Design method and system of heating wire layout of neonatal warming blanket and corresponding product

Abstract

The application relates to a design method and system of a heating wire layout of a neonatal warming blanket and corresponding products, the method comprising: obtaining a body surface contact footprint image obtained by simulating wrapping of a neonatal physical model; performing grid quantization processing on the body surface contact footprint image to obtain a plurality of grid cells and generate a corresponding density distribution feature matrix, wherein each element in the density distribution feature matrix represents the proportion of the effective contact area of the corresponding grid cell; based on the density distribution feature matrix, a multi-dimensional feature vector is constructed for each grid cell, and a pre-trained machine learning model is used to process the multi-dimensional feature vector to determine the predicted wiring density of each grid cell; and according to the predicted wiring density, vector layout information for controlling the differential wiring operation of an automatic wiring device on the heating layer of the warming blanket is generated.

Description

Technical Field

[0001] This application belongs to the field of heating equipment and its design technology, and in particular relates to a design method, system and corresponding product for the layout of heating wires in a newborn heating blanket. Background Technology

[0002] Maintaining stable body temperature is crucial for ensuring stable vital signs in newborns, especially premature infants. Because the thermoregulatory center in newborns is not fully developed, their skin has a thin stratum corneum, and their body surface area is relatively large, they are highly susceptible to heat loss, leading to hypothermia and related complications. Therefore, external heat sources such as warming blankets are commonly used in clinical practice to provide warmth and support to newborns.

[0003] Currently, existing heated blanket technologies mainly focus on using sensors to acquire the user's body position and contact status, and then using a control system to adjust the power and control the heating zones in a preset area. Typical solutions include the following: The first type is a zone-switching control scheme based on a pressure sensor matrix. This scheme divides the heating blanket into multiple regular rectangular heating zones. By detecting the number of pressure sensors within each rectangular heating zone, it determines the user's body position and controls the heating wires in the corresponding area to turn on or off. However, this scheme only performs simple regular geometric partitioning on a macroscopic level; the physical arrangement of the heating wires within the heating zones remains uniform. Newborns have significant differences in blood vessel density and heat absorption and dissipation capabilities in different body parts. For example, the core trunk requires more heat while the limbs dissipate heat more easily. Traditional uniformly arranged heating wires cannot adapt to these physiological differences, easily leading to uneven heat distribution and localized heat supply-demand mismatches.

[0004] The second category is a multi-mode dynamic power control scheme based on non-contact duration. This type of scheme combines pressure sensors and infrared sensors to obtain the non-contact time between the heating element and the human body, dynamically dividing the heating blanket into continuous contact areas, intermittent contact areas, and non-contact areas, and applying different power heating modes to different areas to achieve gradient temperature control. The disadvantage of this type of scheme is that it is essentially still a sensor-dependent, hysteresis-based adjustment mechanism. Newborns are extremely light and short, and traditional sensors are not sensitive enough to their minute changes in body position, which can easily lead to temperature adjustment lag or failure. At the same time, the underlying physical heating wire network structure remains unchanged, and simply changing the power supply cannot fundamentally solve the spatial matching problem between heat dissipation and the absorption characteristics of the newborn's skin microcirculation vascular network.

[0005] The third type is a solution based on user weight data and pressure distribution maps for posture recognition and target temperature control. This solution periodically acquires total pressure values ​​and combines them with a set weight data threshold to create a pressure distribution map that identifies the user's head and body areas, thereby locking in the main lying posture range and controlling heat only in the target area. The drawback of this solution is that its logic is primarily designed for adults with a certain weight who can turn over independently. Since newborns lack the ability to adjust their sleeping position to avoid overheated areas, this solution, which relies on external weight data for later calculations and area control, poses a safety hazard. Sensor recognition errors or system response delays can easily lead to excessive heat accumulation in localized areas or insufficient heating in peripheral areas.

[0006] In summary, the common flaws of existing technical solutions are that they all rely excessively on external sensors and algorithms for post-processing temperature and power adjustment, without innovating on the physical structure of the heating blanket itself, i.e., the layout of the heating wires. Due to the lack of pre-processed biomimetic structural design based on the characteristics of newborn skin contact, existing products suffer from low heat transfer efficiency, poor physiological adaptability, and pose safety hazards such as localized overheating burns or cold stress. Summary of the Invention

[0007] In view of the shortcomings of the prior art, the purpose of the invention is to provide a design method, system and corresponding product for the heating wire layout of a newborn heating blanket. By using contact imprint images based on the physical model of the newborn, and by using gridded quantization processing and machine learning models, the invention achieves pre-data-driven bionic layout design of the differential density of the heating wires from the physical level, so as to solve the problems of relying on post-sensor adjustment, low heat transfer efficiency and poor physiological adaptability in the prior art.

[0008] The first aspect of this application proposes a design method for the layout of heating wires in a neonatal warming blanket, comprising: acquiring a body surface contact imprint image obtained by simulating wrapping a neonatal physical model; performing gridding and quantization processing on the body surface contact imprint image to obtain multiple grid cells and generating a corresponding density distribution feature matrix, wherein each element in the density distribution feature matrix represents the effective contact area ratio of the corresponding grid cell; constructing a multi-dimensional feature vector for each grid cell based on the density distribution feature matrix, and processing the multi-dimensional feature vector using a pre-trained machine learning model to determine the predicted wiring density of each grid cell; and generating vector layout information for controlling an automatic wiring device to perform differentiated wiring operations on the heating layer of the warming blanket according to the predicted wiring density.

[0009] According to a second aspect of the present disclosure, a newborn bionic heating blanket is provided, comprising an outer covering layer, a flexible heating layer, a heat preservation layer, and a temperature control module; the outer covering layer, the flexible heating layer, and the heat preservation layer are stacked; heating wires are arranged on the flexible heating layer; the temperature control module is electrically connected to the heating wires and is used to control the heating power of the heating wires; the flexible heating layer includes multiple regions with different heating wire arrangement densities, and the heating wire arrangement density is determined by the method described in the first aspect of the present disclosure.

[0010] According to a third aspect of the present disclosure, a storage medium is provided, the storage medium including a stored program, wherein, when the program is executed, a processor performs the method described in any of the above embodiments.

[0011] According to a fourth aspect of the present disclosure, a design system for the heating wire layout of a neonatal heating blanket is provided, comprising: an image acquisition module for acquiring a body surface contact imprint image obtained by simulating wrapping a neonatal physical model; a density distribution feature matrix generation module for performing gridded quantization processing on the body surface contact imprint image to obtain multiple grid cells and generating a corresponding density distribution feature matrix, wherein each element in the density distribution feature matrix represents the effective contact area ratio of the corresponding grid cell; a wiring density prediction module for constructing a multi-dimensional feature vector for each grid cell based on the density distribution feature matrix, and processing the multi-dimensional feature vector using a pre-trained machine learning model to determine the predicted wiring density of each grid cell; and a vector layout information generation module for generating vector layout information for controlling an automatic wiring device to perform differentiated wiring operations on the heating layer of the heating blanket according to the predicted wiring density.

[0012] According to a fifth aspect of the present disclosure, a design system for the heating wire layout of a neonatal warming blanket is provided, comprising: a processor; and a memory connected to the processor, configured to provide the processor with instructions for processing the following steps: acquiring a body surface contact imprint image obtained by simulating wrapping a neonatal physical model; performing gridding and quantization processing on the body surface contact imprint image to obtain multiple grid cells and generating a corresponding density distribution feature matrix, wherein each element in the density distribution feature matrix represents the effective contact area ratio of the corresponding grid cell; constructing a multidimensional feature vector for each grid cell based on the density distribution feature matrix, and processing the multidimensional feature vector using a pre-trained machine learning model to determine the predicted wiring density of each grid cell; and generating vector layout information for controlling an automatic wiring device to perform differentiated wiring operations on the heating layer of the warming blanket based on the predicted wiring density.

[0013] The beneficial effects of this application are as follows: (1) This application obtains images of contact imprints on the body surface of a newborn's body using a simulated physical model, and generates a density distribution feature matrix based on gridded quantization processing. This standardizes the contact distribution information into quantifiable feature data, and then uses a pre-trained machine learning model to predict the wiring density of each grid cell. Finally, it generates vector layout information to drive an automatic wiring device, thus establishing a complete pre-data-driven biomimetic design method. This method completely changes the traditional design logic of heating blankets, which relies on adjusting the power supply for post-processing sensing. It innovates from the source of the underlying physical heating wire network structure, precisely matching the differentiated heat needs of different body parts of the newborn, achieving physiological-level precise heat transfer, and significantly improving heat transfer efficiency and physiological adaptability.

[0014] (2) This application adopts a multi-task learning network to simultaneously perform classification sub-tasks and regression sub-tasks, outputting coverage partition level labels and predicted wiring density respectively. The two sub-tasks work together through weighted joint optimization of the global loss function, realizing integrated processing from macro partition determination to micro density prediction. Furthermore, the prediction results can be directly converted into vector layout information, opening up a complete closed loop from data-driven design to industrial automated manufacturing.

[0015] (3) The newborn bionic heating blanket provided in this application has heating wire density in different areas of its flexible heating layer determined by the above method. High-density wiring is used in areas with high contact demand, while low-density wiring or blanking is used in areas with low contact demand. This achieves precise spatial matching between heat dissipation and the contact characteristics of the newborn's skin from the physical underlying structure. Compared with the existing wiring scheme that relies on sensors for post-adjustment, this design solves the problem of local heat accumulation or insufficient heating caused by physiological supply and demand mismatch from the source. While improving heat transfer efficiency, it significantly reduces the safety risks of local overheating burns and edge cold stress. Attached Figure Description

[0016] The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Throughout the drawings, the same reference numerals denote the same components. Obviously, the drawings described below are merely some embodiments described in this application, and those skilled in the art can obtain other drawings based on these drawings.

[0017] Figure 1 This is a hardware structure block diagram of a computing device for implementing the method described in Embodiment 1 of this disclosure; Figure 2 This is a flowchart of the design method for the heating wire layout of the newborn warming blanket according to Embodiment 1 of this application; Figure 3This is a schematic diagram of a body surface contact imprint image according to Embodiment 1 of this application; Figure 4 This is a schematic diagram of the density distribution feature matrix according to Embodiment 1 of this application; Figure 5 This is a schematic diagram of the design system for the heating wire layout of the newborn warming blanket according to Embodiment 2 of this application; Figure 6 This is a schematic diagram of the design system for the heating wire layout of the newborn warming blanket according to Embodiment 3 of this application. Detailed Implementation

[0018] To enable those skilled in the art to better understand the technical solutions of this disclosure, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this disclosure.

[0019] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0020] Example 1

[0021] According to this embodiment, a method embodiment of a design method for the heating wire layout of a newborn warming blanket is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0022] The method embodiments provided in this example can be executed on a server or similar computing device. Figure 1 A hardware block diagram of a computing device is shown, illustrating a design method for implementing the heating wire layout of a neonatal warming blanket. Figure 1As shown, a computing device may include one or more processors (processors may include, but are not limited to, microprocessors such as MCUs or programmable logic devices such as FPGAs), a memory for storing data, a transmission device for communication functions, and an input / output interface. The memory, transmission device, and input / output interface are connected to the processor via a bus. In addition, it may also include a display, keyboard, and cursor control device connected to the input / output interface. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, a computing device may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0023] It should be noted that the aforementioned one or more processors and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element in a computing device. As involved in the embodiments of this disclosure, the data processing circuits serve as processor control (e.g., selection of a variable resistor termination path connected to an interface).

[0024] The memory can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the design method of the heating wire layout of the newborn heating blanket in this embodiment of the present disclosure. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, to implement the design method of the heating wire layout of the newborn heating blanket described above. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to the computing device via a network. Examples of the above-mentioned networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0025] The transmission device is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the computing device's communication provider. In one example, the transmission device includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device may be a Radio Frequency (RF) module used for wireless communication with the Internet.

[0026] The display can be, for example, a touchscreen liquid crystal display (LCD), which allows users to interact with the user interface of the computing device.

[0027] It should be noted here that, in some optional embodiments, the above... Figure 1 The computing device shown may include hardware elements (including circuitry), software elements (including computer code stored on a computer-readable medium), or a combination of both hardware and software elements. It should be noted that... Figure 1 This is only one instance of a specific particular instance, and is intended to illustrate the types of components that may exist in the aforementioned computing devices.

[0028] Under the above operating environment, according to the first aspect of this embodiment, a design method for the heating wire layout of a newborn warming blanket is provided. Figure 2 A flowchart illustrating the method is shown below. (Refer to...) Figure 2 As shown, the method includes: Step 1: Obtain images of body surface contact imprints obtained by simulating wrapping a physical model of a newborn.

[0029] In this embodiment of the invention, the method is applied to the design of the heating wire layout for a warming blanket used in neonatal clinical care. For clinical and operational convenience, instead of directly using a real newborn for contact imprint collection, a neonatal physical model is used as a substitute. A neonatal physical model refers to a physical dummy model made based on standard neonatal vital signs (such as weight, length, and body circumference), whose external dimensions and surface contours are consistent with those of a real newborn. When designing the heating wire layout for the warming blanket, imprint paper or a pressure-sensitive medium is placed on the actual contact interface between the neonatal physical model's surface (such as the back or torso) and the outer wrapping blanket. The neonatal physical model is then wrapped in a simulated manner according to standard clinical wrapping methods (such as supine wrapping), and the preset posture is maintained for a certain duration (such as 30 to 60 seconds), allowing the neonatal physical model to form a realistic contact imprint on the imprint paper or pressure-sensitive medium under natural pressure and wrapping tension. Then, remove the rubbing paper or pressure-sensitive medium and photograph it using an imaging device (such as a digital camera or scanner) to obtain a digital image of the area retaining the effective imprint, i.e., a body surface contact imprint image, such as... Figure 3 As shown. Figure 3 This is a schematic diagram of the body surface contact imprint image according to Embodiment 1 of this application. The black imprint area in the figure is the actual contact area between the body surface of the newborn physical model and the swaddling blanket, while the white area is the non-contact area.

[0030] Therefore, by using the above method, the surface contact imprint images obtained from a neonatal physical model in a simulated clinical swaddling scenario realistically reflect the contact distribution between the neonatal's body surface and the warming blanket while swaddled. This method avoids the risks associated with directly manipulating real neonates and abandons the erroneous assumption of "assuming a flat and uniform surface contact" in traditional approaches, providing a data foundation that conforms to the actual physical contact state for subsequent gridded quantization processing and biomimetic layout design.

[0031] Step 2: Perform gridding and quantization processing on the contact imprint image on the body surface to obtain multiple grid cells and generate a corresponding density distribution feature matrix. Each element in the density distribution feature matrix represents the effective contact area ratio of the corresponding grid cell.

[0032] In this embodiment of the invention, in order to transform the unstructured contact distribution information contained in the body surface contact imprint image obtained in step 1 into standardized quantitative data that can be read and processed by machine learning models, it is necessary to perform gridded quantization processing on the body surface contact imprint image. The gridded quantization processing refers to the process of transforming the contact distribution information reflected in the body surface contact imprint image from an unstructured image form into a standardized data structure composed of multiple grid units that can be quantified and analyzed. The core of this processing is: discretizing the image space into multiple grid units, each grid unit corresponding to a physical location unit on the heating layer of the heating blanket; simultaneously, for each grid unit, extracting its effective contact area percentage from the image as the contact density characterization value of that grid unit. Organizing the effective contact area percentages of all grid units according to the spatial arrangement order of the grid units generates the density distribution feature matrix. The density distribution feature matrix is ​​a two-dimensional numerical matrix, whose number of rows and columns corresponds to the number of rows and columns of the grid division, respectively. The value of each element in the matrix is ​​the effective contact area percentage of the grid unit at the corresponding spatial location; a higher value indicates a closer contact between the newborn's body surface and the heating blanket at that location, and a greater heat demand.

[0033] like Figure 4 As shown, Figure 4 This is a schematic diagram of the density distribution feature matrix according to Embodiment 1 of this application. Figure 4 The rows and columns represent the grid cell numbers (e.g., for a 10×10 grid, the number range is 1 to 10); the color of each cell in the diagram represents the coverage of that grid cell, with a value ranging from 0 to 1, indicated by the color bar on the right side of the diagram. Specifically, the yellow area (value close to 1.0) indicates that the grid cell is completely covered, meaning it has the closest contact with the newborn's skin; the green area (value approximately 0.5 to 0.8) indicates that the grid cell is covered on average by more than half; the blue area (value approximately 0.2 to 0.4) indicates that the grid cell has a low coverage; and the purple area (value close to 0) indicates that the grid cell is almost completely uncovered. This diagram visually identifies the distribution of the degree of contact between the warming blanket and the newborn's skin at different spatial locations, providing a visual data basis for subsequently determining the predicted wiring density of each grid cell.

[0034] Thus, through the aforementioned gridding and quantization processing, the blurred and spatially heterogeneous contact information in the original body surface contact imprint image was successfully transformed into a density distribution feature matrix with fixed dimensions and standardized values. Each element in this matrix corresponds one-to-one with a grid cell, representing the degree of contact tightness and relative heat requirement at different locations on the newborn's body surface in precise numerical form. This provides a standardized data input format for subsequent construction of multidimensional feature vectors and prediction of wiring density using machine learning models.

[0035] Step 3: Based on the density distribution feature matrix, construct a multi-dimensional feature vector for each grid cell, and use a pre-trained machine learning model to process the multi-dimensional feature vector to determine the predicted wiring density of each grid cell.

[0036] In this embodiment of the invention, in order to transform the contact distribution information and spatial location information contained in the density distribution feature matrix generated in step 2 into an accurate prediction of the wiring density of each grid cell, it is necessary to construct a multi-dimensional feature vector for each grid cell based on the density distribution feature matrix and process it using a pre-trained machine learning model.

[0037] Specifically, a multidimensional feature vector is constructed for each grid cell based on the density distribution feature matrix. This multidimensional feature vector is a numerical combination extracted from the density distribution feature matrix to describe the contact and spatial characteristics of the corresponding grid cell. Its function is to fuse and express the information of the grid cell in both the contact distribution and spatial location dimensions, enabling the machine learning model to gain a comprehensive understanding of that grid cell. The multidimensional feature vector integrates at least the contact density information of the corresponding grid cell and the spatial location information of that grid cell on the heating blanket surface.

[0038] The pre-trained machine learning model is trained using a large amount of historical contact imprint sample data. During the training phase, the model learns and establishes a mapping relationship between the contact distribution characteristics and spatial location characteristics of grid cells and the optimal wiring density. In the application phase, the constructed multidimensional feature vector is input into the pre-trained machine learning model. After internal inference operations, the model outputs the predicted wiring density for each grid cell. The predicted wiring density is a numerical value representing the relative density of heating wires that should be placed at the location of each grid cell. A higher value means that the location is in closer contact with the newborn's skin and the heat demand is greater, thus requiring a denser arrangement of heating wires.

[0039] To facilitate understanding of the physical meaning of the predicted wiring density, an example is provided below. Assume the grid resolution in step 2 is 0.5cm × 0.5cm, meaning each grid cell corresponds to a 0.5cm × 0.5cm square area on the heating layer of the heating blanket. If the machine learning model outputs a predicted wiring density of 0.85 for a certain grid cell, it means that the heating wire coverage area within that 0.5cm × 0.5cm area should be approximately 85%, indicating that a relatively dense arrangement of heating wires is needed to meet the high heat demand at that location. If the predicted wiring density for another grid cell is 0.15, it means that the heating wire coverage area within that area only needs to be approximately 15%, indicating that sparse wiring is required, resulting in a correspondingly lower heat output. Therefore, the predicted wiring density directly corresponds to the density of the heating wire arrangement in each tiny area of ​​the heating blanket's heating layer, and the value has a linear relationship with the physical coverage.

[0040] Thus, through the above processing, a pre-trained machine learning model is used to intelligently analyze and reason about multi-dimensional feature vectors, automatically completing the accurate mapping from contact imprint distribution to heating wire layout density. This process automates and automates design decisions, eliminating the need for human experience-based judgment, and the output predicted wiring density highly matches the contact characteristics of a newborn's skin, directly guiding subsequent differentiated wiring manufacturing.

[0041] Step 4: Based on the predicted wiring density, generate vector layout information for controlling the automatic wiring equipment to perform differentiated wiring operations on the heating layer of the heating blanket.

[0042] In this embodiment of the invention, in order to transform the predicted wiring density of each grid cell determined in step 3 from the digital prediction result into a production instruction that can directly drive industrial manufacturing equipment, it is necessary to generate vector layout information based on the predicted wiring density to control the automatic wiring equipment to perform differentiated wiring operations on the heating layer of the heating blanket.

[0043] Specifically, the predicted wiring density of each grid cell output in step 3 is converted into vector layout information that can be recognized and executed by the automatic wiring equipment, based on the spatial correspondence of the grid cells on the heating layer of the heating blanket. This vector layout information is a digital manufacturing instruction file containing the wiring path, wiring density, and correspondence between the heating wires at each physical location on the heating layer of the heating blanket and each grid cell area. Specifically, for heating layer areas corresponding to grid cells with higher predicted wiring density, the wiring instructions generated in the vector layout information are denser to form a high-density arrangement of heating wires in that area; for heating layer areas corresponding to grid cells with lower predicted wiring density, the wiring instructions generated in the vector layout information are sparser or simply blank instructions are generated to form a low-density or no-heating-wire arrangement in that area. This vector layout information can be directly input into the control system of the automatic wiring equipment, which then performs the physical laying operation of the heating wires on the heating layer of the heating blanket according to this vector layout information. The automatic wiring equipment can be, for example, a CNC automatic wiring machine or a flexible circuit (FPC) hot-pressing composite equipment.

[0044] Thus, through the above steps, a seamless connection from data-driven design to automated industrial manufacturing is achieved. The digital wiring density scheme predicted by the machine learning model is directly converted into vector layout information that drives automated wiring equipment for physical manufacturing, completely opening up the entire closed loop of "design-manufacturing". The heating blanket manufactured based on this vector layout information has a heating wire arrangement density that precisely matches the contact characteristics of a newborn's skin, fundamentally realizing a biomimetic layout design for on-demand heating.

[0045] As described in the background section, existing heating blanket technologies rely excessively on external sensors and algorithms for post-processing temperature and power adjustment, without innovating the physical layout of the heating wires themselves. This results in low heat transfer efficiency, poor physiological adaptability, and potential safety hazards such as localized overheating burns or cold stress.

[0046] In view of this, this application obtains surface contact imprint images of a newborn physical model and performs gridded quantization processing to generate a density distribution feature matrix. A pre-trained machine learning model is then used to determine the predicted wiring density of each grid cell, and based on this, vector layout information is generated to drive an automated wiring device to perform differentiated wiring operations. This establishes a complete pre-data-driven biomimetic design methodology. This method innovates from the source of the underlying physical heating wire network structure, precisely matching the differentiated heat needs of different body parts of the newborn, achieving physiological-level precise heat transfer, and effectively solving the problems of existing methods such as reliance on sensor lag adjustment, low heat transfer efficiency, and poor physiological adaptability.

[0047] Optionally, the operation of performing gridded quantization processing on the body surface contact imprint image to obtain multiple grid cells includes: extracting the boundary of the target contact area in the body surface contact imprint image and constructing a corresponding binary mask; performing grayscale conversion and threshold segmentation on the effective contact area determined by the binary mask to extract the contact imprint area; constructing a local parameterized coordinate system based on the vertices of the contact imprint area, and mapping the contact imprint area to a standard two-dimensional plane of a preset size through affine transformation; and discretizing the standard two-dimensional plane according to a preset resolution to obtain the multiple grid cells.

[0048] In this embodiment of the invention, in order to transform the body surface contact imprint image obtained in step 1 from an unstructured original image form that may have tilt and distortion into a standardized spatial structure composed of multiple precisely positioned and equally sized grid units, a specific grid quantization processing implementation method is provided.

[0049] First, the boundaries of the target contact area in the body surface contact imprint image are extracted, and a corresponding binary mask is constructed. Specifically, in a workstation environment (such as Python, using libraries such as OpenCV and NumPy), the body surface contact imprint image is processed using annotation software (such as LabelMe) to annotate the boundaries of the target contact area and extract the boundaries of the target quadrilateral or polygonal region. This target contact area is typically the working area of ​​the wrapping blanket. Based on the annotated boundaries, a corresponding binary mask is constructed, preserving the target region (ROI) and whitening the background pixels outside the polygonal region to eliminate background interference. This binary mask is a binary image of the same size as the original image, where pixels located inside the target contact area are assigned foreground values, and background pixels outside the target contact area are eliminated, thereby separating the target contact area from the complex background of the original image.

[0050] Then, grayscale conversion and threshold segmentation are performed on the effective contact area determined by the binary mask to extract the contact imprint region. Since the binary mask only defines the range of the target contact area, which contains both effective contact imprints (black imprints) and invalid blank areas, it is necessary to further refine the morphology of the effective contact imprints. Specifically, grayscale conversion is performed on the ROI image within the effective contact area determined by the binary mask, converting the color image to a grayscale image. Then, fixed threshold segmentation (such as using OpenCV's findContours method) is performed on the grayscale image to accurately extract the black imprint region, i.e., the contact imprint region.

[0051] Secondly, based on the vertices of the contact imprint region, a local parameterized coordinate system is constructed, and the contact imprint region is mapped to a standard two-dimensional plane of a preset size through affine transformation. Since the contact imprint region in the original body surface contact imprint image may have non-standardized factors such as tilting and perspective distortion, it cannot be directly divided into regular meshes and needs to be mapped to a standardized plane. Traditional image processing algorithms struggle to handle such irregularly shaped polygons with distortions. This step standardizes and reduces the dimensionality of the irregular physical contact boundary, providing high-precision mathematical labels for machine learning. Specifically, based on the quadrilateral vertices of the extracted contact imprint region, a local parameterized coordinate system is constructed. Under this local parameterized coordinate system, the tilted and distorted target pressure area is uniformly mapped and unified to a standard two-dimensional plane of a preset size through affine transformation, for example, mapped to a standard physical plane of 50cm × 50cm.

[0052] Next, the standard two-dimensional plane is discretized according to a preset resolution to obtain the multiple grid cells. Specifically, the standard two-dimensional plane after affine transformation is finely discretized according to a preset ultra-high resolution. For example, a 50cm×50cm standard plane is divided into 0.5cm×0.5cm units, forming 100×100 units, totaling 10,000 independent grid cells. Each grid cell has a unique row and column index on the standard plane, corresponding to a specific physical location unit on the heating layer of the heating blanket. This step ensures that every contact pixel in the real physical space is strictly and uniquely assigned to a specific local grid cell, eliminating data distortion caused by the distortion of the original image.

[0053] Thus, through the above steps, a complete method is provided to transform the original body surface contact imprint image into multiple standardized grid cells. This method achieves background separation through binary masks, accurate extraction of effective imprints through grayscale conversion and thresholding, distortion correction and standardized mapping through local parameterized coordinate systems and affine transformations, and generation of regular grid cells through discretization. This implementation ensures that each contact pixel in the real physical space is strictly and uniquely assigned to a specific local grid cell, eliminating data distortion caused by the distortion of the original image and providing an accurate spatial basis for the subsequent accurate calculation of the density distribution feature matrix.

[0054] Optionally, the operation of generating the corresponding density distribution feature matrix includes: traversing each of the grid cells and calculating the total pixel area of ​​each grid cell. and the pixel area of ​​the effective contact imprint within the grid cell. Where i and j represent the row and column indices of the grid cell in the spatial arrangement, respectively; the coverage of each grid cell is calculated according to the following formula. As the percentage of the effective contact area: ; Coverage of all grid cells The density distribution feature matrix is ​​formed by summarizing the row and column indices in the spatial arrangement.

[0055] In this embodiment of the invention, in order to further quantify the contact distribution information within each grid cell into a specific coverage value based on obtaining multiple standardized grid cells, and organize them according to their spatial arrangement to generate a standardized density distribution feature matrix that can be used as input to a machine learning model, a specific matrix generation implementation method is provided.

[0056] Specifically, all obtained grid cells are traversed, and two basic area data are calculated for each grid cell: the total pixel area of ​​the grid cell. and the pixel area of ​​the effective contact imprint within the grid cell. Where i represents the row index of the grid cell in the spatial arrangement, and j represents the column index of the grid cell in the spatial arrangement. For example, on a standard 50cm×50cm plane, 100×100 grid cells are formed with a resolution of 0.5cm×0.5cm, and each grid cell from i=1 to 100 and j=1 to 100 is counted sequentially.

[0057] Subsequently, the coverage of each grid cell is calculated according to the following formula. As the percentage of the effective contact area: ; in, Let i be the coverage of the grid cell in the i-th row and j-th column. The effective contact imprint pixel area within this grid cell. This represents the total pixel area of ​​the grid cell. Coverage. The value ranges from 0 to 1, and this value directly reflects the degree of contact between the newborn's skin and the warming blanket at the location of the grid cell.

[0058] After that, the coverage of all grid cells will be increased. The density distribution feature matrix is ​​formed by summing the data according to the row index i and column index j in the spatial arrangement. This density distribution feature matrix is ​​a two-dimensional numerical matrix, with the number of rows and columns consistent with the number of rows and columns in the grid division. The value of the element in the i-th row and j-th column of the matrix is ​​the coverage of the corresponding grid cell. .

[0059] In practical implementation, the calculation results can be output in the form of image visualization and CSV tables. Image visualization output is as follows: Figure 4 As shown, Figure 4 This is a schematic diagram of the density distribution feature matrix according to Embodiment 1 of this application. Figure 4 The rows (horizontal axis) and columns (vertical axis) represent the grid cell numbers (e.g., for a 10×10 grid, the numbering ranges from 1 to 10); the color of each cell in the graph represents the coverage of that grid cell. The values ​​range from 0 to 1, indicated by the color bars on the right side of the graph. The yellow area (values ​​close to 1.0) indicates that the grid cell is almost always completely covered, meaning it has the closest contact with the newborn's skin; the green area (values ​​approximately 0.5 to 0.8) indicates that the grid cell is covered on average by more than half; the blue area (values ​​approximately 0.2 to 0.4) indicates that the grid cell has low coverage; and the purple area (values ​​close to 0) indicates that the grid cell is almost completely uncovered. The CSV table output stores the coverage values ​​of each grid cell in comma-separated text format, facilitating subsequent data processing and machine learning model retrieval.

[0060] Thus, through the above steps, a complete implementation method is provided to quantify the contact distribution information of each grid cell into a standard density distribution feature matrix. This method, through four steps—traversing grid cells, calculating contact area, calculating coverage, and summarizing by spatial location—successfully transforms the clinically ambiguous and unstructured surface contact distribution information into a precisely quantifiable coverage matrix feature. This matrix perfectly quantifies the contact extrema and heat flow demand distribution of the newborn's body surface, providing extremely accurate mathematical labels for subsequent machine learning and physical wiring, and solving the technical challenge of standardizing and reducing the dimensionality of irregular boundaries.

[0061] Optionally, the operation of constructing a multi-dimensional feature vector for each grid cell based on the density distribution feature matrix includes: obtaining the coverage rate of the current grid cell itself from the density distribution feature matrix as the first dimension feature of the current grid cell; calculating the statistical feature value of the coverage rate of all grid cells within a preset neighborhood window centered on the current grid cell as the second dimension feature of the current grid cell, wherein the statistical feature value includes the mean and / or standard deviation; normalizing the row index and column index of the current grid cell in the spatial arrangement to obtain the normalized coordinates of the spatial position as the third dimension feature of the current grid cell; and combining the first dimension feature, the second dimension feature, and the third dimension feature to obtain the multi-dimensional feature vector.

[0062] In this embodiment of the invention, in order for the machine learning model to perceive the spatial location of a specific grid cell and its neighborhood relationship with surrounding grid cells, rather than simply treating the grid cell as an isolated data point, the pure density numerical matrix in the generated density distribution feature matrix needs to be transformed into a multi-dimensional feature vector for each decision unit (i.e., each grid cell) as the standard input of the machine learning model.

[0063] Specifically, for each grid cell, a feature vector containing at least the following three dimensions is constructed.

[0064] First dimension: Obtain the coverage of the current grid cell itself from the density distribution feature matrix. This serves as the first dimension feature of the current grid cell. This feature directly reflects the degree of contact between the current grid cell and is the most fundamental feature component in the multidimensional feature vector.

[0065] The second dimension: Within a preset neighborhood window centered on the current grid cell, calculate the statistical characteristic value of the coverage of all grid cells within the preset neighborhood window, which serves as the second dimension feature of the current grid cell. The statistical characteristic value includes the mean and / or standard deviation. For example, taking the current grid cell (i,j) as the center, take a 3×3 neighborhood window, and calculate the mean and standard deviation of the coverage of the 9 grid cells within this window. The mean feature reflects the degree of thermal contact aggregation in the area surrounding the grid cell, helping the machine learning model distinguish between "contact core areas" and "noise isolated points"; the standard deviation feature reflects the uniformity of contact distribution in the area surrounding the grid cell.

[0066] The third dimension: Normalize the row index i and column index j of the current grid cell in the spatial arrangement to obtain the normalized coordinates of the spatial position. , This serves as the third dimension feature of the current mesh cell. For example, for a 100×100 mesh, = =100, and the normalized coordinates of the spatial location of grid cell (30,50) are (0.3,0.5). This feature gives the machine learning model the ability to perceive the actual spatial location of the grid cell on the physical surface of the heating blanket, enabling the model to learn the heat demand patterns of different spatial locations.

[0067] Optionally, the distance from the center of the current grid cell to key anatomical landmarks (such as the estimated center point of the child's trunk) can be used as a fourth-dimensional feature or supplementary feature to further enhance the model's ability to perceive spatial structure.

[0068] Subsequently, the first-dimensional feature, the second-dimensional feature, and the third-dimensional feature are combined to obtain the multi-dimensional feature vector. This combination refers to concatenating the various dimensional features into a unified vector data structure according to a preset order. For example, for a grid cell (i,j), its multi-dimensional feature vector can be represented as [ , , , , ],in and These are the average and standard deviation of the coverage within the neighborhood window of the grid cell, respectively.

[0069] Thus, through the above steps, the generated density distribution feature matrix can be successfully transformed into a multidimensional feature vector corresponding to each grid cell. This multidimensional feature vector not only contains the contact density information of the grid cell itself, but also integrates the contact distribution statistics and spatial location information of its surrounding neighborhood. This enables the machine learning model to make comprehensive judgments based on the spatial environment, thereby possessing the ability to automatically and accurately partition various new input boundaries, significantly improving the model's generalization performance and prediction accuracy.

[0070] Optionally, the machine learning model is a multi-task learning network, and the operation of processing the multi-dimensional feature vector using a pre-trained machine learning model to determine the predicted wiring density of each grid cell includes: inputting the multi-dimensional feature vector into the multi-task learning network; generating a coverage partition level label for each grid cell from the classification sub-task in the multi-task learning network, wherein the coverage partition level label is used to determine the density interval to which the wiring density of the grid cell belongs; and generating the predicted wiring density corresponding to each grid cell from the regression sub-task in the multi-task learning network within the density interval determined by the coverage partition level label; wherein the global loss function of the multi-task learning network during training is composed of a weighted joint optimization of the classification loss of the classification sub-task and the regression loss of the regression sub-task.

[0071] In this embodiment of the invention, in order to transform the constructed multidimensional feature vector into the predicted wiring density corresponding to each grid cell, and to solve the problem that traditional single prediction models cannot simultaneously handle macroscopic partitioning and accurate microscopic density prediction, this embodiment uses a multi-task learning network as a machine learning model. The single prediction task is decomposed into two clearly defined cascaded sub-tasks, or they are merged into a multi-task learning network for joint processing. This provides a specific implementation method for machine learning model processing.

[0072] Specifically, the constructed multidimensional feature vector is input into the multi-task learning network. Structurally, this multi-task learning network includes a shared feature extraction layer and two parallel task branches, corresponding to the classification sub-task and the regression sub-task, respectively.

[0073] The classification subtask in the multi-task learning network generates a coverage partition level label for each grid cell. This classification subtask addresses the macroscopic partitioning problem; its input is the multidimensional feature vector, and its output is the coverage partition level label for that grid cell. The coverage partition level label determines the density range to which the wiring density of the grid cell belongs, i.e., macroscopically determining whether the grid cell belongs to a high-coverage-demand area or a low-coverage-demand area, thereby guiding the presence and approximate spacing range of heating wires during physical wiring. This classification subtask can be implemented using tree models such as XGBoost or LightGBM, or fully connected neural networks, with multi-class cross-entropy as the loss function.

[0074] The regression subtask in the multi-task learning network generates the predicted wiring density for each grid cell within the density interval determined by the coverage partition level label. This regression subtask addresses the micro-layout problem; under the density interval constraint defined by the classification subtask, it further outputs a continuous numerical value for the specific wiring density of the grid cell, such as the percentage of pixels covered by the effective heating wire per unit area or the physical length. This regression subtask can be implemented using random forest regression or neural network regression. Through a cascaded collaborative approach where the classification subtask first defines the interval and the regression subtask then defines the specific value, a complete parameter mapping from macro-partitioning to micro-density is achieved.

[0075] During the model training phase, the following training design was adopted to ensure that the multi-task learning network can adapt to new input boundaries and maintain high robustness in changing clinical scenarios.

[0076] In terms of data diversity, contact imprint images of newborn physical models generated under different weights, lengths, and clinical wrapping positions (such as supine and lateral) were continuously collected, and coverage feature matrices were generated according to the methods described in steps 1 to 2, which greatly enriched the feature diversity of the training set samples.

[0077] In terms of scale normalization, for different individual newborn data, their working regions are uniformly affine mapped to a standardized size plane to ensure that the feature matrix dimensions of all input models are strictly consistent.

[0078] Regarding the cross-validation mechanism, K-fold cross-validation is strictly employed during training to evaluate model stability. For example, 5-fold or 10-fold cross-validation is used, dividing the training data into K mutually exclusive subsets. Each time, K-1 subsets are used as the training set, and the remaining subset is used as the validation set, alternating between training and validation K times. This cross-validation is specifically used to verify the model's partition accuracy and macro-average F1 score on unseen newborn individual data (i.e., out-of-distribution data), preventing overfitting and ensuring that the model maintains stable predictive performance when facing newborns with different physical characteristics in real-world applications.

[0079] The global loss function of the multi-task learning network is formed by weighted joint optimization of the classification loss of the classification subtask and the regression loss of the regression subtask. Through this joint optimization mechanism, the two subtasks promote each other and converge collaboratively during training. The classification subtask provides a reasonable density interval constraint for the regression subtask, and the accurate prediction results of the regression subtask within the corresponding interval verify and correct the rationality of the classification boundary, thereby significantly improving the overall prediction accuracy and generalization ability of the model.

[0080] Thus, through the above steps, a multi-task learning network organically integrates macro-level partitioning and micro-level density prediction into an end-to-end processing framework. This framework outputs partition-level labels through a classification subtask to determine density intervals, and outputs specific predicted wiring density within the classification interval through a regression subtask. Co-training of the two subtasks is achieved through weighted joint optimization of the global loss function. Combined with the construction of multi-source heterogeneous datasets, scale normalization, and K-fold cross-validation, the model's generalization ability and prediction stability are effectively guaranteed in diverse clinical scenarios. This approach effectively solves the problems of simple classification models failing to provide continuous density values ​​and simple regression models lacking partition constraints and easily producing unreasonable predictions. It achieves a complete mapping from multi-dimensional feature vectors to accurate predicted wiring density, providing a precise data foundation for the subsequent generation of vector layout information.

[0081] Optionally, the coverage zoning level labels include the following four levels: Level IV coverage area, where the coverage rate CR of the grid cells in the Level IV coverage area is ≥75%; Level III coverage area, where the coverage rate CR of the grid cells in the Level III coverage area is ∈ [50%, 74%]; Level II coverage area, where the coverage rate CR of the grid cells in the Level II coverage area is ∈ [25%, 49%]; and Level I coverage area, where the coverage rate CR of the grid cells in the Level I coverage area is <25%.

[0082] In this embodiment of the invention, in order to establish a clear correspondence between the coverage partition level labels output by the classification subtasks and the specific coverage numerical ranges, thereby providing a clear decision-making basis for the presence and approximate spacing range of heating wires during subsequent physical wiring, a contact coverage classification model needs to be established. A specific four-level partitioning implementation method is provided.

[0083] Specifically, based on the calculated coverage of each grid cell The surface of the heating blanket is divided into the following four levels of labels.

[0084] Level IV High Coverage Zone: The coverage rate (CR) of the grid cells in the Level IV coverage zone is ≥75%. This area corresponds to the part of the newborn's body surface that is in the closest contact with the heating blanket, and has the greatest heat demand. In subsequent physical wiring, this area needs to be equipped with a high-density, continuous layout of heating wires to meet the extremely high heat transfer efficiency requirements.

[0085] Level III Higher Coverage Area: The coverage rate CR of the grid cells in the Level III coverage area is [50%, 74%]. This area corresponds to parts with relatively close contact but slightly lower than Level IV, with moderate to high heat demand. A medium-density layout is adopted in this area during subsequent physical cabling.

[0086] Level II Low Coverage Area: The coverage rate CR of the grid cells in the Level II coverage area is [25%, 49%]. This area corresponds to the edge transition area or the area with relatively loose contact, and the heat demand is low. In subsequent physical wiring, the wiring density in this area is significantly reduced.

[0087] Level I Extremely Low Coverage Area: The coverage rate (CR) of the grid cells in the Level I coverage area is less than 25%. This area corresponds to corners or edges where there is no contact or where heat is extremely easily dissipated. In subsequent physical wiring, this area will have fewer or no major heat-generating units installed, i.e., it will be left blank.

[0088] The aforementioned four-level partition labels serve as the output targets of the classification subtask. During the model training phase, supervised learning is conducted using a large number of labeled samples, enabling the classification subtask to learn to automatically determine the coverage partition level of each grid cell based on the multi-dimensional feature vector.

[0089] Thus, through the aforementioned four-level zoning model, a mapping relationship was established from continuous coverage rate values ​​to discrete zoning level labels. This grading model divides the surface of the heating blanket into four levels according to the degree of contact and heat demand, intuitively reflecting the differentiated heat demands of different body parts of newborns: high-frequency contact areas such as the core and torso correspond to Level IV, medium-contact areas such as the limbs correspond to Level III, edge transition areas correspond to Level II, and areas with no contact or extremely high heat dissipation correspond to Level I. This grading result provides a constraint basis for subsequent regression subtasks to accurately predict wiring density within the corresponding density ranges, and also directly guides the decision-making on the arrangement density of heating wires during physical wiring.

[0090] According to a second aspect of this embodiment, a newborn bionic heating blanket is provided, comprising an outer covering layer, a flexible heating layer, a heat preservation layer, and a temperature control module; the outer covering layer, the flexible heating layer, and the heat preservation layer are stacked; heating wires are arranged on the flexible heating layer; the temperature control module is electrically connected to the heating wires and is used to control the heating power of the heating wires; the flexible heating layer includes multiple regions with different heating wire arrangement densities, and the heating wire arrangement density is determined by the method described in the first aspect of this disclosure.

[0091] In this embodiment of the invention, a newborn bionic heating blanket with differentiated density wiring is provided. The heating wire arrangement density of the heating blanket is designed and determined according to the method described in Embodiment 1, thereby achieving precise matching with the contact characteristics of the newborn's body surface in terms of the underlying physical structure.

[0092] It should be noted that in the neonatal bionic heating blanket provided in this embodiment, the material selection, layering layout, and basic temperature control circuit design of the outer covering layer, the flexible heating layer, the insulation layer, and the temperature control module can all be implemented using existing mature technologies in the field, and this disclosure does not specifically limit them. For example, the outer covering layer can be made of medical-grade soft fabric, the insulation layer can be made of conventional heat insulation material, and the temperature control module can use existing PID temperature control circuits or microprocessor control schemes. The inventive concept and core structural improvement of this disclosure lies in the layout density of the heating wires in the flexible heating layer. That is, through the method described in Embodiment 1, based on the differentiated heating wire arrangement scheme determined after the surface contact imprint image of the neonatal physical model is processed by gridding quantization and predicted by a machine learning model, the precise spatial matching of heat output and neonatal surface contact characteristics is achieved from the underlying physical structure.

[0093] In practice, the traditional method of uniformly coiled heating wires in heating blankets was abandoned. Instead, differentiated density wiring was directly implemented during the physical equipment manufacturing stage based on the grid coverage feature matrix predicted by the neonatal physical model.

[0094] In terms of structural components, the proposed neonatal bionic warming blanket includes an outer covering layer, a flexible heating layer, a heat insulation layer, and a basic temperature control module. The outer covering layer, the flexible heating layer, and the heat insulation layer are stacked, with the flexible heating layer located between the outer covering layer and the heat insulation layer. The outer covering layer, being the outermost layer, directly contacts the external environment and the user, and is typically made of soft fabric, serving protective, aesthetic, and tactile functions. The flexible heating layer is the core functional layer, on which heating wires are arranged in a differentiated pattern based on predicted wiring density. The heat insulation layer is located below the flexible heating layer (i.e., on the side furthest from the body) to reduce downward heat loss and improve thermal efficiency. The temperature control module is electrically connected to the heating wires and controls their heating power to achieve basic temperature regulation.

[0095] Regarding the heating wire arrangement density, the flexible heating layer includes multiple regions with different heating wire arrangement densities, and the heating wire arrangement density of each region is determined by the method described in Example 1. Specifically, based on the wiring density predicted by each grid cell output by the machine learning model in the method described in Example 1, the predicted wiring density is converted into vector layout information of the automatic wiring device, driving the automatic wiring device to physically lay the heating wires on the flexible heating layer according to the differentiated density scheme. Thus, the following multiple regions are formed on the flexible heating layer.

[0096] The first zone corresponds to the Level IV high-coverage zone (high-frequency contact zone) on the newborn's back and core torso. Within this zone, a high-density, continuously arranged heating wire network is installed to meet the extremely high heat transfer efficiency requirements. For areas of the newborn's core and torso with high heat demand, the average coverage of the heating wire network in this zone is set to be greater than or equal to 75%. The heating wires in this zone are arranged in a continuous, winding, and dense pattern to ensure concentrated and uniform heat output.

[0097] The second zone corresponds to the Level III higher coverage area for the limbs and other parts of the newborn's body. A medium-density layout is used in this zone. For the upper and lower limbs of the newborn, the average coverage rate of the heating wire network in this zone is set at 50% to 74%.

[0098] The third zone corresponds to the Level II low-coverage area at the edge transition. In this zone, wiring density is significantly reduced. Corresponding to the non-core contact area at the edge of the heating blanket, the average coverage of the heating wire network in this zone is set at 25% to 49%.

[0099] The fourth zone corresponds to the Level I extremely low coverage area, where there is no contact or heat loss is extremely easy. In this area, the main heating units are reduced or completely omitted, i.e., left blank. For corner areas where there is no risk of contact or heat loss is extremely easy, the average coverage of the heating wire network in this area is set to be less than 25%, or even non-existent.

[0100] Through the above-described differentiated density wiring design, the heating blanket provided in this embodiment achieves the following technical effects.

[0101] First, the ultimate physical heat matching. This design achieves on-demand heating from the underlying device structure, providing high-density heat output to areas of close contact with the newborn's skin, and reducing or stopping heat output to areas of loose or no contact, completely replacing the traditional solution that relies on sensor-based delayed power-off for zoned control.

[0102] Second, it eliminates safety hazards and improves thermal efficiency. Prioritizing coverage of high-frequency contact areas significantly shortens the arrival time to the target temperature zone (36.5℃ to 37.5℃), while simultaneously cutting off ineffective heating from inefficient contact areas. This reduces the risk of localized hot spot accumulation (scalds, erythema) and edge cold stress from the physical source, and significantly lowers the maximum temperature difference and peak temperature on the surface of the heating blanket.

[0103] Therefore, the neonatal bionic heating blanket provided in this embodiment achieves precise spatial matching between heat output and the contact characteristics of the newborn's skin surface from the physical underlying structure by using the differentiated heating wire arrangement density determined by the above method. This heating blanket uses high-density wiring in areas of high contact demand and low-density or blank areas. Compared with existing uniform wiring schemes that rely on sensors for post-adjustment, this fundamentally improves heat transfer efficiency and physiological adaptability, and reduces the safety risks of localized overheating burns and edge cold stress.

[0104] In addition, refer to Figure 1 As shown, according to a third aspect of this embodiment, a storage medium is provided. The storage medium includes a stored program, wherein, when the program is executed, a processor performs any of the methods described above.

[0105] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0106] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0107] Example 2

[0108] Figure 5 A design system for the heating wire layout of a neonatal warming blanket according to this embodiment is shown, which corresponds to the method described according to Embodiment 1. Reference Figure 5 As shown, the system includes: an image acquisition module 510, used to acquire surface contact imprint images obtained by simulating wrapping a newborn physical model; a density distribution feature matrix generation module 520, used to perform gridding and quantization processing on the surface contact imprint images to obtain multiple grid cells and generate corresponding density distribution feature matrices, where each element in the density distribution feature matrix represents the effective contact area ratio of the corresponding grid cell; a wiring density prediction module 530, used to construct multi-dimensional feature vectors for each grid cell based on the density distribution feature matrix, and use a pre-trained machine learning model to process the multi-dimensional feature vectors to determine the predicted wiring density of each grid cell; and a vector layout information generation module 540, used to generate vector layout information for controlling the automatic wiring equipment to perform differentiated wiring operations on the heating layer of the heating blanket according to the predicted wiring density.

[0109] It should be noted that the design system for the heating wire layout of the newborn heating blanket provided in this embodiment can realize all the functions and steps in the above method embodiments, solve the same technical problems, and achieve the same technical effects. The similarities will not be repeated here.

[0110] Example 3

[0111] Figure 6 A design system for the heating wire layout of a neonatal warming blanket according to this embodiment is shown, which corresponds to the method described according to Embodiment 1. Reference Figure 6As shown, the system includes: a processor 610; and a memory 620 connected to the processor 610, used to provide the processor 610 with instructions to process the following steps: acquiring a body surface contact imprint image obtained by simulating wrapping a newborn physical model; performing gridding and quantization processing on the body surface contact imprint image to obtain multiple grid cells and generating a corresponding density distribution feature matrix, wherein each element in the density distribution feature matrix represents the effective contact area ratio of the corresponding grid cell; constructing a multi-dimensional feature vector for each grid cell based on the density distribution feature matrix, and processing the multi-dimensional feature vector using a pre-trained machine learning model to determine the predicted wiring density of each grid cell; and generating vector layout information for controlling the automatic wiring equipment to perform differentiated wiring operations on the heating layer of the heating blanket according to the predicted wiring density.

[0112] It should be noted that the design system for the heating wire layout of the newborn heating blanket provided in this embodiment can realize all the functions and steps in the above method embodiments, solve the same technical problems, and achieve the same technical effects. The similarities will not be repeated here.

[0113] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0114] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0115] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0116] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0117] 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 can be implemented in hardware or as a software functional unit.

[0118] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0119] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A design method for the heating wire layout of a newborn warming blanket, characterized in that, include: Obtain images of skin contact imprints by simulating wrapping a physical model of a newborn; The contact imprint image on the body surface is subjected to gridding and quantization processing to obtain multiple grid cells and generate a corresponding density distribution feature matrix. Each element in the density distribution feature matrix represents the effective contact area ratio of the corresponding grid cell. Based on the density distribution feature matrix, a multidimensional feature vector is constructed for each grid cell, and a pre-trained machine learning model is used to process the multidimensional feature vector to determine the predicted wiring density of each grid cell. Based on the predicted wiring density, vector layout information is generated to control the automatic wiring equipment to perform differentiated wiring operations on the heating layer of the heating blanket.

2. The method according to claim 1, characterized in that, The operation of performing gridded quantization processing on the contact imprint image on the body surface to obtain multiple grid cells includes: Extract the boundary of the target contact area in the body surface contact imprint image and construct the corresponding binary mask; Perform grayscale conversion and threshold segmentation on the effective contact area determined by the binary mask to extract the contact imprint area; Based on the vertices of the contact imprint region, a local parameterized coordinate system is constructed, and the contact imprint region is mapped to a standard two-dimensional plane of a preset size through affine transformation. The standard two-dimensional plane is discretized according to a preset resolution to obtain the multiple grid cells.

3. The method according to claim 1, characterized in that, The operations for generating the corresponding density distribution feature matrix include: Traverse each of the aforementioned grid cells and calculate the total pixel area of ​​each grid cell. and the pixel area of ​​the effective contact imprint within the grid cell. Where i and j represent the row index and column index of the grid cell in the spatial arrangement, respectively; The coverage of each grid cell is calculated using the following formula. As the percentage of the effective contact area: ； Coverage of all grid cells The density distribution feature matrix is ​​formed by summarizing the row and column indices in the spatial arrangement.

4. The method according to claim 1, characterized in that, The operation of constructing multidimensional feature vectors for each grid cell based on the density distribution feature matrix includes: The coverage of the current grid cell itself is obtained from the density distribution feature matrix and used as the first dimension feature of the current grid cell; Within a preset neighborhood window centered on the current grid cell, calculate the statistical characteristic value of the coverage of all grid cells within the preset neighborhood window, and use it as the second dimension feature of the current grid cell. The statistical characteristic value includes the mean and / or standard deviation. The row and column indices of the current grid cell in the spatial arrangement are normalized to obtain the normalized coordinates of the spatial position, which serve as the third dimension feature of the current grid cell. The first dimension feature, the second dimension feature, and the third dimension feature are combined to obtain the multidimensional feature vector.

5. The method according to claim 1, characterized in that, The machine learning model is a multi-task learning network, and the operation of processing the multi-dimensional feature vector using the pre-trained machine learning model to determine the predicted wiring density of each grid cell includes: The multidimensional feature vector is input into the multi-task learning network; The classification subtask in the multi-task learning network generates a coverage partition level label for each grid cell, which is used to determine the density range to which the wiring density of the grid cell belongs. The regression subtask in the multi-task learning network generates the predicted wiring density for each grid cell within the density range determined by the coverage partition level label. The global loss function of the multi-task learning network during training is composed of a weighted joint optimization of the classification loss of the classification sub-task and the regression loss of the regression sub-task.

6. The method according to claim 5, characterized in that, The coverage partition level labels include the following four levels: Level IV coverage area, wherein the coverage rate CR of the grid cells within the Level IV coverage area is ≥75%; Level III coverage area, wherein the coverage rate CR of the grid cells in the Level III coverage area is [50%, 74%]; Level II coverage area, wherein the coverage rate CR of the grid cells in the Level II coverage area is [25%, 49%]; Level I coverage area, wherein the coverage rate CR of the grid cells in the Level I coverage area is less than 25%.

7. A newborn bionic warming blanket, characterized in that, It includes an outer covering layer, a flexible heating layer, an insulation layer, and a temperature control module; The outer covering layer, the flexible heating layer, and the heat insulation layer are stacked in a stacked manner; Heating wires are arranged on the flexible heating layer; The temperature control module is electrically connected to the heating wire and is used to control the heating power of the heating wire; The flexible heating layer includes multiple regions with different heating wire arrangement densities, and the heating wire arrangement density is determined by the method of any one of claims 1 to 6.

8. A storage medium, characterized in that, The storage medium includes a stored program, wherein, when the program is executed, a processor performs the method according to any one of claims 1 to 6.

9. A design system for the heating wire layout of a newborn warming blanket, characterized in that, include: The image acquisition module is used to acquire images of body surface contact imprints obtained by simulating wrapping a physical model of a newborn. The density distribution feature matrix generation module is used to perform gridding and quantization processing on the body surface contact imprint image to obtain multiple grid cells and generate a corresponding density distribution feature matrix. Each element in the density distribution feature matrix represents the effective contact area ratio of the corresponding grid cell. The wiring density prediction module is used to construct a multi-dimensional feature vector for each grid cell based on the density distribution feature matrix, and to process the multi-dimensional feature vector using a pre-trained machine learning model to determine the predicted wiring density of each grid cell. The vector layout information generation module is used to generate vector layout information based on the predicted wiring density to control the automatic wiring equipment to perform differentiated wiring operations on the heating layer of the heating blanket.

10. A design system for the heating wire layout of a newborn warming blanket, characterized in that, include: processor; A memory, connected to the processor, for providing the processor with instructions to perform the following processing steps: Obtain images of skin contact imprints by simulating wrapping a physical model of a newborn; The contact imprint image on the body surface is subjected to gridding and quantization processing to obtain multiple grid cells and generate a corresponding density distribution feature matrix. Each element in the density distribution feature matrix represents the effective contact area ratio of the corresponding grid cell. Based on the density distribution feature matrix, a multidimensional feature vector is constructed for each grid cell, and a pre-trained machine learning model is used to process the multidimensional feature vector to determine the predicted wiring density of each grid cell. Based on the predicted wiring density, vector layout information is generated to control the automatic wiring equipment to perform differentiated wiring operations on the heating layer of the heating blanket.

Images