Severe patient diaphragm function track monitoring system and evaluation method

Abstract

The invention discloses a critical patient diaphragm function track monitoring system and evaluation method, and belongs to the technical field of intensive care medical treatment. The method comprises the following steps: synchronously acquiring an ultrasonic image data stream related to diaphragm of a patient and a respiratory mechanics data stream related to a breathing machine of the patient; based on the ultrasonic image data stream, analyzing and generating a first time sequence physiological parameter sequence in real time; calculating and generating a second time sequence physiological parameter sequence in real time based on the respiratory mechanics data flow; based on conjoint analysis of the first time sequence physiological parameter sequence and the second time sequence physiological parameter sequence, a diaphragm function reserve index used for representing the diaphragm function reserve state of the patient is generated. According to the method, synchronous fusion and dynamic analysis are carried out on the multi-modal physiological data, a composite index capable of objectively, continuously and prospectively quantifying the anti-fatigue capacity of the diaphragm muscle is constructed, and the problems that in the prior art, machine withdrawal evaluation depends on scattered and static indexes, judgment subjectivity is high, and cognitive loads are high are solved.

Description

Technical Field

[0001] This application relates to the field of medical monitoring technology, and in particular, to a system and assessment method for monitoring the diaphragmatic function trajectory of critically ill patients. Background Technology

[0002] In intensive care medicine, mechanical ventilation is a crucial means of maintaining vital signs in critically ill patients. However, prolonged mechanical ventilation can lead to various complications such as ventilator-associated pneumonia and diaphragmatic atrophy. Therefore, accurately determining whether a patient is ready to be weaned off the ventilator and seizing the optimal time for weaning is a core challenge in ICU clinical practice. Current technology mainly relies on clinicians to comprehensively assess a series of scattered physiological indicators such as respiratory rate, tidal volume, and maximum inspiratory pressure, possibly supplemented by bedside ultrasound to observe the diaphragmatic thickening fraction. This method has a fundamental technical contradiction: on the one hand, the data sources are abundant, with high-frequency ventilator data streams and dynamic ultrasound images providing an unprecedented amount of information; on the other hand, the decision-making process is lagging and subjective, requiring physicians to rely on personal experience to perform a high-cognitive-load "mental synthesis" of these temporally fragmented and dimensionally heterogeneous data, making it difficult to form an objective, continuous, and forward-looking quantitative judgment on the "reserve" and "fatigue trend" of diaphragmatic function. This results in limited accuracy and repeatability of weaning decisions, leading to a persistently high weaning failure rate. Summary of the Invention

[0003] The first aspect of this application provides a method for assessing the diaphragmatic function trajectory of critically ill patients, aiming to solve the technical problems in the prior art where the assessment of the timing of ventilator weaning relies on scattered and static indicators, resulting in strong subjectivity in judgment, high cognitive load, and lack of forward-looking predictive ability.

[0004] This application provides a method for assessing the diaphragmatic function trajectory of critically ill patients, comprising: simultaneously acquiring ultrasound image data streams related to a patient's diaphragm and respiratory mechanics data streams related to the patient's mechanical ventilator; based on the ultrasound image data streams, determining a first temporal physiological parameter sequence composed of multiple first physiological parameters characterizing the morphological dynamics of the diaphragm; based on the respiratory mechanics data streams, determining a second temporal physiological parameter sequence composed of multiple second physiological parameters characterizing the patient's respiratory work mechanics characteristics; and generating a diaphragmatic function reserve index to characterize the patient's diaphragmatic function reserve state based on joint analysis of the first temporal physiological parameter sequence and the second temporal physiological parameter sequence within a preset time window.

[0005] Optionally, according to the aforementioned scheme, the first physiological parameter includes instantaneous diaphragm thickness and a diaphragm thickening fraction derived from the instantaneous diaphragm thickness over a respiratory cycle.

[0006] Optionally, according to the aforementioned scheme, determining the first temporal physiological parameter sequence includes: applying a semantic segmentation model to each image frame in the ultrasound image data stream to identify and segment a binary mask of the diaphragm region; obtaining the instantaneous diaphragm thickness based on the binary mask and a preset pixel-to-millimeter calibration scale; and obtaining the diaphragm thickening score based on a preset baseline diaphragm thickness and the peak value of the instantaneous diaphragm thickness within a respiratory cycle.

[0007] Optionally, according to the aforementioned scheme, the second physiological parameter includes the pressure-time product and the work of breathing.

[0008] Optionally, according to the aforementioned scheme, generating the diaphragm function reserve index includes: obtaining a work efficiency component based on the first time-series physiological parameter sequence and the second time-series physiological parameter sequence; obtaining an anti-fatigue ability component based on the change trend of the first time-series physiological parameter sequence within the preset time window; obtaining a recovery elasticity component based on the morphological recovery characteristics corresponding to exhalation in each respiratory cycle in the first time-series physiological parameter sequence; and generating the diaphragm function reserve index based on a weighted combination of the work efficiency component, the anti-fatigue ability component, and the recovery elasticity component.

[0009] Optionally, according to the aforementioned scheme, obtaining the work efficiency component includes: obtaining the correlation between the diaphragm thickening fraction and the pressure-time product within the preset time window; obtaining the fatigue resistance component includes: obtaining the absolute value of the linear regression slope of the diaphragm thickening fraction changing with time within the preset time window, and normalizing the absolute value; obtaining the recovery elasticity component includes: for the expiratory phase of each respiratory cycle within the preset time window, obtaining the time constant of the instantaneous diaphragm thickness decaying with time, and normalizing the reciprocal of the mean of the time constant.

[0010] Optionally, according to the aforementioned scheme, the normalization of the absolute value includes: determining a normalization mapping relationship based on the statistical distribution of the absolute values ​​of the linear regression slope of the diaphragm thickening score over time in a preset reference patient database, and applying the normalization mapping relationship for processing.

[0011] Optionally, according to the aforementioned scheme, the normalization process of the reciprocal of the mean of the time constant includes: determining a normalization mapping relationship based on the statistical distribution of the reciprocal of the mean of the time constant of the instantaneous diaphragm thickness decaying over time in a preset reference patient database, and applying the normalization mapping relationship for processing.

[0012] Optionally, according to the foregoing scheme, before the synchronous acquisition, the method further includes: in response to a calibration command, acquiring and storing a pixel-to-millimeter calibration scale for the resolution of the ultrasound image data stream; and in a baseline calibration phase, acquiring and storing a baseline diaphragm thickness for calculating the diaphragm thickening fraction.

[0013] Optionally, according to the aforementioned scheme, before determining the second time-series physiological parameter sequence, the method further includes: adaptively setting a noise threshold based on the respiratory mechanics data stream collected in the early stage of monitoring; and performing data quality verification on the subsequently collected respiratory mechanics data stream based on the noise threshold, and discarding data whose quality does not meet the preset conditions.

[0014] A second aspect of this application provides a monitoring system for the diaphragmatic function trajectory of critically ill patients, characterized by comprising: a multimodal synchronous data acquisition module configured to synchronously acquire ultrasound image data streams related to a patient's diaphragm and respiratory mechanics data streams related to the patient's mechanical ventilator; an ultrasound image real-time analysis module configured to determine a first temporal physiological parameter sequence composed of multiple first physiological parameters characterizing the morphological dynamics of the diaphragm based on the ultrasound image data stream; a respiratory mechanics parameter calculation module configured to determine a second temporal physiological parameter sequence composed of multiple second physiological parameters characterizing the respiratory work mechanics characteristics of the patient based on the respiratory mechanics data stream; and a functional reserve index calculation module configured to generate a diaphragmatic function reserve index characterizing the diaphragmatic function reserve state of the patient based on joint analysis of the first temporal physiological parameter sequence and the second temporal physiological parameter sequence within a preset time window.

[0015] Optionally, according to the foregoing scheme, the functional reserve index calculation module is further configured to: obtain a work efficiency component based on the first time-series physiological parameter sequence and the second time-series physiological parameter sequence; obtain an anti-fatigue ability component based on the change trend of the first time-series physiological parameter sequence within the preset time window; obtain a recovery elasticity component based on the morphological recovery characteristics corresponding to exhalation in each respiratory cycle in the first time-series physiological parameter sequence; and generate the diaphragm functional reserve index based on a weighted combination of the work efficiency component, the anti-fatigue ability component, and the recovery elasticity component.

[0016] The beneficial effects of this application are as follows: This application constructs a novel, multimodal, fusion-based biodynamic composite index—the Diaphragmatic Functional Reserve Index (DFRI)—transforming the subjective judgment process, which relies on clinical experience, into an algorithm-driven, reproducible, and objective quantitative assessment process. This index forcibly synchronizes and nonlinearly fuses separate diaphragmatic morphological data with respiratory work mechanics data over time. The resulting single, continuous assessment value directly quantifies the diaphragm's fatigue resistance and recovery efficiency, providing a more reliable basis for weaning decisions than any single, static indicator, and thus improving the weaning success rate.

[0017] This application interprets high cognitive load analysis of multi-source, heterogeneous, and dynamic data streams, compressing it into a single, intuitive reserve index through a white-box physiological model. Clinicians no longer need to spend time comparing ultrasound images and ventilator parameters at different time points; instead, they can directly observe the real-time value and trend of the DFRI, greatly simplifying the decision-making process and saving valuable clinical time and effort.

[0018] Compared to existing technologies that mostly rely on post-hoc assessments (such as observing fatigue symptoms like shortness of breath), DFRI (Deferred Action for Breathing) possesses trend prediction capabilities by analyzing the slope and efficiency of parameter changes over a continuous time window. It can advance the assessment from when a patient is already fatigued to when their reserve is declining at a specific rate, predicting they will enter the fatigue zone at a specific time in the future. This provides early warning, allowing physicians to halt immature spontaneous breathing trials before patients show obvious signs of fatigue, thus avoiding harm to the patient. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the functional modules of a diaphragmatic function trajectory monitoring system for critically ill patients provided in an embodiment of this application.

[0021] Figure 2 This is a flowchart illustrating a method for assessing the diaphragmatic function trajectory of critically ill patients, provided in one embodiment of this application.

[0022] Figure 3 A detailed flowchart illustrating the method for calculating the diaphragmatic functional reserve index provided in an embodiment of this application.

[0023] Figure 4This is a schematic diagram of a monitoring system visualization and clinical decision support interface provided in an embodiment of this application.

[0024] Figure 5 This is a schematic diagram of the hardware implementation of a multimodal synchronous data acquisition module provided in an embodiment of this application. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0026] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0027] This application discloses an assessment system for the diaphragmatic function trajectory of critically ill patients. Its architecture is configured to include a multimodal synchronous data acquisition module implemented as a hardware interface and firmware logic, a real-time ultrasound image analysis module, a respiratory mechanics parameter calculation module, and a functional reserve index calculation module. The system is configured to process raw data streams from wearable ultrasound devices and mechanical ventilators. By constructing a composite index that nonlinearly fuses morphological and mechanics parameters, it solves the technical problem in existing technologies where clinical weaning assessment relies on dispersed, heterogeneous, and asynchronous physiological indicators, leading to high cognitive load and strong subjectivity in the assessment process. This achieves the effect of providing objective and quantitative decision-making basis for ICU weaning decisions.

[0028] The technical solution of this application is anchored in a Diaphragm Functional Reserve Index (DFRI) calculation model, which is structured to transform separate, multi-source physiological data streams (morphological and mechanical) into a comprehensive, interpretable quantitative index. The model's operational principle is based on a physiological analogy, where the diaphragm is considered a biomechanical engine, and the DFRI is defined as a quantitative assessment of the engine's overall performance. The model's architecture includes three parallel computational branches, each corresponding to a core physiological dimension. The first computational branch generates a Work Efficiency Score (WES), which quantifies the engine's mechanical efficiency, i.e., assesses the coupling strength between unit respiratory power consumption and the resulting diaphragmatic morphological changes. The second computational branch generates a Fatigue Resistance Score (FRS), which quantifies the rate of decay of the engine's output capacity under sustained load. The third computational branch generates a Recovery Elasticity Score (RES), which quantifies the engine's state recovery rate after a single work-relaxation cycle. The outputs of these three branches are then weighted and combined by a fusion unit to generate the final DFRI.

[0029] Based on this core model architecture, the method and process provided in this application are as follows: Figure 2 As shown, it includes the following steps: S100: Simultaneously acquire ultrasound imaging data streams related to a patient's diaphragm and respiratory mechanics data streams related to the patient's mechanical ventilator.

[0030] In one embodiment, a multimodal synchronous data acquisition module, configured as the system's physical data interface layer, performs this step. This module is implemented as a PCIe data acquisition card integrating at least one DB9 serial port conforming to the EIA-232 standard for establishing a serial communication link with the mechanical ventilator, and a USB 3.1 Gen 2 Type-C interface supporting the USB VideoClass (UVC) v1.5 protocol for receiving raw video streams from the ultrasound equipment. All external data received via the physical interface is routed to an onboard Field-Programmable Gate Array (FPGA) chip before being written to the system main memory. This FPGA is programmed to act as a clock synchronization arbiter, internally running a free-oscillating, nanosecond-resolution counter. In response to a hardware interrupt upon receiving a complete data frame (e.g., an ultrasound image frame or a ventilator data packet) from either interface, the FPGA is configured to immediately capture the current 64-bit counter value and inject this count value as a hardware timestamp into the metadata header of the data frame. Because the timestamp appending is done at the hardware level, before the operating system kernel intervenes, this mechanism ensures that the timestamp error between the two independent data streams is controlled within a single clock cycle of the FPGA, thereby achieving sub-millisecond synchronization accuracy. The module outputs two logically independent data streams, but with strong temporal consistency achieved through the aforementioned hardware timestamps: the raw ultrasonic data stream. and ventilator raw data stream These two data streams are then encapsulated in a unified data frame structure and transmitted at high speed to a designated buffer in the system main memory via a direct memory access (DMA) channel for subsequent parsing by software modules.

[0031] S200: Based on the ultrasound image data stream, determine a first temporal physiological parameter sequence consisting of multiple first physiological parameters characterizing the morphological dynamics of the diaphragm.

[0032] A real-time ultrasound image analysis module, whose core computational tasks are offloaded to a graphics processing unit (GPU), is configured to perform this step. The module's function is to automatically and in real-time analyze ultrasound images. Quantitative morphological information is extracted. Its internal processing pipeline includes the following sub-steps: S210: For each image frame in the ultrasound image data stream, apply a semantic segmentation model to identify and segment the binary mask of the diaphragm region.

[0033] Deployed on the GPU is a deep convolutional neural network for semantic segmentation of the diaphragm region. The network's topology is based on an encoder-decoder paradigm. In one specific implementation, the encoder uses a ResNet-101 backbone network with integrated dilated convolutional layers, while the decoder employs a multi-scale feature fusion structure. This configuration exponentially increases the receptive field of feature extraction without reducing spatial resolution, enabling the network to utilize global contextual information to distinguish diaphragmatic tissue from acoustic artifacts. The original image frame input to this module is represented as a three-dimensional tensor of (height, width, channels). The module is configured to process this input tensor and output a single-channel binary mask tensor with the same spatial dimensions as the input. In this output tensor, pixel locations with a value of 1 correspond to regions identified by the network as the diaphragm, while pixel locations with a value of 0 correspond to the background. This inference process is accelerated using NVIDIA's TensorRT engine to ensure that single-frame processing latency meets clinical real-time requirements.

[0034] S220: Based on the binary mask and a preset pixel-to-millimeter calibration scale, obtain the instantaneous diaphragm thickness.

[0035] Based on the binary mask generated by S210, a morphological processing unit is triggered. This unit executes a Zhang-Suen parallel thinning algorithm to extract a single-pixel-width region skeleton by iteratively stripping away the mask boundary pixels. Subsequently, a geometric analysis subroutine is called, configured to traverse each pixel coordinate on the skeleton and determine the mask boundary along the local normal direction at that coordinate using a scanline algorithm, thereby calculating a series of pixel width values. These discrete width values ​​are fed into a statistical filter and arithmetically averaged to generate a robust average pixel width value. This value is related to a calibration parameter stored in the global configuration register during the system initialization phase (S010). (Unit: pixels / mm) Perform floating-point division. Through this operation, pixel spatial measurements are precisely mapped to physical spatial thickness. (Unit: millimeters), thereby eliminating equipment dependence.

[0036] S230: Based on a preset baseline diaphragm thickness and the peak value of the instantaneous diaphragm thickness within a respiratory cycle, obtain the diaphragm thickening score.

[0037] Once a complete respiratory cycle has been defined, a diaphragmatic thickening fraction (TFdi) calculation unit is activated. This unit retrieves the global baseline parameters for the current patient, determined during the baseline calibration phase (S020), from memory. And generated by S220 during the current respiratory cycle. Peak value of time series Based on these two input floating-point numbers, the unit performs standardized, formulaic calculations: The output of this calculation is a dimensionless percentage value. This value, along with the corresponding cycle end timestamp, is packaged together and output as a data point of the first time-series physiological parameter sequence.

[0038] S300: Based on the respiratory mechanics data stream, determine a second time-series physiological parameter sequence consisting of multiple second physiological parameters characterizing the respiratory work mechanics of the patient.

[0039] A respiratory mechanics parameter calculation module, whose computational task runs on the host CPU, is configured to perform this step. This module is responsible for... The extraction of work-mechanical parameters, its internal logic includes: S305: Perform a data quality check before performing any calculations.

[0040] This module includes a data quality verification subunit that performs statistical analysis on the input airway pressure and flow signals within the first 5 minutes of the monitoring task initiation. Specifically, a statistical analysis submodule loads this initial phase data stream into a buffer and calculates its baseline standard deviation. and The calculation result is multiplied by a preset coefficient (e.g., 5) and stored in a dedicated register as a dynamic noise threshold for subsequent real-time verification. For each subsequent data packet, the verification subunit calculates its standard deviation within a 1-second sliding window. If this value exceeds the corresponding register threshold, the quality flag of the data packet is set to "invalid." This mechanism is used to filter out transient noise caused by coughing or sensor artifacts. Data packets marked as "invalid" are discarded in subsequent calculations.

[0041] S310: The respiratory cycle is segmented using a threshold cross-validation method based on the flow signal, and the pressure-time product (PTP) is calculated.

[0042] The data stream that passes quality verification is fed into a respiratory cycle segmentation unit. This unit is configured to continuously monitor the flow signal. .when When the signal value changes from negative to positive and exceeds the threshold of +2 L / min, the unit records the current timestamp as the inhalation start time. .when When the signal falls back from a positive value and crosses zero for the first time, the unit records the current timestamp as the end of inhalation. The intake window is defined based on these two timestamps. , A PTP calculation unit is triggered. This unit uses positive end-expiratory pressure (PEEP) as the pressure baseline. By analyzing the airway pressure sequence within the window Numerical integration is performed using the trapezoidal rule to calculate the pressure-time product: (Only for) (The part is integrated). Each respiratory cycle thus generates one unit of cmH2O·s. value.

[0043] S320: Calculate the work of breathing (WOB).

[0044] For the same intake window defined by S310, a WOB calculation unit is activated. This unit first processes the flow signal within the window. Perform time integration to reconstruct the change in tidal volume over time. Based on this, a two-dimensional pressure-volume (PV) relationship is established. The unit then numerically integrates this PV relationship during the inspiratory phase to calculate the work of breathing: This calculation generates one unit of joules for each respiratory cycle. value.

[0045] S400: Construct the diaphragm functional trajectory.

[0046] A functional trajectory construction module instantiates a circular data queue of fixed length N (e.g., N=4800, corresponding to 4 hours of data capacity) within a specific address space of the system's main memory. This data structure is used as the Trajectory_Buffer. The module is configured to subscribe to the output of modules S200 and S300. In response to receiving a complete dataset of a new cycle (a structure containing TFdi, PTP, WOB, and their timestamps), the module performs an enqueue operation, writing the structure to the memory location pointed to by the queue's write pointer, and then incrementing the write pointer. In a conditional judgment, if the write pointer coincides with the read pointer, the read pointer is also incremented, thereby achieving logical overwrite of the oldest data element. This O(1) time complexity operation ensures that the buffer always maintains historical data from the most recent N breathing cycles, providing a data foundation for subsequent sliding window analysis.

[0047] S500: Calculate the diaphragmatic functional reserve index (DFRI).

[0048] The computation of DFRI is scheduled as a periodic task, triggered, for example, once per minute. Each time it is triggered, a sliding window manager extracts a subset of data from the Trajectory_Buffer corresponding to the past 30 minutes, and this subset is distributed to three parallel computation units: S510: Obtain Work Efficiency Components .

[0049] A productivity calculation unit receives the TFdi and PTP sequences from the data subset. A Pearson correlation coefficient calculation function is invoked, taking these two sequences as input and outputting a dimensionless pure number between -1 and +1. This number is directly assigned to the productivity component. This is used to quantify the linear coupling between the work done by the diaphragm and the thickening of its morphology.

[0050] S520: Obtaining fatigue resistance components .

[0051] A fatigue resistance calculation unit receives the TFdi time series from the data subset. This unit applies the least squares method to perform linear regression on the series to calculate the slope of the regression line. (Unit: %·min⁻¹). Based on the absolute value of this slope. This unit applies a predefined normalized mapping relation for processing. This mapping relation is implemented as a linear normalization function. Its parameters and It is done by analyzing a pre-defined reference patient database. The values ​​were pre-calibrated by analyzing their statistical distribution (using the 5th and 95th percentiles, respectively). This process maps the original slope values ​​to a dimensionless, standardized fatigue resistance component. .

[0052] S530: Obtain the restoring elasticity component .

[0053] A recovery elasticity calculation unit calculates the expiratory phase of each respiratory cycle in the data subset. The curves are processed. For each curve, the unit employs the Levenberg-Marquardt nonlinear least squares algorithm to apply an exponential decay function. The fitting was used to solve for the time constant. (Unit: s). After processing all respiratory cycles within the window, the obtained series of... The value is the reciprocal of its mean. Similar to the S520, this unit applies a normalized mapping relationship based on statistical calibration of a reference database to... The value is converted into a dimensionless, standardized restoring elastic component. .

[0054] S540: Perform weighted summation to generate the final DFRI score.

[0055] Finally, the three dimensionless components generated by S510, S520, and S530 , , It is organized into a three-dimensional feature vector. This vector is then compared with a pre-defined weight vector, calibrated through offline regression analysis. Perform the dot product operation. The result of the dot product, a floating-point number between 0 and 1, is multiplied by 100 by a scalar multiplier to complete the scaling transformation to the range of 0-100, generating the final DFRI score.

[0056] In addition to the core method and process described above, this application also includes some important auxiliary and initialization steps: S000: System initialization and equipment calibration.

[0057] This step is performed when the system is first started or when a new external device is connected, and is designed to ensure measurement accuracy and device compatibility.

[0058] S010: Acquire and calibrate the pixel-to-millimeter calibration scale.

[0059] To convert pixels measured from ultrasound images to the standard physical unit of millimeters, the system must know the imaging scale of the current ultrasound equipment. In one embodiment, the system has a built-in database containing a list of mainstream brands and models of ultrasound equipment, along with their default probes and settings. The system automatically loads the corresponding scale parameters when the operator first starts the system, selecting the currently connected device model from a drop-down list in a configuration interface. If the device is not in the list, or if the operator wishes to perform more precise on-site calibration, the system will initiate a manual, interactive calibration procedure. This procedure prompts the operator to image an object of known physical length (e.g., a dedicated phantom with a built-in 10mm length scale) using an ultrasonic probe. Then, on the displayed image, the operator marks the two endpoints of the object of known length using a mouse or touchscreen. The system automatically calculates the pixel distance between these two marks and divides it by the known physical length (10mm) to accurately calculate the current device and settings. The value is then stored as a global calibration parameter. This step addresses the feasibility of the solution under different ultrasound equipment environments and ensures... The foundation of measurement accuracy.

[0060] S020: Perform patient baseline calibration.

[0061] To accurately calculate TFdi, a reliable baseline value for end-expiratory diaphragmatic thickness is required. Before commencing long-term monitoring, the system guides the operator through a baseline calibration procedure. The operator first places the wearable ultrasound patch or traditional probe at the intercostal space where the diaphragm is most clearly visualized (usually the 8th-11th intercostal space on the right side, near the anterior axillary line), based on anatomical landmarks. Then, with the patient in a stable assisted ventilation mode (such as A / C mode) without significant spontaneous breathing effort, the system begins a 1-minute baseline data acquisition. During this time, the system automatically identifies the end-expiratory phase of each respiratory cycle and records the instantaneous diaphragmatic thickness at that moment. After 1 minute, the system averages all recorded end-expiratory thickness values ​​and uses this average as the patient's personalized baseline diaphragmatic thickness. Store this personalized baseline value. This personalized baseline will serve as the reference for all subsequent TFdi calculations, and is more stable and accurate than using a fixed value or the initial measured value.

[0062] In addition, this application also provides a monitoring system for the diaphragmatic function trajectory of critically ill patients, such as... Figure 1 As shown, the system is the physical entity that performs the methods described above. In one embodiment, the system may include: A multimodal synchronous data acquisition module M100, the structure and function of which are detailed in step S100.

[0063] The structure and function of the real-time ultrasound image analysis module M200 are detailed in step S200.

[0064] The structure and function of the respiratory mechanics parameter calculation module M300 are detailed in step S300.

[0065] The Functional Trajectory and Reserve Index (DFRI) calculation module M400, the structure and function of which are detailed in steps S400 and S500.

[0066] The M500 is a visualization and clinical decision support module. This module serves as the system's human-computer interface, responsible for presenting the complex results calculated by the M400 to clinicians in an intuitive and easy-to-understand manner. This module typically runs on the display screen of a bedside monitor or central workstation. Its functions may include: Trend plotting: Time series of diaphragmatic function trajectory received from M400 and DFRI time series The system utilizes a graphical user interface library (such as Qt Charts or D3.js) to plot real-time trends of key parameters like TFdi, PTP, and DFRI over time. These graphs help doctors visually observe the trajectory and trend of changes in patient function; for example, whether the DFRI curve is stable, rising, or slowly declining. The interface should also provide zoom and pan functions so doctors can view detailed data for any given time period.

[0067] Dashboard Display and Alarms: To provide immediate, atomized decision support, this module displays the latest DFRI value on a prominent, speedometer-like digital dashboard. The dashboard uses different color zones (e.g., green for the safe zone, yellow for the observation zone, and red for the danger zone) to represent the different clinical meanings of the DFRI value. Simultaneously, the system allows clinicians or departments to set a clinical alarm threshold based on their experience and patient group characteristics. The determination of this threshold should not be based on subjective assumptions, but rather on a white-box definition using statistical methods. For example, a preliminary clinical dataset containing at least 50 local patients can be collected, and the DFRI value for each patient before the spontaneous breathing trial can be calculated. Then, using "successful weaning" as the gold standard, a receiver operating characteristic (ROC) curve can be plotted. The DFRI value that maximizes the Youden index (sensitivity + specificity - 1) is selected as the optimal alarm threshold recommended by the clinical center. For example, the threshold calculated using this method might be 42.5. When the real-time DFRI value received by the M500 falls below this preset threshold, the system will immediately trigger an audible and visual alarm, such as changing the dashboard background color to flashing red and emitting a warning sound to alert medical staff that the spontaneous breathing test may need to be stopped.

[0068] The implementation of this application, through the aforementioned methods and systems, transforms the originally complex, fragmented, and highly experience-dependent process of assessing the timing of diaphragmatic weaning into an automated, data-driven decision support process with objective quantitative indicators. By deeply integrating ultrasound imaging and respiratory mechanics information, it creatively proposes the core indicator DFRI, which not only reflects the instantaneous state of the diaphragm but, more importantly, reveals its functional trajectory and reserve depletion trend under continuous load. This demonstrates significant application value in improving the quality of medical decision-making and reducing clinical risks.

[0069] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit described above can be implemented in hardware.

[0070] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for assessing the functional trajectory of the diaphragm in critically ill patients, characterized in that, include: Simultaneously acquire ultrasound imaging data streams related to a patient's diaphragm and respiratory mechanics data streams related to the patient's mechanical ventilator; Based on the ultrasound image data stream, a first temporal physiological parameter sequence consisting of multiple first physiological parameters characterizing the morphological dynamics of the diaphragm is determined. Based on the respiratory mechanics data stream, a second time-series physiological parameter sequence consisting of multiple second physiological parameters characterizing the patient's respiratory work mechanics characteristics is determined. Based on the joint analysis of the first and second time-series physiological parameter sequences within a preset time window, a diaphragmatic function reserve index is generated to characterize the diaphragmatic function reserve status of the patient.

2. The method according to claim 1, characterized in that, The first physiological parameter includes instantaneous diaphragmatic thickness and the diaphragmatic thickening fraction derived from the instantaneous diaphragmatic thickness over a respiratory cycle.

3. The method according to claim 2, characterized in that, Determining the first time-series physiological parameter sequence includes: For each image frame in the ultrasound image data stream, a semantic segmentation model is applied to identify and segment the binary mask of the diaphragm region; The instantaneous diaphragm thickness is obtained based on the binary mask and a preset pixel-to-millimeter calibration scale. The diaphragm thickening score is obtained based on a preset baseline diaphragm thickness and the peak value of the instantaneous diaphragm thickness within a respiratory cycle.

4. The method according to claim 1, characterized in that, The second physiological parameter includes the pressure-time product and the work of breathing.

5. The method according to claim 1, characterized in that, The generation of the diaphragm functional reserve index includes: Based on the first time-series physiological parameter sequence and the second time-series physiological parameter sequence, a work efficiency component is obtained; Based on the changing trend of the first time-series physiological parameter sequence within the preset time window, a fatigue resistance component is obtained. Based on the morphological recovery characteristics corresponding to exhalation in each respiratory cycle in the first time-series physiological parameter sequence, a recovery elasticity component is obtained. The diaphragm function reserve index is generated based on a weighted combination of the work efficiency component, the fatigue resistance component, and the recovery elasticity component.

6. The method according to claim 5, characterized in that, The step of obtaining the work efficiency component includes: obtaining the correlation between the diaphragm thickening fraction and the pressure-time product within the preset time window; The step of obtaining the anti-fatigue ability component includes: obtaining the absolute value of the linear regression slope of the diaphragm thickening fraction changing with time within the preset time window, and normalizing the absolute value. The step of obtaining the restorative elasticity component includes: for the expiratory phase of each respiratory cycle within the preset time window, obtaining the time constant of the instantaneous diaphragm thickness decaying over time, and normalizing the reciprocal of the mean of the time constant.

7. The method according to claim 6, characterized in that, The normalization process for the absolute value includes: determining a normalization mapping relationship based on the statistical distribution of the absolute values ​​of the linear regression slope of the diaphragm thickening score over time in a preset reference patient database, and applying the normalization mapping relationship for processing.

8. The method according to claim 6, characterized in that, The normalization process for the reciprocal of the mean of the time constant includes: determining a normalization mapping relationship based on the statistical distribution of the reciprocal of the mean of the time constant of the instantaneous diaphragm thickness decaying over time in a preset reference patient database, and applying the normalization mapping relationship for processing.

9. The method according to claim 1, characterized in that, Prior to the synchronous acquisition, the following is also included: In response to a calibration command, a pixel-to-millimeter calibration scale for the resolution of the ultrasound image data stream is acquired and stored; In a baseline calibration phase, baseline diaphragm thickness is acquired and stored for calculating the diaphragm thickening fraction.

10. A monitoring system for the diaphragmatic function trajectory of critically ill patients, characterized in that, include: A multimodal synchronous data acquisition module is configured to synchronously acquire ultrasound image data streams related to a patient's diaphragm and respiratory mechanics data streams related to the patient's mechanical ventilator; An ultrasound image real-time analysis module is configured to determine a first temporal physiological parameter sequence consisting of multiple first physiological parameters characterizing the morphological dynamics of the diaphragm based on the ultrasound image data stream. A respiratory mechanics parameter calculation module is configured to determine a second time-series physiological parameter sequence consisting of multiple second physiological parameters characterizing the respiratory work mechanics characteristics of the patient, based on the respiratory mechanics data stream. A functional reserve index calculation module is configured to generate a diaphragmatic functional reserve index to characterize the diaphragmatic functional reserve status of the patient based on the joint analysis of the first time-series physiological parameter sequence and the second time-series physiological parameter sequence within a preset time window.

Images