitoring of intra-abdominal pressure is a core prerequisite for early intervention and improved patient prognosis in these diseases.

[0003] Currently, the commonly used methods for measuring intra-abdominal pressure in clinical practice are mainly divided into direct measurement and indirect measurement methods. Direct measurement involves connecting a pressure sensor through an abdominal drainage tube or puncture needle. Although the measured value is relatively close to the true value, it is an invasive procedure that is prone to complications including but not limited to infection, bleeding, and organ damage. Furthermore, it has strict requirements on the patient's abdominal cavity conditions and is only suitable for patients undergoing specific surgeries, thus having significant limitations in clinical application.

[0004] Indirect measurement methods, particularly interventional methods, use intravesical pressure measurement as the clinical gold standard. This involves instilling sterile fluid into the bladder via a catheter and then calculating the intra-abdominal pressure based on the hydrostatic pressure of the fluid. This method is cumbersome, time-consuming, and cannot achieve continuous real-time monitoring. The measurement accuracy is easily affected by factors including, but not limited to, bladder fullness, catheter blockage, patient position, and abdominal muscle tension. Furthermore, it carries risks including, but not limited to, urinary tract infections, venous thrombosis, and gastric mucosal damage, making it unsuitable for long-term monitoring needs.

[0005] Existing non-invasive indirect measurement technologies mostly rely on single types of sensors, including but not limited to pressure sensors and strain sensors. Measurement accuracy is easily affected by factors such as individual abdominal wall characteristics, respiratory movements, changes in body position, and the external environment, resulting in poor repeatability and large errors, making it difficult to meet the needs of precise clinical monitoring. Furthermore, most non-invasive devices use rigid shell designs, lacking ergonomic flexibility to conform to the curves of the human abdomen, leading to poor wearing comfort, skin discomfort, and an inability to achieve long-term continuous monitoring. They are particularly unsuitable for special populations, including but not limited to infants, children with sensitive skin, obese individuals, or those who are underweight.

[0006] Current non-invasive intra-abdominal pressure measurement technology suffers from several systemic defects: First, at the algorithm level, it generally employs a fixed algorithm and fixed weight multi-sensor fusion strategy, which cannot dynamically adjust parameters according to the real-time monitoring scenario of the patient, nor can it adapt to the differences in physiological characteristics of different patients, resulting in a high data loss rate and a significant increase in measurement error in dynamic scenarios. Second, it has insufficient battery life; the built-in batteries of existing devices have short battery life, which cannot support long-term continuous wear, and frequent charging also increases the burden on clinical care. Third, it has narrow scenario coverage, only suitable for adults in a static supine position, and cannot meet the needs of multiple scenarios including but not limited to ICU patient care, postoperative patient rehabilitation, and home-based chronic disease management. Fourth, it lacks intelligent remote management functions, and medical staff cannot obtain real-time early warning information of abnormal intra-abdominal pressure in patients, resulting in low efficiency in clinical data traceability and analysis.

[0007] In summary, current intra-abdominal pressure measurement technologies suffer from multiple problems, including high invasive risks, complex operation, insufficient measurement accuracy, poor adaptability to dynamic scenarios, low level of intelligence, short battery life, and limited applicable population. There is an urgent need for an intra-abdominal pressure measurement solution that combines non-invasive safety, high precision, flexible wearability, intelligent dynamic optimization, long battery life, and adaptability to multiple scenarios and all population groups. Summary of the Invention

[0008] This application provides an AI-based multimodal flexible intra-abdominal pressure measurement device and method, which can solve the technical problems existing in the prior art, such as reliance on a single sensing mode, insufficient measurement accuracy due to fixed algorithms, poor scene adaptability, and inability to continuously monitor.

[0009] In a first aspect, this application provides an AI-based multimodal flexible intra-abdominal pressure measurement device, comprising: A flexible wearable carrier with adjustable elastic fixing components and positioning markers; The multimodal sensor module, housed within the flexible wearable carrier, includes a pressure sensor, a stress-strain sensor, a sound sensor, an acceleration sensor, and a blood oxygen saturation infrared sensor. These sensors are used to acquire direct pressure signals from the abdominal wall in contact with the body surface, deformation signals of the abdominal wall caused by changes in intra-abdominal pressure, sound signals related to the movement of intra-abdominal organs, respiration, and disturbances, real-time body position and physical activity signals of the patient, and minute fluctuations in blood perfusion caused by changes in intra-abdominal pressure. The signal conditioning and intelligent wireless transmission module is electrically connected to the multimodal sensor module and includes a signal conditioning circuit, a multifunctional wireless transmission unit, and a miniature demodulation module, which are respectively used to amplify and reduce noise of the original sensor signal, send the processed data to the terminal device, and convert the spectral signal of the flexible fiber optic sensor unit into an electrical signal. The data processing module is configured to run an AI algorithm to perform fusion processing and intra-abdominal pressure calculation on the multi-source synchronous data collected by the multimodal sensor module, and obtain the final intra-abdominal pressure measurement value.

[0010] Furthermore, the pressure sensor includes a piezoresistive pressure sensor and a capacitive pressure sensor; The piezoresistive pressure sensor is used to stably capture static pressure signals from the abdominal wall. The capacitive pressure sensor is used to detect minute dynamic fluctuations in intra-abdominal pressure.

[0011] Furthermore, the flexible wearable carrier is equipped with a standardized sensor expansion interface for connecting one or more of temperature sensors, electromyography sensors, and electrocardiogram sensors according to clinical needs; The adjustable elastic fixation component includes Velcro straps and medical hypoallergenic silicone tape; The location markers include fluorescent markers that are identifiable in low-light environments.

[0012] Furthermore, the device also includes: The bioenergy harvesting module is electrically connected to the signal conditioning and intelligent wireless transmission module, and its internal components include a micro thermoelectric generator, a piezoelectric film, a power management unit, and an energy storage unit. The micro thermoelectric generator is used to generate electricity by utilizing the temperature difference between the abdominal wall and the environment. The piezoelectric film is used to generate electricity through body deformation. The power management unit is used to rectify, stabilize, and manage the output power of the micro thermoelectric generator and the piezoelectric film. The energy storage unit is used to store the managed energy to power the device. The bioenergy harvesting module is connected to the signal conditioning and intelligent wireless transmission module via a snap-on detachable structure, and the energy storage unit is a flexible solid-state lithium battery.

[0013] Furthermore, the multi-functional wireless transmission unit includes one or more of Wi-Fi, Bluetooth, and mobile networks, supporting local offline data storage and automatic synchronization after network recovery; the signal conditioning and intelligent wireless transmission module also includes a cross-modal data synchronization calibration unit, which performs timing consistency calibration on the output signals of the multi-modal sensor module based on high-precision timestamp alignment technology and clock synchronization protocol.

[0014] Furthermore, the data processing module includes: The scene classification unit is used to determine the current real-time monitoring scene based on the body position data of the accelerometer and the acoustic features of the sound sensor through a pre-trained AI scene classification model. The dynamic weight fusion unit is used to call a predefined scene weight mapping database according to the determined monitoring scene, dynamically allocate fusion weights to the multimodal sensor data, and calculate the original value of intra-abdominal pressure using a weighted fusion algorithm. The individual difference compensation unit is used to calculate personalized correction coefficients based on the input individual physiological parameters through a pre-trained AI individual difference compensation model, to compensate for the original intra-abdominal pressure value and obtain the final intra-abdominal pressure measurement value.

[0015] Furthermore, the data processing module also includes a risk warning unit, which is used to determine the risk level of intra-abdominal hypertension based on the final intra-abdominal pressure measurement value and historical trend data through a pre-trained risk assessment model, generate personalized clinical intervention suggestions, and push abnormal warning information to the smart terminal. The device employs end-to-end encrypted transmission, hierarchical access control, distributed cloud backup, and a preset periodic automatic data cleanup mechanism to ensure the security of medical data.

[0016] Secondly, this application provides an AI-based multimodal flexible intra-abdominal pressure measurement method, applied to the AI-based multimodal flexible intra-abdominal pressure measurement device described above, comprising the following steps: The device is aligned with the target area using the positioning marker and then worn and secured to the patient's abdomen. The device is activated, and the timing calibration of the multimodal sensors is completed through the cross-modal data synchronization calibration unit. Multi-dimensional physiological signals are synchronously acquired through the multimodal sensor module. The physiological signal is amplified, filtered, and noise-reduced by the signal conditioning and intelligent wireless transmission module, and then encrypted and transmitted to the data processing module. The data processing module uses AI algorithms to calculate and obtain the raw value of intra-abdominal pressure based on the received synchronous data, and then corrects it based on the input individual physiological parameters to obtain the final intra-abdominal pressure measurement value. The final intra-abdominal pressure measurement value is displayed in real time on the smart terminal, stored locally, and synchronized to the cloud management platform.

[0017] Furthermore, the data processing module, based on the received data, calculates the raw intra-abdominal pressure value through algorithmic fusion, and corrects it based on the input individual physiological parameters to obtain the final intra-abdominal pressure measurement value. This specifically includes the following steps: Based on real-time monitoring of body position changes by an accelerometer and respiratory status and environmental acoustic features collected by a sound sensor, the current real-time monitoring scene is determined by a pre-trained AI scene classification model. The multimodal features include posture features extracted from acceleration data and acoustic features extracted from audio signals. Based on the determined current real-time monitoring scenario, a predefined scenario weight mapping database is invoked to dynamically assign fusion weights to the multimodal sensor data. A weighted fusion algorithm is then used to integrate and calculate the multi-source heterogeneous data from pressure, stress-strain, sound, acceleration, and blood oxygen saturation infrared sensors, and output the raw value of intra-abdominal pressure. The integration calculation includes normalization processing of the multi-source heterogeneous sensor data based on the mean and standard deviation. Based on the user's pre-input individual physiological parameters, a personalized correction coefficient is calculated through an AI individual difference compensation model, and the original intra-abdominal pressure value is compensated to obtain the final intra-abdominal pressure measurement value.

[0018] Furthermore, the body position change data monitored in real time by the accelerometer and the respiratory state and environmental acoustic features collected by the sound sensor are used to determine the current real-time monitoring scene through a pre-trained AI scene classification model. The multimodal features include posture features extracted from the acceleration data and acoustic features extracted from the audio signal, specifically including the following steps: The three-axis attitude data from the accelerometer and the audio signal from the sound sensor are acquired and preprocessed simultaneously to extract multimodal features, which include attitude features extracted from the acceleration data and acoustic features extracted from the audio signal. Multimodal features are fused together to construct a standardized scene feature vector, which is then input into a pre-trained AI scene classification model for data processing to obtain the probability distribution. The current real-time monitoring scenario is determined based on the obtained probability distribution.

[0019] The beneficial effects of the technical solutions provided in this application include at least the following: This application adopts a completely non-invasive, flexible wearable design. It does not require puncture, intubation or sterile fluid infusion. Monitoring can be started simply by wearing the device. No professional medical staff are required to operate it. This avoids the risk of complications from invasive procedures and greatly reduces the burden of clinical care. It is also suitable for home monitoring scenarios.

[0020] This application utilizes multiple core sensors to form multi-dimensional data complementarity, combined with AI dynamic scene recognition, adaptive weight fusion, and individual difference compensation algorithms. It solves the pain point that the fixed algorithms of existing technologies cannot adapt to dynamic scenes and individual differences. The measurement accuracy is significantly improved compared with existing non-invasive technologies, and it is highly correlated with the clinical gold standard. Its anti-interference ability in dynamic scenes is significantly better than existing technologies.

[0021] The flexible wearable carrier of this application uses medical-grade skin-friendly and breathable materials, conforms to the curve of the human abdomen, has no rigid structure, and can be worn continuously for a long time; at the same time, it provides multiple size designs, with medical low-allergy fixation components, covering all groups including but not limited to infants, adults, the elderly, and patients with sensitive skin.

[0022] This application enables the self-generation of human energy through a detachable bioenergy harvesting module, significantly improving endurance and meeting the needs of long-term continuous monitoring. It is also adaptable to various scenarios, including but not limited to continuous monitoring of critically ill patients in the ICU, postoperative patient rehabilitation monitoring, and home management of patients with chronic diseases, supporting full-state monitoring.

[0023] This application features real-time risk warning, remote data management, and historical trend analysis functions, which can assist medical staff in achieving early warning and timely intervention for intra-abdominal hypertension. It can also seamlessly connect with hospital information systems to achieve standardized management and traceability of clinical data, significantly improving diagnostic and treatment efficiency. Attached Figure Description

[0024] Figure 1 A structural block diagram of the AI-based multimodal flexible intra-abdominal pressure measurement device provided in the embodiments of this application; Figure 2 This is a functional block diagram of the multimodal sensor module provided in the embodiments of this application; Figure 3 A functional block diagram of the bioenergy harvesting module provided in the embodiments of this application; Figure 4 This is a schematic diagram illustrating the process of multimodal sensor data fusion processing and intra-abdominal pressure calculation provided in an embodiment of this application. Detailed Implementation

[0025] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0026] First, some of the technical terms used in this application will be explained to help those skilled in the art understand this application.

[0027] AI: Artificial Intelligence; FBG: Fiber Bragg Grating; BMI: Body Mass Index; Wi-Fi: Wireless Fidelity; BLE: Bluetooth Low Energy.

[0028] Firstly, such as Figure 1 As shown, this application provides an AI-based multimodal flexible intra-abdominal pressure measurement device, including a flexible wearable carrier 10, a multimodal sensor module 20, a signal conditioning and intelligent wireless transmission module 30, and a data processing module 40.

[0029] The flexible wearable carrier 10 is equipped with adjustable elastic fixation components and positioning markers. The tensile elastic modulus of the flexible wearable carrier 10 is within a reasonable range for clinical application. Specifically, the flexible wearable carrier 10 is made of skin-friendly, breathable, and highly elastic medical-grade flexible material, preferably medical-grade polydimethylsiloxane (PDMS), with a thickness of approximately 0.3 mm, a breathability ≥3500 g / (m²·24h), and skin irritation compliance with GB / T16886.10-2017. The carrier shape conforms to the curve of the human abdomen, and the edges are rounded to avoid friction against the skin and reduce discomfort during long-term wear. The carrier surface is equipped with adjustable elastic fixation components, which can adaptively adjust the tightness according to the patient's waist circumference and abdominal wall curvature to ensure that the sensor fits tightly against the abdominal wall without affecting the patient's normal breathing, turning over, and other activities.

[0030] In one embodiment, the adjustable elastic fixation component is a Velcro strap or medical hypoallergenic tape to achieve individualized tightness. Patients or medical staff can flexibly adjust the tightness according to their waist circumference and abdominal wall curvature to ensure that the sensor array fits tightly against the abdominal wall. This avoids discomfort or breathing difficulties caused by excessive tightness, while also preventing sensor displacement or signal acquisition failure due to excessive looseness. Specifically, the length of the Velcro strap can be adjusted from 60cm to 120cm. Meanwhile, the medical hypoallergenic tape is made of a gentle material, effectively reducing the risk of irritation to the delicate skin of infants and other special populations. Furthermore, flexible wearable carriers 10 of corresponding sizes are designed for the abdominal physiological characteristics of different age groups, optimizing the material of the fixation component and the layout of the sensor array. For example, the infant version of the carrier is one-third the size of the adult version, the sensor array is reduced to a 3cm area around the navel, and the medical hypoallergenic tape is made of silicone, showing no redness or allergic reactions after skin irritation testing.

[0031] In one embodiment, the positioning marker facilitates quick and accurate wearing, avoiding measurement errors caused by sensor position deviation. Specifically, it is a fluorescent marker point, which is easy to identify in low-light environments. Medical staff or patients can quickly align the center of the sensor array with the target area on the abdomen (adults align with the navel, infants align with 1cm around the navel) based on the marker point, thus avoiding measurement errors caused by wearing position deviation from the source and improving the reliability and repeatability of the data. Preferably, four fluorescent positioning marker points can be set in a rectangular distribution to assist in accurate positioning.

[0032] In one embodiment, the flexible wearable carrier 10 has a reserved standardized sensor expansion interface, which can flexibly connect to other medical monitoring sensors such as temperature sensors and blood oxygen saturation sensors according to clinical needs. The expansion process does not require changes to the core structure and data processing logic of the device; only simple software configuration is needed to achieve the acquisition and fusion of new data. In a modified embodiment of this application, this interface can also connect to sensors such as muscle tension sensors and electrocardiogram sensors to expand monitoring functions. Specifically, muscle tension sensors can be evenly arranged along the circumference of the carrier to collect abdominal wall muscle state signals in real time, and AI algorithms can be used to eliminate muscle activity interference.

[0033] The multimodal sensor module 20 is disposed within the flexible wearable carrier 10, such as... Figure 2 As shown, the multimodal sensor module 20 includes at least five sensors, including a pressure sensor, a stress-strain sensor, a sound sensor, an acceleration sensor, and a blood oxygen saturation infrared light sensor; the blood oxygen saturation infrared light sensor is implemented using a flexible fiber optic sensor unit (such as a fiber Bragg grating, FBG). Among them, the pressure sensor is used to acquire the direct pressure signal between the abdominal wall and the body surface; the stress-strain sensor is used to acquire the deformation signal of the abdominal wall caused by pressure changes; the sound sensor is used to acquire the sound signals such as the movement of organs in the abdominal cavity and respiration; the acceleration sensor is used to acquire the patient's real-time body position and body activity status signals; and the blood oxygen saturation infrared light sensor is used to acquire the minute fluctuation signal of intra-abdominal pressure through spectral changes.

[0034] In one embodiment, the multimodal sensor module 20 is based on five types of core sensors, but is not limited to these five types of sensors. Other suitable medical sensors can be integrated according to clinical monitoring needs. All sensors are miniaturized and can be embedded in the flexible wearable carrier 10 in a preset array layout using MEMS technology to avoid direct contact with the skin and causing irritation or discomfort.

[0035] In one embodiment, the pressure sensor includes a piezoresistive pressure sensor and a capacitive pressure sensor. The piezoresistive sensor is used to stably capture the static pressure signal of the abdominal wall, ensuring the stability of the basic measurement. Specifically, its range can be 0-100 mmHg, and its accuracy can reach 0.1 mmHg. It can be arranged at the center of the carrier. The capacitive sensor is used to accurately sense minute fluctuations in intra-abdominal pressure, improving the sensitivity to subtle pressure changes. Specifically, its resolution can reach 0.05 mmHg. It can be symmetrically arranged on both sides of the piezoresistive sensor, forming a point-surface combined layout of "center-both sides".

[0036] In one embodiment, stress-strain sensors are arranged circumferentially along the flexible wearable carrier 10 to detect the tensile or contractile deformation of the abdominal wall caused by changes in intra-abdominal pressure. The degree of deformation is used to infer the overall trend of intra-abdominal pressure changes, forming a complementary "point-to-surface" data with the "point pressure signal" of the pressure sensor. Specifically, four stress-strain sensors can be evenly arranged circumferentially.

[0037] In one embodiment, the sound sensor employs a miniature high-sensitivity microphone to collect sounds of organ activity within the abdominal cavity. Changes in sound frequency and intensity help determine the intra-abdominal pressure status, while also identifying interfering actions such as breathing and coughing, providing a basis for data filtering. Specifically, a MEMS microphone can be used, positioned 2 cm above the umbilicus, with a sampling frequency of up to 44.1 kHz.

[0038] In one embodiment, an accelerometer is used to monitor changes in patient position and limb movement in real time. The position and movement data provide a dynamic correction benchmark for the data collected by pressure and stress-strain sensors, eliminating measurement deviations caused by factors other than intra-abdominal pressure. Specifically, it can be a triaxial MEMS accelerometer, arranged at the center of the carrier, with a sampling frequency of 100Hz.

[0039] In one embodiment, the blood oxygen saturation infrared sensor is implemented using a flexible fiber optic sensor unit, which is a fiber Bragg grating sensor. It is alternately arranged with the pressure sensor on the flexible wearable carrier 10, realizing the spatial synchronous acquisition of small intra-abdominal pressure fluctuation signals and body surface contact pressure signals, providing a complementary and interference-resistant sensing basis for multimodal data fusion.

[0040] The signal conditioning and intelligent wireless transmission module 30 is electrically connected to the multimodal sensor module 20 and includes a signal conditioning circuit, a multifunctional wireless transmission unit, and a miniature demodulation module. These components are used to amplify and reduce noise in the original sensor signal, transmit the processed data to the terminal device, and convert the spectral signal of the flexible fiber optic sensor unit into an electrical signal, respectively. The signal conditioning circuit has a certain level of waterproof and dustproof capability, and the wireless transmission module has electromagnetic interference resistance to ensure stable operation in complex environments. Specifically, the signal conditioning circuit has an IP67 waterproof rating and can withstand short-term immersion. The wireless transmission module has a signal loss rate of ≤0.5% under 30V / m electromagnetic interference, making it suitable for dense clinical medical equipment environments.

[0041] In one embodiment, the signal conditioning and intelligent wireless transmission module 30 is integrated into the edge region of the flexible wearable carrier 10, and mainly includes a signal conditioning circuit, a multi-functional intelligent wireless transmission unit, and a miniature demodulation module. The signal conditioning circuit amplifies, filters, and reduces noise in the weak electrical signals output by each sensor, effectively removing signal impurities such as environmental electromagnetic interference and skin contact noise, improving the quality of the original data, and laying the foundation for accurate processing by subsequent AI algorithms. Specifically, this circuit can be a multi-channel programmable amplification and filtering circuit with an adjustable amplification factor of 10-1000 times, a filtering bandwidth of 0.1-1000Hz, and a sampling frequency ≥100Hz.

[0042] This multi-functional intelligent wireless transmission unit supports various wireless transmission methods such as Wi-Fi, Bluetooth, and mobile networks, enabling real-time and stable transmission of processed sensor data to computer terminals, smartphones, and home health management platforms. The module also features a built-in micro-storage chip for local data storage, preventing data loss due to network interruptions and automatically synchronizing to terminal devices once the network is restored. Furthermore, the module integrates low-power management, dynamically adjusting sensor acquisition and data transmission frequencies based on the energy output of the bioenergy harvesting module 50 using AI algorithms to extend the device's usage time on a single charge. Specifically, the unit supports Bluetooth 5.0 (including Bluetooth Low Energy / BLE mode) and Wi-Fi, and its built-in storage chip has a capacity of 1GB.

[0043] The miniature demodulation module connects to the flexible fiber optic sensor unit, converting the spectral wavelength drift signal acquired by the flexible fiber optic sensor unit into an electrical signal that can be processed by the signal conditioning circuit. Its sampling frequency is consistent with other sensors to ensure synchronized data acquisition. The signal conditioning circuit has a certain level of waterproof and dustproof capability, and the wireless transmission module has electromagnetic interference resistance to ensure stable operation in complex environments.

[0044] In one embodiment, the multi-functional wireless transmission unit is one or more of Wi-Fi, Bluetooth, and mobile networks, constructing a multi-mode, adaptive, and highly reliable communication system. This enables seamless and stable transmission of monitoring data in different scenarios such as clinical settings and home settings, providing crucial communication support for the intelligent management of the device, the realization of remote monitoring functions, and ultimately improving the timeliness and coverage of medical services.

[0045] In one embodiment, the signal conditioning and intelligent wireless transmission module 30 further includes a cross-modal data synchronization calibration unit. Based on high-precision timestamp alignment technology and clock synchronization protocols (such as the PTP protocol), this unit performs timing consistency calibration on the output signals of pressure, strain, sound, acceleration, blood oxygen, and muscle tension sensors. This unit can eliminate delay errors in multi-source data acquisition and transmission, ensuring that all sensor data are aligned on a unified timeline, providing a synchronization basis for subsequent AI fusion algorithms and reducing cross-modal fusion errors. Specifically, its time synchronization error can be ≤1ms.

[0046] Furthermore, the signal conditioning and intelligent wireless transmission module 30 integrates an electromagnetic interference shielding unit and an environmental noise adaptive filtering unit. The electromagnetic interference shielding unit encapsulates the entire module with a flexible conductive material, which can attenuate electromagnetic radiation in the clinical environment (shielding effectiveness ≥30dB). The environmental noise adaptive filtering unit, based on an AI noise recognition model (such as the LMS algorithm), identifies and filters motion artifacts, clothing friction noise, etc. in real time, and dynamically adjusts the filtering parameters to ensure signal purity.

[0047] The data processing module 40 is configured to run an AI algorithm to fuse and calculate intra-abdominal pressure on the data collected by the multimodal sensor module 20, obtaining the final intra-abdominal pressure measurement value. The intra-abdominal pressure measurement covers common clinical pressure ranges, and the measurement error is controlled within a clinically acceptable range. The sensor signal acquisition response is rapid, meeting the needs of real-time monitoring. Specifically, this module can be deployed on a local smart terminal or a cloud server, and its AI algorithm includes a scene classification unit, a dynamic weight fusion unit, an individual difference compensation unit, and a risk warning unit.

[0048] In one embodiment, the scene classification unit is used to determine the current real-time monitoring scene (such as supine with stable breathing, or a change in body position) based on the body position data from the accelerometer and the acoustic features from the sound sensor, using a pre-trained AI scene classification model. Specifically, the model can be built based on a CNN+LSTM network structure and trained with a large number of clinical samples, for example, using 500 clinical samples covering different body positions, respiratory states, ages, and BMIs, achieving a scene classification accuracy of ≥98%.

[0049] In one embodiment, the dynamic weight fusion unit is used to dynamically allocate fusion weights to multimodal sensor data by calling a predefined scene weight mapping database according to the determined monitoring scene, and calculate the original value of intra-abdominal pressure using a weighted fusion algorithm. Specifically, the integration calculation may include Z-score normalization processing of multi-source heterogeneous sensor data based on mean and standard deviation. There are weight allocation examples for different scenes. For example, in the supine breathing stable scenario, the predefined weight combination is: pressure sensor weight 0.35, stress strain sensor weight 0.20, acceleration sensor weight 0.15, sound sensor weight 0.15, and fiber optic sensor (blood oxygen saturation infrared sensor) weight 0.15.

[0050] In one embodiment, the individual difference compensation unit is used to calculate personalized correction coefficients based on input individual physiological parameters (such as abdominal wall thickness and BMI) using a pre-trained AI individual difference compensation model, thereby compensating for the original intra-abdominal pressure value and obtaining the final intra-abdominal pressure measurement. Specifically, the model can be trained based on the gradient boosting tree algorithm, fitted using 1000 clinical sample data, with a model fit degree R. 2 ≥0.92.

[0051] The data processing module 40 also includes a risk warning unit, which is used to determine the risk level of intra-abdominal hypertension based on the final intra-abdominal pressure measurement value and historical trend data through a pre-trained risk assessment model, generate personalized clinical intervention suggestions, and push abnormal warning information to the smart terminal.

[0052] In one embodiment, such as Figure 3 As shown, the device also includes a bioenergy harvesting module 50, connected to the signal conditioning and intelligent wireless transmission module 30. Internally, it integrates a micro-thermoelectric generator, a piezoelectric film, and an energy storage unit. The micro-thermoelectric generator generates electricity using the temperature difference between the abdominal wall and the environment, the piezoelectric film generates electricity through body deformation, and the energy storage unit stores the electrical energy generated by these two components to power the device. Specifically, the bioenergy harvesting module 50 is connected to the main module via a snap-on detachable structure. Internally, it can integrate four micro-thermoelectric generators (5mm x 5mm x 0.2mm, operating temperature difference ≥ 2℃), a PVDF piezoelectric film (0.1mm thick), a power management unit, and a flexible solid-state lithium battery.

[0053] The bioenergy harvesting module 50 in this embodiment generates its own power by comprehensively utilizing two types of environmental energy: body temperature difference and body activity deformation. This significantly improves the device's endurance, reduces dependence on and replacement frequency of external batteries, and provides key technical support for long-term or even permanent implantation or wearing of the device. It also enhances its practicality in long-term clinical monitoring and daily home use scenarios.

[0054] In one embodiment, the bioenergy harvesting module 50 is detachably connected to the signal conditioning and intelligent wireless transmission module 30. Users can flexibly decide whether to install this module based on specific monitoring scenarios and duration requirements. For example, for short-term postoperative monitoring, installation may not be necessary, as the device can meet its power requirements with its built-in battery, making the device lighter and easier to wear. However, for long-term home monitoring of chronic diseases, installing this module can significantly extend battery life, reduce charging frequency, and achieve near-unnoticeable continuous monitoring. This "on-demand" model allows a single product to perfectly adapt to diverse application scenarios, from short-term clinical use to long-term home use. Through modular and configurable design, the product's adaptability and flexibility to different monitoring scenarios and user groups are greatly enhanced, while also considering cost control, ease of maintenance, and wearing comfort.

[0055] In one embodiment, during the disassembly of the bioenergy harvesting module 50, the operator first presses the latching mechanisms on both sides of the module to unlock it, and then performs a plug-and-play separation along the guide rail. The disassembled module can be charged independently at a standard 5V voltage, with a full charging time not exceeding 2 hours, or it can be directly replaced with a new module. During installation, the module guide rail must be aligned with the interface and inserted smoothly. Once the latches are locked, the assembly is complete. At this point, the connection between the module and the host can achieve a strength of over 5N, ensuring its stability and safety during daily use.

[0056] In one embodiment, to ensure data security, the device employs end-to-end encrypted transmission (such as the AES-256 algorithm), hierarchical access control (such as three-level permissions for patients, doctors, and administrators), distributed cloud backup, and an automatic data cleanup mechanism (retaining data for one year by default). This ensures that sensitive information such as intra-abdominal pressure complies with medical privacy regulations during collection, transmission, and storage, preventing leakage or misuse.

[0057] In one embodiment, the intra-abdominal pressure measurement device based on AI provided in this application has an intra-abdominal pressure measurement range of 0-40 mmHg and an error of ≤±1.5 mmHg, meeting the standards for precise clinical monitoring. The sensor signal acquisition response time is ≤100ms, enabling real-time capture of minute fluctuations in intra-abdominal pressure. When powered only by the built-in battery, the device has a battery life of ≥72 hours; when equipped with the bioenergy harvesting module 50, the battery life is ≥30 days under normal activity conditions and ≥60 days under static monitoring conditions. These parameters enable the device to meet the core clinical needs for non-invasive, precise, and continuous 24 / 7 monitoring of intra-abdominal pressure in scenarios such as postoperative care for critically ill patients and chronic disease management. The signal conditioning circuit has an IP67 waterproof rating, allowing it to withstand short-term immersion. The wireless transmission module has a signal loss rate of ≤0.5% under 30V / m electromagnetic interference, making it suitable for scenarios with dense clinical medical equipment. In a study of 30 patients with intra-abdominal hypertension, the correlation between the measured values ​​of this device and the intravesical pressure measurement (clinical gold standard) was 0.95, with an average deviation of 0.8 mmHg, which is significantly better than existing non-invasive measurement devices.

[0058] To more intuitively and quantitatively demonstrate the comprehensive advantages of the technical solution presented in this application compared to existing non-invasive intra-abdominal pressure measurement technologies, the key performance parameters are compared in Table 1 below. This comparison covers multiple core dimensions, including measurement accuracy, clinical relevance, battery life, scenario adaptability, applicable population, sensor configuration, and algorithm intelligence, comprehensively reflecting the significant progress made by this application in achieving non-invasive, accurate, continuous, and comfortable monitoring.

[0059] Table 1. Advantages and disadvantages of the technical solution in this application compared with existing non-invasive intra-abdominal pressure measurement technologies. This application embodiment uses multiple types of sensors to collaboratively collect multi-dimensional data, combines AI intelligent algorithms to achieve dynamic data fusion and error correction, and features an ergonomic flexible carrier design to achieve non-invasive, real-time, and high-precision measurement of intra-abdominal pressure; it constructs a multi-scenario data interaction system to adapt to clinical and home monitoring needs; at the same time, it retains sensor integration and expandability, allowing the addition of other physiological parameter monitoring functions according to clinical needs, broadening the applicability of the device, reducing the operational burden on medical staff, and improving patient comfort and monitoring compliance.

[0060] Secondly, this application provides an AI-based multimodal flexible intra-abdominal pressure measurement method, applied to the AI-based multimodal flexible intra-abdominal pressure measurement device described above, comprising the following steps: Step S1: Wear the device on the patient's abdomen; Step S2: Activate the AI-based multimodal flexible intra-abdominal pressure measurement device, enabling the multimodal sensor module 20 to collect physiological signals. The physiological signals are processed by the signal conditioning and intelligent wireless transmission module 30 and then transmitted to the data processing module 40. Based on the received data, the data processing module 40 calculates the raw intra-abdominal pressure value through algorithm fusion and corrects it based on the input individual physiological parameters to obtain the final intra-abdominal pressure measurement value. The final intra-abdominal pressure measurement value is then displayed and stored on the smart terminal.

[0061] In one embodiment, such as Figure 4 As shown, in step S2, the data processing module 40 calculates the raw intra-abdominal pressure value based on the received data using an algorithm, and corrects it based on the input individual physiological parameters to obtain the final intra-abdominal pressure measurement value. Specifically, this includes the following steps: Step S21: Based on the real-time postural change data monitored by the accelerometer and the respiratory status and environmental acoustic features collected by the sound sensor, the current real-time monitoring scene is determined through a pre-trained AI scene classification model. The multimodal features include posture features extracted from the acceleration data and acoustic features extracted from the audio signal. The model is trained based on clinical sample data covering different body positions, respiratory states, age groups, and body mass index (BMI) to ensure its generalization ability and clinical applicability. Specifically, the model in this embodiment was trained using 500 ethically reviewed clinical samples, with ages ranging from 18 to 80 years, aiming to enable the model to accurately adapt to a wide range of people from youth to old age and different body types. Step S22: Based on the determined current real-time monitoring scenario, a predefined scenario weight mapping database is invoked. This database contains multiple composite scenarios formed by combinations of body position, respiratory state, and activity interference. For each scenario, corresponding weights are assigned according to the reliability of sensor data. Fusion weights are dynamically assigned to multimodal sensor data, and a weighted fusion algorithm is used to integrate and calculate the multi-source heterogeneous data from pressure, stress-strain, sound, acceleration, and blood oxygen saturation infrared sensors, outputting the raw value of intra-abdominal pressure. The integration calculation includes normalization of the multi-source heterogeneous sensor data based on the mean and standard deviation. Specifically, a typical monitoring scenario and its corresponding sensor weight allocation example are as follows: In a supine, stable breathing scenario, the pressure sensor weight is 0.4, the flexible fiber optic sensor unit weight is 0.3, and other sensors total 0.3; In a body position change scenario, the acceleration sensor weight is increased to 0.5, and the remaining sensors are allocated proportionally.

[0062] Step S23: Based on the individual physiological parameters pre-input by the user, a personalized correction coefficient is calculated using an AI individual difference compensation model. This AI individual difference compensation model is trained based on a machine learning algorithm, and its model fit meets the accuracy requirements of clinical measurement. It also compensates for the original intra-abdominal pressure value to obtain the final intra-abdominal pressure measurement value. In this embodiment, the AI ​​individual difference compensation model is trained based on a gradient boosting tree algorithm, with a model fit R² ≥ 0.92, and is verified to have no systematic bias by 100 test samples.

[0063] This embodiment dynamically allocates weights through an AI scene classification model, significantly improving measurement stability under different scenarios. Unlike existing AI medical devices that monitor only a single parameter, this application integrates the function of jointly monitoring intra-abdominal pressure and blood oxygen saturation. It can assist in judging the peritoneal blood perfusion status through multi-parameter correlation analysis, providing a more comprehensive reference for clinical diagnosis and treatment.

[0064] In one embodiment, step S21: Based on the body position change data monitored in real time by the accelerometer and the breathing state and environmental acoustic features collected by the sound sensor, the current real-time monitoring scene is determined by a pre-trained AI scene classification model. The multimodal features include posture features extracted from the acceleration data and acoustic features extracted from the audio signal, specifically including the following steps: Step S211: Synchronously collect and preprocess the three-axis attitude data from the accelerometer and the audio signal from the sound sensor, and extract multimodal features. The multimodal features include attitude features extracted from the acceleration data and acoustic features extracted from the audio signal. Specifically, the current real-time monitoring scene is determined by a pre-trained AI scene classification model. The multimodal features include attitude features extracted from the acceleration data and acoustic features extracted from the audio signal. The AI ​​scene classification model is trained based on a certain number of clinical sample data with different body positions and breathing states, covering different age groups and body mass index populations. set up For triaxial acceleration data Operators for extracting attitude and activity intensity features, To obtain from audio signals Operators for extracting acoustic features (such as MFCC and short-time energy) are used to obtain multimodal feature vectors. The extraction process can be represented as: in, and For the same timestamp Data collected synchronously below, This represents the concatenation operation of feature vectors; Step S212: Fuse and construct multimodal features into a standardized scene feature vector. The data is input into a pre-trained AI scene classification model for data processing to obtain the probability distribution, as shown in the following formula: in, and These are the feature mean vector and standard deviation vector calculated during the training phase, respectively, which are used to perform Z-score standardization on the fused features to eliminate the influence of dimensions. The standardized scene feature vector Input into the pre-trained AI scene classification model The model (e.g., a Softmax classifier) ​​outputs a probability distribution vector. , which represents the probability that the current data belongs to each predefined scenario.

[0065] in, The total number of predefined scene categories (e.g., supine, side-lying, active, etc.). Indicates that given input features Under these conditions, the model predicts the current scene to be a predefined scene. The probability; the probability distribution satisfies ; Step S213: Determine the current real-time monitoring scene based on the obtained probability distribution, specifically implemented as follows: Receive the probability distribution vector output from the AI ​​scene classification model. ,in This indicates that the current moment belongs to a predefined scenario. The probability of (e.g., "stable breathing while lying supine", "rapid breathing while lying on one's side", "changing position", "coughing interference") K The total number of predefined scene categories; From probability vector Find the maximum probability value and its corresponding current real-time monitoring scenarios Currently, real-time monitoring scenarios mainly include a variety of composite scenarios based on body position, respiratory status, and activity disturbances.

[0066] Specifically, assume that the predefined scene weight mapping database contains the following 5 typical composite scenes: C1: Supine position with steady breathing; C2: Breathing is stable while lying on one's side; C3: Supine position with rapid breathing; C4: During a change of body position (such as turning over); C5: Coughing interference exists; After analyzing the current data frame, the AI ​​scene classification model outputs a probability vector. : Maximum probability value The probability 0.75 is the second element in the vector. According to the predefined list, The corresponding current real-time monitoring scenario is Breathing is stable when lying on one's side.

[0067] This embodiment integrates body position change data provided by an accelerometer with respiratory status and environmental acoustic features collected by a sound sensor. It then uses a pre-trained AI scene classification model for real-time analysis and judgment, accurately identifying the patient's current monitoring scene (such as stable breathing while supine, activity while lying on their side, or interference from coughing). This provides crucial scene perception basis for subsequent dynamic weight fusion algorithms, effectively solving the problem of measurement environment complexity caused by body position changes, respiratory fluctuations, and external interference. It significantly improves the system's dynamic adaptability and intelligent discrimination capability to real application scenarios.

[0068] In one embodiment, step S22: Based on the determined current real-time monitoring scenario, a predefined scenario weight mapping database is invoked to dynamically allocate fusion weights to the multimodal sensor data, and a weighted fusion algorithm is used to integrate and calculate the multi-source heterogeneous data from pressure, stress-strain, sound, acceleration, and blood oxygen saturation infrared sensors, outputting the raw value of intra-abdominal pressure. The integration calculation includes normalization processing of the multi-source heterogeneous sensor data based on the mean and standard deviation, specifically including the following steps: Step S221: Based on the current real-time monitoring scenario, index and call the preset optimal sensor weight combination for that scenario in the predefined scenario weight mapping database. This weight combination is a weight vector. ,in These represent the fusion weights of the pressure sensor, stress-strain sensor, sound sensor, and flexible fiber optic sensor units in the current scenario. The sum of all weights is 1, i.e. .

[0069] Step S222: Normalize the multi-source heterogeneous sensor data received from the signal conditioning module after amplification and filtering to eliminate differences in the dimensions and magnitudes of different sensors, and obtain the normalized data vector; let the sensor data vector at a certain moment be... ,in: This is the pressure signal value. The strain signal value, For sound feature values, For the fiber optic sensing signal value, the normalization formula is: ; in, For the first i Normalized data vectors from sensor-like devices and The first i The mean and standard deviation of sensor-like data obtained during the training phase; Note: Accelerometer data has been used for scene classification. The weighted fusion calculation in this step mainly targets data from pressure, stress-strain, sound, and blood oxygen saturation infrared (flexible fiber) sensors.

[0070] Step S223: Normalize the data vector and weight vector of each sensor. The raw value of intra-abdominal pressure was obtained by performing a weighted summation. : .

[0071] This calculation process is performed in real time during each data acquisition cycle, ensuring that the output raw intra-abdominal pressure value dynamically reflects the continuous value of the current intra-abdominal pressure change.

[0072] This embodiment calls a predefined scene weight mapping database to dynamically allocate fusion weights for multimodal sensors based on the real-time determined monitoring scene, and uses a weighted fusion algorithm to integrate and calculate multi-source heterogeneous data. This effectively overcomes the measurement limitations of a single sensor in a specific scene, significantly improves the anti-interference ability, measurement stability and overall accuracy of the raw intra-abdominal pressure value in different dynamic scenes, and provides a highly reliable data foundation for subsequent personalized calibration.

[0073] In one embodiment, step S23: Based on the individual physiological parameters pre-input by the user, a personalized correction coefficient is calculated using an AI individual difference compensation model, and the original intra-abdominal pressure value is compensated to obtain the final intra-abdominal pressure measurement value. This specifically includes the following steps: Step S231: Input the user's individual physiological parameters through the patient information interface of a computer terminal or smartphone APP. These parameters mainly include abdominal wall thickness (obtained by ultrasound measurement, unit: cm), body mass index (BMI) (calculated from the input height and weight, unit: kg / m²), and abdominal wall elasticity coefficient (obtained by clinical palpation combined with standard scale assessment, a dimensionless parameter). Standardize and preprocess the input parameters to eliminate the influence of dimensions and form a standardized individual feature vector. ; Step S232: Invoke the pre-trained AI individual variability compensation model (this model is based on an individual parameter-measurement error correlation model trained on a large amount of clinical data), and convert the standardized individual feature vectors into... The input model outputs a personalized correction coefficient Δ (unit: mmHg), as shown in the following formula: in, For the intercept term, , , The values ​​were learned from clinical data by the model and correspond to abdominal wall thickness, respectively. BMI coefficient and abdominal wall elasticity coefficient The weighting parameter is ε, where ε is the random error term.

[0074] The correction factor Δ reflects the systematic measurement bias caused by the specific abdominal wall characteristics (such as thickness and elasticity) of the user. For example, if the model calculates Δ=+1.1mmHg, it means that for this specific patient, the original measurement value is generally about 1.1mmHg lower than the true value, and positive compensation is required. Step S233: Obtain the raw calculated value of intra-abdominal pressure from the AI ​​dynamic weighted fusion algorithm. (Unit: mmHg), compare the correction factor Δ with the original value of intra-abdominal pressure. The values ​​are added together to obtain the final intra-abdominal pressure measurement. : This embodiment utilizes an AI-based individual difference compensation model to transform user-input static physiological parameters into dynamic, personalized correction coefficients. This provides real-time compensation for the raw intra-abdominal pressure values ​​calculated from multimodal sensor fusion, effectively eliminating systematic measurement biases caused by individual differences in abdominal wall characteristics. This significantly improves the consistency and accuracy of measurement results across different patient groups, ensuring that the final output intra-abdominal pressure measurement meets clinical medical-grade accuracy standards. Clinical sample testing of this device has shown that the measurement results are highly correlated with the clinical gold standard, meeting clinical diagnostic reference requirements.

[0075] In one embodiment, this application provides an AI-based multimodal flexible intra-abdominal pressure measurement method, which further includes the following steps: Step S24: Based on the final intra-abdominal pressure measurement and historical trend data, the risk level is determined through a risk assessment model, and personalized intervention suggestions are generated from the clinical database, such as "suggesting a semi-recumbent position and limiting fluid intake", and the early warning information is pushed to the smart terminal in real time. Step S25: Based on the original static parameters, incorporate dynamic data such as real-time respiratory rate and muscle tension, and calculate correction coefficients using an AI model to achieve full-state adaptive compensation. For example, dynamically adjust the weight of the pressure sensor in conjunction with muscle tension signals to eliminate real-time physiological fluctuation errors.

[0076] It should be noted that the sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0077] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device to execute the AI-based multimodal flexible intra-abdominal pressure measurement method described in the various embodiments of this application.

[0078] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A multimodal flexible intra-abdominal pressure measurement device based on AI, characterized in that, include: A flexible wearable carrier with adjustable elastic fixing components and positioning markers; The multimodal sensor module, housed within the flexible wearable carrier, includes a pressure sensor, a stress-strain sensor, a sound sensor, an acceleration sensor, and a blood oxygen saturation infrared sensor. These sensors are used to acquire direct pressure signals from the abdominal wall in contact with the body surface, deformation signals of the abdominal wall caused by changes in intra-abdominal pressure, sound signals related to the movement of intra-abdominal organs, respiration, and disturbances, real-time body position and physical activity signals of the patient, and minute fluctuations in blood perfusion caused by changes in intra-abdominal pressure. The signal conditioning and intelligent wireless transmission module is electrically connected to the multimodal sensor module and includes a signal conditioning circuit, a multifunctional wireless transmission unit, and a miniature demodulation module, which are respectively used to amplify and reduce noise of the original sensor signal, send the processed data to the terminal device, and convert the spectral signal of the flexible fiber optic sensor unit into an electrical signal. The data processing module is configured to run an AI algorithm to perform fusion processing and intra-abdominal pressure calculation on the multi-source synchronous data collected by the multimodal sensor module, and obtain the final intra-abdominal pressure measurement value.

2. The AI-based multimodal flexible intra-abdominal pressure measurement device according to claim 1, characterized in that, The pressure sensors include piezoresistive pressure sensors and capacitive pressure sensors; The piezoresistive pressure sensor is used to stably capture static pressure signals from the abdominal wall. The capacitive pressure sensor is used to detect minute dynamic fluctuations in intra-abdominal pressure.

3. The AI-based multimodal flexible intra-abdominal pressure measurement device according to claim 1, characterized in that, The flexible wearable carrier is equipped with a standardized sensor expansion interface, which can be used to connect one or more of temperature sensors, electromyography sensors, and electrocardiogram sensors according to clinical needs. The adjustable elastic fixation component includes Velcro straps and medical hypoallergenic silicone tape; The location markers include fluorescent markers that are identifiable in low-light environments.

4. The AI-based multimodal flexible intra-abdominal pressure measurement device according to claim 1, characterized in that, The device further includes: The bioenergy harvesting module is electrically connected to the signal conditioning and intelligent wireless transmission module, and its internal components include a micro thermoelectric generator, a piezoelectric film, a power management unit, and an energy storage unit. The micro thermoelectric generator is used to generate electricity by utilizing the temperature difference between the abdominal wall and the environment. The piezoelectric film is used to generate electricity through body deformation. The power management unit is used to rectify, stabilize, and manage the output power of the micro thermoelectric generator and the piezoelectric film. The energy storage unit is used to store the managed energy to power the device. The bioenergy harvesting module is connected to the signal conditioning and intelligent wireless transmission module via a snap-on detachable structure, and the energy storage unit is a flexible solid-state lithium battery.

5. The AI-based multimodal flexible intra-abdominal pressure measurement device according to claim 1, characterized in that, The multi-functional wireless transmission unit includes one or more of Wi-Fi, Bluetooth, and mobile networks, supporting local offline data storage and automatic synchronization after network recovery; the signal conditioning and intelligent wireless transmission module also includes a cross-modal data synchronization calibration unit, which performs timing consistency calibration on the output signals of the multi-modal sensor module based on high-precision timestamp alignment technology and clock synchronization protocol.

6. The AI-based multimodal flexible intra-abdominal pressure measurement device according to claim 1, characterized in that, The data processing module includes: The scene classification unit is used to determine the current real-time monitoring scene based on the body position data of the accelerometer and the acoustic features of the sound sensor through a pre-trained AI scene classification model. The dynamic weight fusion unit is used to call a predefined scene weight mapping database according to the determined monitoring scene, dynamically allocate fusion weights to the multimodal sensor data, and calculate the original value of intra-abdominal pressure using a weighted fusion algorithm. The individual difference compensation unit is used to calculate personalized correction coefficients based on the input individual physiological parameters through a pre-trained AI individual difference compensation model, to compensate for the original intra-abdominal pressure value and obtain the final intra-abdominal pressure measurement value.

7. The AI-based multimodal flexible intra-abdominal pressure measurement device according to claim 1, characterized in that, The data processing module also includes a risk warning unit, which is used to determine the risk level of intra-abdominal hypertension based on the final intra-abdominal pressure measurement value and historical trend data through a pre-trained risk assessment model, generate personalized clinical intervention suggestions, and push abnormal warning information to the smart terminal. The device employs end-to-end encrypted transmission, hierarchical access control, distributed cloud backup, and a preset periodic automatic data cleanup mechanism to ensure the security of medical data.

8. A multimodal flexible intra-abdominal pressure measurement method based on AI, characterized in that, The method is applied to the AI-based multimodal flexible intra-abdominal pressure measurement device as described in any one of claims 1 to 7; the method includes the following steps: The device is aligned with the target area using the positioning marker and then worn and secured to the patient's abdomen. The device is activated, and the timing calibration of the multimodal sensors is completed through the cross-modal data synchronization calibration unit. Multi-dimensional physiological signals are synchronously acquired through the multimodal sensor module. The physiological signal is amplified, filtered, and noise-reduced by the signal conditioning and intelligent wireless transmission module, and then encrypted and transmitted to the data processing module. The data processing module uses AI algorithms to calculate and obtain the raw value of intra-abdominal pressure based on the received synchronous data, and then corrects it based on the input individual physiological parameters to obtain the final intra-abdominal pressure measurement value. The final intra-abdominal pressure measurement value is displayed in real time on the smart terminal, stored locally, and synchronized to the cloud management platform.

9. The AI-based multimodal flexible intra-abdominal pressure measurement method according to claim 8, characterized in that, The data processing module calculates the raw intra-abdominal pressure value based on the received data using an algorithm, and then corrects it based on the input individual physiological parameters to obtain the final intra-abdominal pressure measurement value. Specifically, this includes the following steps: Based on real-time monitoring of body position changes by an accelerometer and respiratory status and environmental acoustic features collected by a sound sensor, the current real-time monitoring scene is determined by a pre-trained AI scene classification model. The multimodal features include posture features extracted from acceleration data and acoustic features extracted from audio signals. Based on the determined current real-time monitoring scenario, a predefined scenario weight mapping database is invoked to dynamically assign fusion weights to the multimodal sensor data. A weighted fusion algorithm is then used to integrate and calculate the multi-source heterogeneous data from pressure, stress-strain, sound, acceleration, and blood oxygen saturation infrared sensors, and output the raw value of intra-abdominal pressure. The integration calculation includes normalization processing of the multi-source heterogeneous sensor data based on the mean and standard deviation. Based on the user's pre-input individual physiological parameters, a personalized correction coefficient is calculated through an AI individual difference compensation model, and the original intra-abdominal pressure value is compensated to obtain the final intra-abdominal pressure measurement value.

10. The AI-based multimodal flexible intra-abdominal pressure measurement method according to claim 8, characterized in that, The system uses real-time body position change data monitored by an accelerometer and respiratory status and environmental acoustic features collected by a sound sensor to determine the current real-time monitoring scene through a pre-trained AI scene classification model. The multimodal features include posture features extracted from the acceleration data and acoustic features extracted from the audio signal. Specifically, the system includes the following steps: The three-axis attitude data from the accelerometer and the audio signal from the sound sensor are acquired and preprocessed simultaneously to extract multimodal features, which include attitude features extracted from the acceleration data and acoustic features extracted from the audio signal. Multimodal features are fused together to construct a standardized scene feature vector, which is then input into a pre-trained AI scene classification model for data processing to obtain the probability distribution. The current real-time monitoring scenario is determined based on the obtained probability distribution.

Images