Intraoperative patient body temperature distribution monitoring method and system based on infrared image

Abstract

The invention relates to the technical field of medical infrared image processing and intraoperative body temperature monitoring, and discloses an intraoperative patient body temperature distribution monitoring method and system based on an infrared image. The method comprises the steps of infrared thermal imaging image acquisition and temperature calibration preprocessing, human body exposed body surface area intelligent segmentation and interference heat source elimination, anatomical area temperature field characteristic parameter extraction and analysis and intraoperative low body temperature risk dynamic evaluation and early warning. The limitation of traditional single-point temperature measurement is overcome, and adaptive monitoring parameter adjustment is realized based on a closed-loop feedback mechanism.

Description

Technical Field

[0001] This invention relates to the field of medical infrared image processing and intraoperative body temperature monitoring technology, and in particular to a method and system for monitoring the distribution of patient body temperature during surgery based on infrared images. Background Technology

[0002] Intraoperative hypothermia is a very common adverse event during surgery, typically defined as a core body temperature below 36°C. Numerous clinical studies have shown that approximately 50% to 70% of patients undergoing surgery experience hypothermia to varying degrees during the procedure, particularly in major surgeries, prolonged surgeries, and surgeries on infants and elderly patients. Intraoperative hypothermia can lead to a series of serious clinical consequences, including increased intraoperative blood loss due to coagulation disorders, delayed anesthesia recovery due to decreased drug metabolism rates, increased postoperative infection risk due to immunosuppression, and increased oxygen consumption due to shivering. Therefore, accurate and timely intraoperative temperature monitoring is crucial for perioperative patient safety management.

[0003] Currently, commonly used intraoperative temperature monitoring methods in clinical practice mainly include single-point contact temperature measurement methods such as esophageal temperature probes, nasopharyngeal temperature probes, tympanic membrane infrared thermometers, and zero-heat-flow temperature sensors. While these traditional methods can provide core body temperature or temperature values ​​for a specific location, they are all single-point measurement modes, reflecting only the temperature information of a localized area of ​​the patient's body and failing to comprehensively represent the spatial distribution characteristics of the patient's entire body surface temperature during surgery. In actual surgical scenarios, due to factors such as surgical area exposure, intraoperative irrigation, continuous illumination by surgical lights, and uneven coverage of warming measures, temperature changes in different parts of the patient's body often exhibit significant spatial heterogeneity; that is, some areas may already be hypothermic while other areas remain within the normal range. Traditional single-point temperature measurement methods struggle to capture this spatial difference in temperature distribution, leading to inaccurate assessment of the location and severity of intraoperative hypothermia, and thus limiting the targeted nature of warming interventions.

[0004] More importantly, intraoperative hypothermia typically follows a pathological process that spreads gradually from the periphery to the core. Specifically, the extremities and lower limbs often show the first temperature drop, followed by a gradual expansion of the hypothermic area into the core trunk. However, traditional temperature measurement methods often place the probe in the esophagus, nasopharynx, or forehead—core or near-core locations. This monitoring strategy essentially confirms a hypothermic event after it has progressed to a certain stage, making it difficult to issue an early warning when the peripheral regions first show a temperature drop, thus missing a valuable window for early intervention. Therefore, a comprehensive temperature distribution monitoring technology that can simultaneously cover both the core and peripheral regions and track the dynamic expansion of the hypothermic area is of significant clinical importance for early warning and proactive prevention of intraoperative hypothermia.

[0005] Infrared thermal imaging, as a non-contact temperature field measurement method, can acquire two-dimensional temperature distribution information by detecting infrared radiation emitted from the surface of an object, and has been widely used in the medical field. For example, Chinese invention patent application CN110298902A discloses a medical infrared image reconstruction method. This method involves applying cold stimulation to the area to be measured, then continuously acquiring dynamic image sequences during the temperature recovery process using an infrared thermal imager. The ratio of the temperature recovery rate to the cooling rate of each pixel is calculated to obtain an activity ratio image, thereby improving the contrast and contour clarity of the infrared image. However, this method is mainly designed for static in vitro detection scenarios, requiring pre-treatment with cold stimulation and the subject to remain stationary during acquisition. It is unsuitable for dynamic intraoperative monitoring scenarios in operating rooms where patients are under anesthesia and their bodies are continuously affected by surgical procedures. Furthermore, this method does not address the identification and elimination of interference heat sources specific to the operating room, nor does it establish a low body temperature risk early warning mechanism based on the temperature time-series change trend.

[0006] While some studies in recent years have attempted to introduce infrared thermal imaging technology into the field of clinical body temperature monitoring—for example, using infrared thermal imagers for intraoperative brain tumor boundary identification and surgical area temperature monitoring—these studies have largely focused on local temperature field analysis at the surgical site. Their goal is to assist the surgical procedure itself rather than monitoring the patient's overall body temperature distribution. Furthermore, these studies have not fully considered the impact of various interfering heat sources unique to the operating room environment on the accuracy of infrared thermometry, nor have they established risk assessment and early warning models based on the temporal trends of body temperature changes. From a systems engineering perspective, current technologies lack a closed-loop collaborative monitoring system that integrates infrared thermal imaging acquisition, temperature calibration, intelligent image segmentation, temperature field analysis, and risk warning, making it difficult to meet the urgent clinical needs for continuous monitoring of the entire intraoperative body temperature distribution and early warning of hypothermia.

[0007] Therefore, there is an urgent need in the existing technology for an intelligent body temperature monitoring method and system that can monitor the patient's body temperature distribution in real time, non-contact, and over the entire area during surgery, while effectively eliminating the influence of various interfering heat sources in the complex environment of the operating room, and providing low body temperature risk warnings based on temperature change trends. Summary of the Invention

[0008] To address the technical problems in existing intraoperative temperature monitoring technologies, such as single-point temperature measurement failing to reflect the overall body temperature distribution, interference from heat sources in the operating room affecting the accuracy of temperature measurement, and the lack of a low body temperature risk early warning mechanism.

[0009] This invention provides a method for monitoring intraoperative patient body temperature distribution based on infrared images, comprising the following steps:

[0010] Step S1, Infrared thermal imaging image acquisition and temperature calibration preprocessing: Using an infrared thermal imager installed on the ceiling of the operating room, continuous infrared thermal imaging is performed on the patient's body surface area on the operating table at a preset acquisition frame rate to obtain an infrared thermal imaging image sequence containing body surface temperature radiation information; temperature calibration processing and environmental radiation correction processing are performed on the acquired infrared thermal imaging image sequence to obtain a temperature distribution image.

[0011] Step S2, Intelligent segmentation and elimination of interfering heat sources of the exposed human body surface area: The temperature distribution image processed in step S1 is used to automatically identify and segment the patient's exposed body surface area using a human body region segmentation network based on deep learning. At the same time, interfering heat sources such as the radiation area of ​​the surgical shadowless lamp, the working area of ​​the electrosurgical unit, and the area covered by the irrigation fluid are detected and eliminated to obtain an effective body surface temperature area image.

[0012] Step S3, Extraction and Analysis of Temperature Field Feature Parameters in Anatomical Regions: Based on the effective body surface temperature region image obtained in Step S2, the region is divided into multiple anatomical regions according to the human anatomy zoning standard. For each anatomical region, the body surface temperature distribution feature parameters are extracted, including the regional average temperature, temperature standard deviation, temperature gradient amplitude, and core-periphery temperature difference.

[0013] Step S4, dynamic assessment and early warning of intraoperative hypothermia risk: Based on the surface temperature distribution characteristic parameters of each anatomical region extracted in step S3, the temperature drop rate and the expansion speed of the low temperature area are calculated by combining the time-series temperature sequence, and the risk level of intraoperative hypothermia is comprehensively assessed. The risk level is then fed back to step S1 to dynamically adjust the acquisition frame rate.

[0014] Step S5, Generating and Outputting a Temperature Distribution Heatmap and Monitoring Report: Based on the processing results of steps S3 and S4, a whole-body temperature distribution heatmap is generated, and an intraoperative temperature monitoring report containing temperature statistics for each anatomical region and a low body temperature warning level is output.

[0015] This invention also provides an intraoperative patient body temperature distribution monitoring system based on infrared images, including an infrared thermal imaging image acquisition and temperature calibration preprocessing module, an intelligent segmentation and interference heat source elimination module for exposed human body surface areas, an anatomical area temperature field characteristic parameter extraction and analysis module, an intraoperative hypothermia risk dynamic assessment and early warning module, and a body temperature distribution heat map generation and monitoring report output module. Each module corresponds one-to-one with steps S1 to S5 of the method described above. The modules are interconnected through cascaded data flow transmission and control signal feedback loops, forming a deeply coupled closed-loop collaborative system architecture. The risk level and parameter adjustment instructions output by the intraoperative hypothermia risk dynamic assessment and early warning module can be transmitted back to the acquisition and segmentation modules via feedback channels, enabling dynamic adaptive adjustment of monitoring parameters.

[0016] The beneficial effects of this invention are as follows: First, by using an overhead infrared thermal imager in the operating room, non-contact, comprehensive, and continuous monitoring of the patient's body surface temperature during surgery is achieved. This overcomes the limitations of traditional single-point temperature measurement methods, which cannot reflect the spatial distribution characteristics of body temperature throughout the entire body. It can simultaneously acquire temperature information from multiple anatomical regions of the patient's body, providing global information support for intraoperative temperature management. Second, through the synergistic operation of a deep learning-based human body region segmentation network and three types of interference heat source elimination mechanisms, the invention effectively solves the problem of the impact of various interference heat sources on the accuracy of body temperature measurement in the complex environment of the operating room, such as continuous radiation from the operating room lamp, intermittent high temperatures from the electrosurgical unit, and low temperatures from the irrigation fluid. This allows infrared thermal imaging technology to be reliably applied to body temperature monitoring in the special environment of the operating room. Third, through the cascade analysis of anatomical region temperature field characteristic parameters extraction and dynamic assessment of low body temperature risk, it is possible not only to calculate the instantaneous temperature statistical characteristics of each anatomical region but also to capture the temperature change trend over time and the spatial expansion dynamics of low-temperature regions, achieving a leapfrog improvement from static temperature monitoring to dynamic risk early warning. Fourth, through the feedback adjustment mechanism between the post-risk assessment module and the forward acquisition and segmentation module, a complete closed-loop collaborative control architecture is formed. The system can adaptively adjust the acquisition frame rate and segmentation sensitivity according to the patient's real-time body temperature, effectively controlling the consumption of computing resources while ensuring monitoring sensitivity, and has good engineering practicality. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the method for monitoring intraoperative patient body temperature distribution based on infrared images provided in an embodiment of the present invention.

[0018] Figure 2 This is a schematic diagram of the architecture of an intraoperative patient body temperature distribution monitoring system based on infrared images provided in an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0020] See Figure 1 The intraoperative patient body temperature distribution monitoring method based on infrared images provided in this invention comprises five core steps: infrared thermal imaging image acquisition and temperature calibration preprocessing; intelligent segmentation and interference heat source elimination of exposed human body surface areas; extraction and analysis of temperature field characteristic parameters of anatomical areas; dynamic assessment and early warning of intraoperative hypothermia risk; and generation of body temperature distribution heatmap and output of monitoring report. These steps form a data-driven cascaded processing flow, where the risk assessment results of subsequent steps can inversely adjust the acquisition and segmentation parameters of preceding steps, thus constituting a closed-loop collaborative intelligent monitoring system. The following provides a detailed description of each step.

[0021] Step S1: Infrared thermal imaging image acquisition and temperature calibration preprocessing. In one embodiment of the present invention, an uncooled long-wave infrared thermal imager is installed on the ceiling of the operating room. Its operating wavelength is 8–14 μm, which corresponds to the peak wavelength of human body surface radiation and can effectively penetrate water vapor and aerosol interference in the operating room. Preferably, the detector resolution of the infrared thermal imager is 640×512 pixels, the temperature resolution (NETD) is not greater than 0.05K, and the frame rate is configurable from 1 to 30 frames / s. The infrared thermal imager is fixed to the ceiling track near the surgical lamp arm by an adjustable bracket, with its lens facing the operating table. The installation height is 2.0–3.0m above the operating table surface to ensure that the field of view covers the entire effective area of ​​the operating table. In a preferred embodiment of the present invention, the installation height is set to 2.5m. At this time, a lens with a focal length of 19mm can be used to obtain a field of view of about 32°×25°, which corresponds to an imaging coverage area of ​​about 1.4m×1.1m on the operating table, which is sufficient to cover the entire body surface area of ​​an adult patient.

[0022] In actual operation, the infrared thermal imager continuously acquires infrared thermal images of the patient's body surface area on the operating table at a preset acquisition frame rate. Preferably, the default acquisition frame rate is set to 2 frames / s, which effectively reduces the computational load of data transmission and processing while ensuring the temporal resolution of temperature changes. It should be noted that this acquisition frame rate is not fixed; the low body temperature risk dynamic assessment module in subsequent step S4 will dynamically adjust it based on the real-time assessment results. The specific adjustment strategy will be described in detail in step S4. The raw infrared thermal image output by the infrared thermal imager is transmitted to the image processing host in the form of a 14-bit digital signal, with each frame containing the radiation intensity values ​​of 640×512 pixels.

[0023] Temperature calibration is a crucial step in converting infrared radiation intensity values ​​into absolute temperature values. In one embodiment of this invention, a surface-source blackbody radiation source is used for pre-calibration during the system deployment phase. The calibration temperature range is set to 20–42°C, with a calibration step size of 1°C, resulting in the acquisition of radiation intensity-temperature correspondence data for 23 temperature points. Based on the above calibration data, a fourth-order polynomial fitting is used to establish a mapping function from radiation intensity to temperature, with a fitting residual not exceeding ±0.1°C. The mapping function is expressed as: ,in, For pixels The calibrated temperature value is in °C. For pixels The original radiation intensity value ranges from 0 to 16383 (14-bit quantization). For the first Polynomial fitting coefficients of order 1 The value range is 0 to 4, and each coefficient is obtained by fitting the blackbody calibration data using the least squares method. In a specific embodiment of the present invention, , , , , .

[0024] The purpose of environmental radiation correction is to eliminate the systematic bias of background radiation in the operating room on body surface temperature measurement. Specifically, environmental temperature and humidity sensors are installed in the operating room to acquire real-time data on the ambient temperature. and relative humidity The formula for calculating the contribution of ambient background radiation is:

[0025] ,

[0026] in, Pixels after environmental radiation correction The temperature value, in °C; The environmental emissivity is set at 0.90–0.95 for typical operating room wall and ceiling materials. The emissivity of human skin is taken as 0.98 according to literature reports; The Stefan-Boltzmann constant has a value of W / (m 2 ·K 4 Preferably, a humidity correction factor is also introduced. Compensation for temperature deviations caused by water vapor absorption:

[0027] ,

[0028] in, This is the final corrected temperature value, in °C. This is the humidity correction factor, with a value ranging from 0.001 to 0.005℃ / %RH, representing the temperature measurement deviation caused by each percentage change in relative humidity; The relative humidity is measured in real time, and the unit is % %. The reference humidity for calibration is usually set to 50%. The temperature equivalent deviation corresponds to the transmittance attenuation coefficient of water vapor in the 8–14 μm band, with a value ranging from 0.02 to 0.08 °C. After the above temperature calibration and environmental radiation correction, an accurate temperature distribution image is obtained, with a temperature measurement accuracy within ±0.3 °C.

[0029] It should be further noted that, to ensure the long-term stability of temperature calibration, in a preferred embodiment of the present invention, the system automatically executes a rapid calibration and verification procedure before each surgery. Specifically, an infrared image of the operating table surface is acquired while the operating table is empty, and its temperature reading is compared with the measurement value of a contact platinum resistance temperature sensor installed on the operating table surface. If the deviation exceeds ±0.5℃, a recalibration process is triggered. Furthermore, the non-uniformity correction mechanism built into the infrared thermal imager is automatically executed every 30 minutes to compensate for spatial temperature non-uniformity caused by the response drift of each pixel in the detector. The above calibration and verification and automatic correction mechanisms ensure the reliability of temperature measurement under long-term continuous operation conditions.

[0030] For data transmission, the temperature distribution images output by the infrared thermal imager are transmitted to the image processing host via Gigabit Ethernet using the GigEVision standard protocol. Each frame contains approximately 640 × 512 × 2 bytes, or 655,360 bytes of data. At the highest acquisition frame rate of 10 frames / s, the required transmission bandwidth is approximately 6.25 MB / s, while the effective transmission bandwidth of Gigabit Ethernet is approximately 110 MB / s, providing ample bandwidth. After receiving the data, the image processing host stores it in a circular buffer for pipelined processing. The buffer depth is set to 100 frames to handle short-term fluctuations in computational load.

[0031] Step S2 involves intelligent segmentation and elimination of interfering heat sources in the exposed body surface area. This step receives the temperature distribution image output from step S1 as input and uses a deep learning-based human body region segmentation network to automatically identify and accurately segment the patient's exposed body surface area. In one embodiment of the invention, the human body region segmentation network employs a semantic segmentation network based on an encoder-decoder architecture, specifically an improved design based on the U-Net architecture. The encoder uses ResNet-50 as the backbone network to extract multi-scale features. ResNet-50 consists of four residual block groups, outputting feature maps at scales of 1 / 4, 1 / 8, 1 / 16, and 1 / 32 of the original resolution, with channel numbers of 256, 512, 1024, and 2048, respectively. The decoder employs a layer-by-layer bilinear upsampling and skip connection mechanism to fuse high-resolution low-level features from each level of the encoder with high-level semantic features, gradually restoring the spatial resolution to the original input size. Preferably, an attention gating mechanism is introduced after the skip connection. By calculating the attention weights of the encoder features and decoder features pixel by pixel, the feature responses of irrelevant background regions are adaptively suppressed, thereby enhancing the segmentation accuracy of the body surface region boundaries.

[0032] Regarding training data preparation, in one embodiment of this invention, 2000 intraoperative infrared thermal imaging images were collected as the training set. Each image was pixel-level annotated by experienced annotators, who added annotations to the patient's exposed body surface area. The annotation categories included four types: exposed body surface area, surgical drape area, surgical instrument area, and background area. Training employed a weighted combination of cross-entropy loss and Dice loss as the optimization objective, with a weight ratio of 0.5:0.5. The Adam optimizer was used, and the initial learning rate was set to [value missing]. The training batch size was 8, and the total number of training rounds was 100. The learning rate was reduced to 0.1 times its previous value in rounds 60 and 80, respectively. After training, the average intersection-over-union (mIoU) of the segmentation network on the test set reached 0.92, with the mIoU of the exposed body surface region reaching 0.95, which meets the accuracy requirements for real-time intraoperative monitoring.

[0033] During model training, to improve the segmentation network's generalization ability to patients in different positions, surgical types, and body types, one embodiment of this invention employs multiple data augmentation strategies. These include random horizontal flipping, random rotation (angle range -15° to +15°), random scaling (scale range 0.8 to 1.2), random brightness adjustment (simulating infrared image temperature shift), and random cropping and padding. Furthermore, considering the significant variations in the exposed area and position of the patient's body surface in operating room scenarios, a Mosaic stitching enhancement technique is introduced. This stitches four infrared images from different surgical scenarios into a single training sample, forcing the network to learn a more robust representation of body surface region features.

[0034] Regarding model inference optimization, in a preferred embodiment of this invention, the trained segmentation network is quantized and optimized using the TensorRT inference engine, employing FP16 half-precision floating-point inference mode. This reduces the single-frame inference time from the original 120ms to 45ms while maintaining a segmentation accuracy loss of no more than 0.5%. To further improve processing efficiency, a keyframe-interpolation frame alternating processing strategy is adopted for consecutive frames: a complete segmentation network inference is performed every 5 frames as the keyframe result, and intermediate frames undergo optical flow-guided motion compensation interpolation based on the segmentation masks of the preceding and following keyframes to obtain approximate segmentation results. This strategy further reduces the average single-frame processing time to approximately 15ms.

[0035] It is particularly important to note that various interfering heat sources exist in the operating room environment. Their temperature characteristics in infrared thermal imaging images differ significantly from the human body surface temperature, but failure to identify and eliminate them will severely affect the accuracy of subsequent temperature field analysis. In one embodiment of the present invention, the elimination of interfering heat sources specifically includes the following three types of processing.

[0036] The first category excludes areas radiated by the surgical shadowless lamp. When in operation, the surgical shadowless lamp continuously emits infrared radiation towards the operating table. The irradiated area appears as a bright spot in the infrared image with a temperature significantly higher than the normal human body surface temperature. This invention performs connected component analysis on the temperature distribution image, first extracting areas with temperatures consistently above 45°C. Then, it performs a morphological dilation operation on these areas to cover the radiation gradient transition regions. The size of the dilated structuring element is preferably 5×5 pixels. Finally, the dilated areas are marked as surgical shadowless lamp radiation areas and excluded.

[0037] The second category involves excluding the electrosurgical operating area. During surgery, the electrosurgical unit generates a momentary high-temperature region near the surgical incision, with temperatures reaching several hundred degrees Celsius. This region is relatively small in size and exhibits intermittent bursts over time. This invention calculates the difference image of the temperature distribution in adjacent frames to detect areas where the temperature change exceeds 20°C and the connected region area is less than 100 pixels. These areas are marked as the electrosurgical operating area. Preferably, the electrosurgical operating area is extended with three frames before and after in the time dimension to eliminate the residual heat effect after the electrosurgical unit stops operating.

[0038] The third category excludes areas covered by irrigation fluid. The saline irrigation fluid used during surgery is typically colder than the body surface temperature. In infrared images, areas covered by irrigation fluid appear as low-temperature patches where the temperature suddenly drops to near room temperature. This invention identifies connected regions in the temperature distribution image with a temperature below 28°C and a temperature decrease exceeding 5°C compared to the previous frame as areas covered by irrigation fluid. Considering the flow and diffusion characteristics of the irrigation fluid on the body surface, a morphological dilation operation is performed on the marked regions, with the dilated structural element having a size of 7×7 pixels.

[0039] After human body region segmentation and elimination of three types of interfering heat sources, an image of the effective body surface temperature region after interference removal was obtained. This image retains only the true temperature information of the patient's exposed body surface area; all pixels affected by interfering heat sources are marked as invalid regions and are not included in subsequent temperature field analysis calculations.

[0040] Step S3: Extraction and analysis of temperature field feature parameters of anatomical regions. This step receives the effective body surface temperature region image output from step S2 as input, performs anatomical partitioning, and extracts the temperature field feature parameters of each region. In one embodiment of the present invention, according to the anatomical partitioning standard for clinical body temperature monitoring, the effective body surface temperature region is divided into six anatomical regions: head and neck region, chest and abdomen region, left upper limb region, right upper limb region, left lower limb region, and right lower limb region. The partitioning is implemented by using the segmentation mask output by the human region segmentation network in step S2, combined with the detection results of human skeletal key points, to determine the boundaries of each anatomical region. Preferably, the skeletal key point detection is implemented using a heatmap regression-based method, and the detected key points include 15 key points: top of head, neck, both shoulders, both elbows, both wrists, center of trunk, both hips, both knees, and both ankles. Based on the coordinate positions of the above key points, the pixel range of each anatomical region in the image is determined by predefined partitioning rules.

[0041] Specifically, the zoning rules are defined as follows: the neck key points serve as the horizontal boundary between the head and neck region and the trunk region; the line connecting the two shoulder key points extending downwards to the line connecting the two hip key points forms the left and right boundaries of the chest and abdomen region; the area extending outwards from the two shoulder key points to the two wrist key points defines the upper limb region; and the area extending downwards from the two hip key points to the two ankle key points defines the lower limb region. Preferably, a transition buffer zone with a width of 10 pixels is set at the zoning boundaries. Pixels within the transition buffer zone do not participate in the feature calculation of any anatomical region to avoid the interference of temperature gradient effects at the zoning boundaries on the regional statistical results. When some key points are detected as being unable to be accurately located due to occlusion by the surgical covering, the system automatically uses an estimation method based on a priori human proportion model to estimate the position of the occluded key points, ensuring the robustness of the anatomical zoning.

[0042] For each anatomical region, the present invention extracts the following four temperature field characteristic parameters.

[0043] The first item is the regional average temperature. The calculation formula is as follows:

[0044] ,

[0045] in, For the first The average temperature of each anatomical region, in °C. The value range is 1 to 6, corresponding to the above 6 anatomical regions; For the first The set of effective pixels within a anatomical region; For set The total number of valid pixels; The corrected temperature value is the output of step S1.

[0046] The second item is the temperature standard deviation. The calculation formula is as follows:

[0047] ,

[0048] in, For the first The temperature standard deviation of an anatomical region, expressed in °C, reflects the uniformity of temperature distribution within that region. A larger temperature standard deviation indicates a more uneven temperature distribution, potentially suggesting the presence of localized low-temperature or hotspot areas. Preferably, when... The local abnormal temperature detection mechanism is triggered when the temperature exceeds 1.5℃.

[0049] The third item is the temperature gradient magnitude. The Sobel gradient operator was applied to the temperature distribution image of each anatomical region to calculate the horizontal gradient component. and vertical gradient components The formula for calculating the magnitude of the temperature gradient is:

[0050] ,

[0051] in, For the first The average temperature gradient amplitude of each anatomical region, in °C / pixel; and Each pixel The horizontal and vertical temperature gradient components at a given location are calculated using a 3×3 Sobel operator convolution. The magnitude of the temperature gradient reflects the degree of spatial variation in body surface temperature; areas with larger gradient magnitudes typically correspond to the boundaries of thermal insulation coverage or the junction between the surgically exposed area and the covered area.

[0052] Preferably, in one embodiment of the present invention, in addition to extracting the above four macroscopic statistical feature parameters, local abnormal temperature point detection is also performed on each anatomical region. Specifically, within each anatomical region, a temperature block matrix is ​​constructed using 8×8 pixels as units. The average temperature of each block is calculated, and then the median absolute deviation method is used to detect blocks with abnormal temperatures. When the average temperature of a block deviates from the median temperature of the anatomical region by more than 2.5 times the median absolute deviation, it is marked as a local abnormal temperature point. The spatial location information of the abnormal temperature points is recorded and transmitted to the report generation module in step S5 for visualization on the body temperature distribution heatmap using special markers. This local abnormal temperature point detection mechanism can discover local low-temperature regions masked by macroscopic statistical parameters, improving the system's sensitivity to early local hypothermia detection.

[0053] Furthermore, to improve the stability of temperature field characteristic parameters over time and reduce the impact of acquisition noise, one embodiment of the present invention performs sliding window time smoothing on each characteristic parameter. Specifically, a moving average filter with a window width of 3 frames is used for the regional average temperature and temperature standard deviation, and a moving average filter with a window width of 5 frames is used for the temperature gradient magnitude and core-periphery temperature difference. The characteristic parameters after time smoothing are more stable, effectively reducing the jitter of characteristic parameters caused by infrared thermal imager detector noise or instantaneous disturbances in ambient airflow, providing more reliable input data for subsequent hypothermia risk assessment.

[0054] The fourth item is the core-periphery temperature difference. The calculation formula is as follows:

[0055] ,

[0056] in, This represents the temperature difference between the core area and the outer perimeter area, expressed in °C. The average temperature of the core region (chest and abdomen region); It is a collection of peripheral regions, including four regions: the left upper limb, the right upper limb, the left lower limb, and the right lower limb. This represents the number of peripheral regions, with a value of 4. The core-periphery temperature difference is a key indicator for assessing intraoperative body temperature distribution; normally, this difference is maintained within the range of 2–4°C. Temperatures exceeding 4°C indicate insufficient peripheral perfusion or significant hypothermia in the peripheral region, requiring clinical attention.

[0057] Step S4: Dynamic assessment and early warning of intraoperative hypothermia risk. This step receives the temperature field characteristic parameters of each anatomical region output from step S3 as input, and combines them with time-series temperature data to perform dynamic assessment and graded early warning of hypothermia risk. In one embodiment of the present invention, the temperature decrease rate of each anatomical region is first calculated. Specifically, for the most recent... Average temperature sequence of the same anatomical region in frame temperature distribution images Perform linear regression fitting to obtain the slope of the linear change in temperature over time as the rate of temperature decrease. The calculation formula is as follows:

[0058] ,in, For the first The rate of temperature decrease for each anatomical region, in °C / min; This represents the number of frames within the sliding time window. A preferred value is the current acquisition frame rate multiplied by the number of frames corresponding to 5 minutes. For example, at the default acquisition frame rate of 2 frames / second... frame; For the first The first frame The average temperature of each anatomical region. When A negative value indicates that the temperature is decreasing.

[0059] It is particularly noteworthy that, to reduce the calculation error of the temperature drop rate caused by acquisition noise and transient temperature disturbances, in a preferred embodiment of the present invention, before performing linear regression fitting, median filtering smoothing is first applied to the average temperature time series data of each anatomical region, with a filtering window width of 5 frames. The temperature series after median filtering eliminates pulsed temperature fluctuations caused by the use of electrosurgical units or brief coverage by irrigation fluid, making the linear regression fitting results more stable and reliable. Furthermore, when the goodness of fit of the linear regression... When the value is less than 0.6, it indicates that the linear assumption of the temperature change trend does not hold. At this time, the system switches to the exponential moving average method to estimate the temperature change trend in order to better adapt to the nonlinear change pattern of temperature that first drops rapidly and then tends to level off.

[0060] Low temperature region expansion rate The calculation method is as follows: First, the temperature distribution image of each frame is binarized, and the temperature is below a preset low temperature threshold. The pixels are marked as low-temperature pixels, and the preset low-temperature threshold is set to 34°C. Then, the first pixel is counted. Total area of ​​low-temperature pixels in the frame (In pixels), the formula for calculating the expansion rate of the low-temperature region is:

[0061] ,in, The rate of expansion in the low-temperature region is expressed in pixels per minute. This represents the total area of ​​low-temperature pixels in the current frame. for The total area of ​​low-temperature pixels before the time interval; For calculation time intervals, a setting of 2 minutes is preferred. A positive value indicates that the low-temperature region is expanding.

[0062] Based on the rate of temperature decrease and the speed of expansion of the low-temperature region, this invention establishes a three-level hypothermia risk assessment model. A first rate threshold is defined. ℃ / min, second rate threshold ℃ / min, first extended threshold pixels / min, second extended threshold Pixels / min. The risk level determination rule is as follows: when the temperature decrease rate of all anatomical areas... All greater than (i.e., the absolute value of the descent rate is less than 0.2℃ / min) and When the risk level is determined to be low, the system maintains the normal monitoring mode; when any anatomical area is classified as low-risk... and ,or When the risk level is determined to be medium, the system issues a yellow alert and automatically increases the frame rate from the current value to 5 frames / s to encrypt the temperature sampling frequency; when any anatomical area is identified... or When the risk level is determined to be high, the system issues a red alert, increases the frame rate to 10 frames / s, and pushes an emergency low body temperature warning to the operating room display terminal.

[0063] It is worth noting that the aforementioned risk level feedback mechanism constitutes the closed-loop collaborative core of the method of this invention. When the risk level increases, the acquisition frame rate in step S1 increases accordingly, thereby providing temperature data with higher temporal resolution for steps S3 and S4, making the detection of temperature decline trends more sensitive. When the risk level returns to low risk, the acquisition frame rate returns to the default value to save computational resources. In addition, in medium-risk and high-risk states, the segmentation threshold of the human body region segmentation network in step S2 is reduced by 0.05 accordingly to ensure that the boundaries of exposed body surface areas can still be accurately identified even when the overall body surface temperature decreases. This feedback adjustment mechanism enables the entire monitoring system to adaptively adjust its operating parameters according to the patient's real-time body temperature, achieving deep coupling and closed-loop collaboration between the pre-acquisition step and the post-evaluation step.

[0064] Step S5: Generating a body temperature distribution heatmap and outputting a monitoring report. This step receives the temperature field characteristic parameters output in step S3 and the low body temperature risk level assessment results output in step S4 as inputs to generate an intuitive body temperature distribution heatmap and a structured monitoring report. In one embodiment of the present invention, the generation process of the whole-body temperature distribution heatmap is as follows: First, the temperature value of each effective pixel in the effective body surface temperature area image obtained in step S2 is normalized to the range of 0 to 1. The normalized temperature mapping range is set to 30 to 38°C, with temperature values ​​below 30°C mapped to 0 and temperature values ​​above 38°C mapped to 1. Then, the normalized temperature values ​​are mapped to RGB color values ​​using the JET color mapping table, where the temperature is mapped from low to high as dark blue, light blue, green, yellow, orange, and red. Preferably, the boundary contour lines of each anatomical region are superimposed on the heatmap, and the average temperature value of each anatomical region is marked at the geometric center position of each region.

[0065] The intraoperative temperature monitoring report includes the following elements: monitoring timestamp, current frame number, regional average temperature and standard deviation of each anatomical region, core-peripheral temperature difference, rate of temperature decrease, percentage of hypothermic area, current hypothermia risk level, and historical records of risk level changes. Preferably, the report also includes a trend curve of temperature changes over time for each anatomical region, displaying dynamic temperature changes in 15-minute time windows. In high-risk situations, the report additionally outputs spatial location information of hypothermic areas, indicating which anatomical regions are mainly distributed in the hypothermic areas and the specific coordinates of the most severe hypothermia, providing precise spatial guidance for clinical medical staff to implement targeted warming interventions.

[0066] Furthermore, in one embodiment of the present invention, the body temperature monitoring report is automatically refreshed and pushed to the monitoring display terminal in the operating room and the anesthesiologist's workstation every 30 seconds. In medium- and high-risk conditions, the report refresh frequency is increased to every 10 seconds to ensure that clinical staff can obtain the latest body temperature distribution information in a timely manner. All monitoring data is stored and archived in the form of a time-series database. The body temperature monitoring records for each surgery can be retrospectively analyzed postoperatively, supporting the evaluation of postoperative body temperature management quality and the source analysis of hypothermia events.

[0067] See Figure 2 The present invention also provides an intraoperative patient body temperature distribution monitoring system based on infrared images. This system corresponds one-to-one with the steps described in the above method embodiments, including an infrared thermal imaging image acquisition and temperature calibration preprocessing module 1, an intelligent segmentation and interference heat source elimination module for exposed human body surface areas 2, an extraction and analysis module for temperature field characteristic parameters of anatomical areas 3, an intraoperative low body temperature risk dynamic assessment and early warning module 4, and a body temperature distribution heat map generation and monitoring report output module 5.

[0068] The infrared thermal imaging image acquisition and temperature calibration preprocessing module 1 includes a hardware acquisition unit and a preprocessing calculation unit. The core component of the hardware acquisition unit is an uncooled long-wave infrared thermal imager mounted on the ceiling track of the operating room. Its operating wavelength is 8–14 μm, the detector resolution is 640 × 512 pixels, and the temperature resolution is no greater than 0.05 K. In one embodiment of the invention, the infrared thermal imager transmits the acquired raw infrared image data to the preprocessing calculation unit via a gigabit Ethernet interface. The preprocessing calculation unit is deployed in the equipment room outside the operating room and is implemented using an industrial-grade embedded computing platform. Preferably, this computing platform is equipped with a multi-core processor and at least 8 GB of memory, running a real-time operating system to ensure the deterministic timing of image processing. The preprocessing calculation unit stores radiation intensity-temperature mapping function parameters obtained in advance through blackbody calibration and acquires operating room environmental parameters in real time through an environmental temperature and humidity sensor interface. The specific processing flow of this module is consistent with the temperature calibration processing and environmental radiation correction processing described in step S1 of the method embodiment, and will not be repeated here.

[0069] The intelligent segmentation and interference heat source elimination module 2 for exposed human body surface areas is deployed on an image processing server equipped with a GPU accelerator card. In one embodiment of the present invention, the GPU accelerator card has a video memory capacity of not less than 4GB and a computing power of not less than 6 TFLOPS to meet the computing requirements of real-time inference of the deep learning-based human body region segmentation network. Preferably, the TensorRT inference engine is used to optimize the deployment of the trained segmentation network model, compressing the model inference time to no more than 50ms per frame, ensuring that real-time processing requirements can still be met at the highest acquisition frame rate of 10 frames / s. This module also integrates an interference heat source detection and elimination submodule to realize the automatic identification and elimination of the radiation area of ​​the shadowless lamp, the working area of ​​the electrosurgical unit, and the area covered by the rinsing fluid. The specific detection and judgment logic is consistent with step S2 in the method embodiment.

[0070] The anatomical region temperature field feature parameter extraction and analysis module 3 receives the segmented effective body surface temperature region image data and performs human anatomical partitioning and feature parameter extraction. This module includes a human skeleton key point detection submodule and a temperature field feature calculation submodule. The skeleton key point detection submodule uses a lightweight pose estimation network to achieve real-time localization of 15 human key points, with a single-frame inference time of no more than 20ms. The temperature field feature calculation submodule calculates four feature parameters for each of the six anatomical regions: regional average temperature, temperature standard deviation, temperature gradient amplitude, and core-periphery temperature difference. The calculation formulas for all feature parameters are completely consistent with those described in step S3 of the method embodiment. Preferably, this module also maintains a 600-frame temporal feature parameter buffer to store historical temporal data of feature parameters for each anatomical region, providing time-series input for subsequent dynamic assessment of hypothermia risk.

[0071] The intraoperative hypothermia risk dynamic assessment and early warning module 4 is the core functional module of this system for achieving closed-loop feedback control. This module receives the current frame feature parameters and historical time-series data output by the temperature field feature parameter extraction and analysis module, and performs real-time risk assessment according to the temperature drop rate calculation method and low-temperature region expansion speed calculation method described in step S4 of the method embodiment. The output of this module includes the current hypothermia risk level and the corresponding acquisition frame rate adjustment command. The acquisition frame rate adjustment command is transmitted back to the infrared thermal imaging image acquisition and temperature calibration preprocessing module through the system's internal communication bus to achieve closed-loop dynamic adjustment of the acquisition frame rate. In addition, in medium-risk and high-risk states, this module also sends a segmentation threshold adjustment command to the intelligent segmentation and interference heat source elimination module for exposed human body surface areas, reducing the judgment threshold of the segmentation network by 0.05 to adapt to the overall temperature drop scenario. Through the above dual feedback channels, the five modules of this system form a complete closed-loop collaborative architecture from data acquisition to risk assessment to parameter adjustment, enabling the system to adaptively switch working modes according to the patient's real-time body temperature status.

[0072] The body temperature distribution heatmap generation and monitoring report output module 5 is responsible for transforming the processing results into intuitive visual output. This module includes two sub-modules: a heatmap rendering engine and a report generation engine. The heatmap rendering engine uses a GPU-accelerated color mapping algorithm to complete the real-time rendering of the whole-body temperature distribution heatmap at a speed of no more than 10ms per frame. The report generation engine automatically integrates the feature parameters, risk levels, and time-series trend data output by each module according to a predefined report template format to generate a structured intraoperative body temperature monitoring report. Preferably, the generated heatmap and monitoring report are pushed to the wall-mounted monitoring monitor in the operating room via an HDMI video output interface, and simultaneously pushed to the software client of the anesthesiologist's workstation via a local area network, and stored in the time-series database of the hospital information system in a standardized data format for postoperative review and analysis.

[0073] In a preferred embodiment of the present invention, the five modules interact with each other through a message middleware based on a publish-subscribe model. Each module runs as an independent process, and efficient data transfer between modules is achieved through shared memory and message queues. This architecture design enables each module to process data from different frames in parallel, forming a pipelined parallel processing mode, which improves the overall system throughput by approximately three times compared to the serial processing mode. Furthermore, the system provides a standardized configuration management interface, allowing clinical engineers to adjust various working parameters, including the acquisition frame rate range, low temperature threshold, risk level determination threshold, report refresh frequency, and display color mapping scheme, to adapt to the personalized needs of different surgical types and clinical departments. The wall-mounted monitoring monitor in the operating room uses a 27-inch 4K resolution LCD screen, with the display screen divided into four areas: the upper left displays a real-time body temperature distribution heatmap, the upper right displays temperature value dashboards for each anatomical region, the lower left displays a temperature time-series trend curve, and the lower right displays a low body temperature risk level indicator and warning information. Under normal monitoring conditions, the display interface is predominantly green; under medium-risk conditions, it switches to predominantly yellow with intermittent flashing indicators; under high-risk conditions, it switches to predominantly red and simultaneously triggers an audible and visual alarm to ensure that the surgical team can detect low body temperature warning signals immediately.

[0074] To verify the technical effectiveness of the method and system described in this invention, a clinical trial was conducted in the operating room environment of a tertiary hospital. The test environment was a standard laminar flow operating room, with the ambient temperature maintained at 22–24°C and the relative humidity maintained at 40%–60%. An infrared thermal imager with a resolution of 640×512 and uncooled long-wave infrared thermal imager was installed 2.5m above the operating table. A total of 50 patients undergoing abdominal surgery under general anesthesia were included in the test, with operation times ranging from 90 to 240 minutes.

[0075] Regarding the accuracy of temperature measurement, using the core body temperature measured by the esophageal temperature probe as a reference, the average deviation between the average temperature of the chest and abdominal region (core region) output by the system of this invention and the esophageal temperature is 0.35℃, with a standard deviation of 0.28℃ and a Pearson correlation coefficient of 0.94. Compared with traditional infrared thermometers, due to the introduction of environmental radiation correction and interference heat source elimination mechanisms in this invention, the temperature measurement error under continuous irradiation by the operating room shadowless lamp is reduced by approximately 62%, from ±1.2℃ in the traditional method to ±0.45℃ in the method of this invention.

[0076] Regarding the accuracy of human body region segmentation, the human body region segmentation network trained on 2000 training data sets achieved an average intersection-over-union (IoU) ratio of 0.92 on 200 independent test datasets, with an IoU ratio of 0.95 for exposed body surface regions and 0.89 for surgical covering regions. In terms of interference heat source elimination, the detection rate for areas radiated by operating lights reached 98.5%, the detection rate for electrosurgical working areas reached 95.2%, and the detection rate for areas covered by irrigation fluid reached 93.8%, with a false exclusion rate of less than 2.1%.

[0077] Regarding hypothermia early warning performance, using retrospectively confirmed postoperative hypothermia events as the gold standard, the system of this invention achieved a hypothermia early warning sensitivity of 94.7%, a specificity of 89.3%, a positive predictive value of 85.6%, and a negative predictive value of 96.2%. Compared with traditional single-point esophageal temperature probes, the system of this invention can detect the hypothermia trend in the peripheral region an average of 12 minutes in advance, providing valuable early warning time for clinical medical staff to implement warming interventions. Furthermore, because this invention provides spatial information on the whole-body temperature distribution, clinicians can accurately locate the site of hypothermia, significantly improving the targeting and effectiveness of warming interventions.

[0078] Regarding system real-time performance, the end-to-end processing latency of this invention does not exceed 200ms at the default acquisition frame rate of 2 frames / s, and does not exceed 120ms at the highest frame rate of 10 frames / s, meeting the clinical needs of real-time intraoperative monitoring. The entire monitoring process requires no physical contact with the patient, does not interfere with the surgical procedure, does not increase the risk of cross-infection, and has good clinical applicability and safety.

[0079] Regarding the advance warning of hypothermia, a comparison of the system of this invention with a traditional single-point esophageal temperature probe in 50 test cases showed that, for the 18 patients ultimately confirmed to have hypothermia, the traditional esophageal temperature probe only issued a hypothermia alarm when the core body temperature dropped below 36°C. However, the system of this invention, by monitoring the temperature decline trend in peripheral areas (mainly the lower limbs and upper limb extremities), could detect the abnormal expansion of the peripheral low-temperature area while the core body temperature was still within the normal range, issuing a medium-risk warning on average 12 minutes earlier. In 8 of these patients, because clinical staff implemented timely warming interventions such as covering with warming blankets and warming intravenous fluids based on the early warning provided by the system, the patients' core body temperature ultimately did not drop below 36°C, successfully preventing hypothermia events. These results indicate that the system of this invention has significant clinical value in the early warning and preventive intervention of intraoperative hypothermia.

[0080] Regarding core-peripheral temperature difference monitoring, data analysis of 50 test cases showed that within the first 30 minutes after surgery, the core-peripheral temperature difference rapidly increased, averaging from 2.1℃ preoperatively to 3.8℃, with the most significant temperature drop observed in the lower extremities. This invention's system can capture this dynamic temperature difference change in real time and automatically trigger a medium-risk warning when the core-peripheral temperature difference exceeds 4℃, providing a quantitative reference for clinical assessment of peripheral perfusion status and optimization of insulation strategies.

[0081] The test results above show that the intraoperative patient body temperature distribution monitoring method and system based on infrared images described in this invention exhibits significantly better technical performance than traditional single-point temperature measurement methods in several key dimensions, including temperature measurement accuracy, human body region segmentation precision, ability to eliminate interfering heat sources, low body temperature early warning sensitivity, and real-time system processing performance. This provides strong technical support for the transformation of intraoperative patient body temperature management from an experience-driven mode to a data-driven and image-guided mode.

[0082] The embodiments of the present invention are not limited to the specific embodiments described above. Those skilled in the art can make various equivalent changes or substitutions based on the technical solutions of the present invention, and all such changes or substitutions should be included within the protection scope of the present invention.

Claims

1. A method for monitoring intraoperative patient body temperature distribution based on infrared images, characterized in that, Includes the following steps: Step S1, Infrared thermal imaging image acquisition and temperature calibration preprocessing: Using an infrared thermal imager installed on the ceiling of the operating room, continuous infrared thermal imaging is performed on the patient's body surface area on the operating table at a preset acquisition frame rate to obtain an infrared thermal imaging image sequence containing body surface temperature radiation information; temperature calibration processing is performed on the acquired infrared thermal imaging image sequence, converting the infrared radiation intensity value into an absolute temperature value according to the preset blackbody calibration parameters, and performing environmental radiation correction processing based on the operating room ambient temperature and humidity parameters to eliminate the interference of operating room ambient background radiation on body surface temperature measurement, and obtaining a temperature distribution image after temperature calibration and environmental radiation correction; Step S2, Intelligent segmentation and elimination of interfering heat sources in the exposed body surface area: The temperature distribution image processed in step S1 is processed by a human body region segmentation network based on deep learning to automatically identify and segment the patient's exposed body surface area; at the same time, the interfering heat source areas in the temperature distribution image are detected and eliminated. The interfering heat sources include the radiation area of ​​the surgical shadowless lamp, the working area of ​​the electrosurgical unit, and the area covered by the irrigation fluid, to obtain an effective body surface temperature area image after eliminating interfering heat sources; Step S3, Extraction and analysis of temperature field feature parameters of anatomical region: Based on the effective body surface temperature region image obtained in step S2, the effective body surface temperature region is divided into multiple anatomical regions according to the human anatomy zoning standard, and the body surface temperature distribution feature parameters are extracted for each anatomical region. Step S4, dynamic assessment and early warning of intraoperative hypothermia risk: Based on the surface temperature distribution characteristic parameters of each anatomical region extracted in step S3, the temperature drop rate and the expansion speed of the low temperature region in each anatomical region are calculated in combination with the time-series temperature sequence. The intraoperative hypothermia risk level is comprehensively assessed based on the temperature drop rate and the expansion speed of the low temperature region, and the assessed hypothermia risk level is fed back to step S1 to dynamically adjust the acquisition frame rate of infrared thermal imaging images.

2. The method for monitoring intraoperative patient body temperature distribution based on infrared images according to claim 1, characterized in that, In step S1, the infrared thermal imager operates in the band of 8–14 μm, with a temperature resolution of no more than 0.05 K, and the preset acquisition frame rate is 1–10 frames / s; the blackbody calibration parameters include the radiation intensity-temperature mapping curve obtained under a calibration temperature range of 20–42 °C.

3. The method for monitoring intraoperative patient body temperature distribution based on infrared images according to claim 1, characterized in that, In step S1, the environmental radiation correction process includes: acquiring the operating room ambient temperature and relative humidity, calculating the contribution of ambient background radiation, subtracting the contribution of ambient background radiation from the radiation intensity of the infrared thermal imaging image, and remapping it to a corrected temperature value.

4. The method for monitoring intraoperative patient body temperature distribution based on infrared images according to claim 1, characterized in that, In step S2, the human body region segmentation network is a semantic segmentation network based on an encoder-decoder architecture. The encoder uses a residual network structure to extract multi-scale features, and the decoder restores spatial resolution through layer-by-layer upsampling and skip connections, outputting a pixel-level human body surface region mask.

5. The method for monitoring intraoperative patient body temperature distribution based on infrared images according to claim 1, characterized in that, In step S2, the elimination of interfering heat sources includes: performing connected component analysis on the temperature distribution image, identifying connected regions with temperatures higher than the upper limit of the normal human body surface temperature range, marking connected regions with temperatures consistently higher than 45°C as shadowless lamp radiation areas, marking regions where the temperature change amplitude between adjacent frames exceeds a preset temperature change threshold and the spatial area is smaller than a preset threshold as electrosurgical working areas, and marking connected regions with temperatures lower than the ambient temperature as rinsing fluid covered areas.

6. The method for monitoring intraoperative patient body temperature distribution based on infrared images according to claim 5, characterized in that, In step S3, the surface temperature distribution characteristic parameters include the regional average temperature, temperature standard deviation, temperature gradient amplitude, and temperature difference between the core region and the peripheral region. The anatomical regions include the head and neck region, the chest and abdomen region, the upper limb region, and the lower limb region. The temperature gradient amplitude is obtained by performing the Sobel gradient operator on the temperature distribution image of each anatomical region. The core region is the chest and abdomen region, and the peripheral regions are the upper limb region and the lower limb region.

7. The method for monitoring intraoperative patient body temperature distribution based on infrared images according to claim 6, characterized in that, In step S4, the temperature drop rate is obtained by linear regression fitting of the average temperature of the same anatomical region in multiple consecutive temperature distribution images; the low temperature region expansion rate is obtained by calculating the rate of change of the area of ​​pixel regions below a preset low temperature threshold at adjacent time points, and the preset low temperature threshold is set to 34℃.

8. The method for monitoring intraoperative patient body temperature distribution based on infrared images according to claim 7, characterized in that, In step S4, the low body temperature risk level is divided into three levels: low risk, medium risk, and high risk. Specifically: when the temperature drop rate of all anatomical regions is less than the first rate threshold and the expansion rate of the low-temperature region is less than the first expansion threshold, it is determined to be low risk; when the temperature drop rate of any anatomical region is greater than or equal to the first rate threshold and less than the second rate threshold, it is determined to be medium risk; when the temperature drop rate of any anatomical region is greater than or equal to the second rate threshold or the expansion rate of the low-temperature region is greater than or equal to the second expansion threshold, it is determined to be high risk.

9. The method for monitoring intraoperative patient body temperature distribution based on infrared images according to claim 8, characterized in that, The procedure also includes step S5, generating a body temperature distribution heatmap and outputting a monitoring report: Based on the temperature field characteristic parameters of step S3 and the low body temperature risk level assessment results of step S4, a whole-body body temperature distribution heatmap is generated, and an intraoperative body temperature monitoring report containing temperature statistics for each anatomical region and a low body temperature warning level is output. The whole-body body temperature distribution heatmap uses a color mapping method to map temperature values ​​to visual colors, where temperatures from low to high are mapped to blue, green, yellow, and red respectively. The intraoperative body temperature monitoring report also includes time-series temperature change curves for each anatomical region and a trend graph of low body temperature risk level changes over time.

10. An intraoperative patient body temperature distribution monitoring system based on infrared images, used to implement the intraoperative patient body temperature distribution monitoring method based on infrared images as described in claim 9, characterized in that, include: The infrared thermal imaging image acquisition and temperature calibration preprocessing module is used to continuously acquire infrared thermal images of the patient's body surface area on the operating table using an infrared thermal imager installed on the top of the operating room at a preset acquisition frame rate. The module performs temperature calibration processing and environmental radiation correction processing on the acquired infrared thermal imaging image sequence to obtain a temperature distribution image. The intelligent segmentation and interference heat source elimination module for exposed human body surface area is used to automatically identify and segment the patient's exposed body surface area using a human body region segmentation network based on deep learning. At the same time, it detects and eliminates interference heat sources such as the radiation area of ​​the surgical shadowless lamp, the working area of ​​the electrosurgical unit, and the area covered by the irrigation fluid, so as to obtain an effective body surface temperature area image. The module for extracting and analyzing the characteristic parameters of the temperature field in the anatomical region is used to divide the effective body surface temperature area into multiple anatomical regions according to the human anatomical zoning standard, and extract the characteristic parameters of body surface temperature distribution for each anatomical region, including the regional average temperature, temperature standard deviation, temperature gradient amplitude, and core-periphery temperature difference. The intraoperative hypothermia risk dynamic assessment and early warning module is used to calculate the temperature drop rate and the expansion speed of the low temperature area based on the surface temperature distribution characteristic parameters of each anatomical region, comprehensively assess the intraoperative hypothermia risk level, and feed the risk level back to the infrared thermal imaging image acquisition and temperature calibration preprocessing module to dynamically adjust the acquisition frame rate. The body temperature distribution heatmap generation and monitoring report output module is used to generate a whole-body body temperature distribution heatmap based on temperature field characteristic parameters and low body temperature risk level assessment results, and output an intraoperative body temperature monitoring report.

Images