A diabetic foot ulcer risk early warning system and method based on multi-modal sensing
By using multimodal sensing units and modular processing, accurate early warning of diabetic foot ulcer risk is achieved, solving the problems of low monitoring efficiency, susceptibility to interference, and poor adaptability in existing technologies, and achieving efficient and stable early warning results.
Patent Information
- Application Number
- CN202610220586.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-02-24
- Publication Date
- 2026-06-09
AI Technical Summary
Existing diabetic foot ulcer risk warning technologies suffer from low monitoring efficiency, poor real-time performance, difficulty in capturing weak signals in the early stages of ulceration, simple signal processing algorithms in multimodal sensing systems, susceptibility to environmental interference, poor contact stability, and inability to adapt to individual differences, leading to false or missed warnings.
The system employs a multimodal sensing unit combined with a preprocessing module, an amplification module, an interference suppression module, and a control module. Through multimodal collaborative design, it achieves closed-loop processing of the entire process from signal acquisition to data output, including fusion filtering algorithms, combined amplification structures, and multi-dimensional interference suppression. It can detect skin contact status in real time and dynamically adjust parameters.
It accurately captures weak signals in the early stage of ulceration, improves system stability and adaptability, reduces power consumption, enables long-term wear, improves the accuracy and reliability of early warning, and is suitable for patients at different stages of the disease.
Smart Images
Figure CN122163144A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of medical and health monitoring technology, and in particular to a risk early warning system and method for diabetic foot ulcers based on multimodal sensing. Background Technology
[0002] Diabetic foot is one of the most serious chronic complications of diabetes. Its core cause is diabetic neuropathy and vascular disease, leading to impaired sensation and blood circulation in the foot. Minor skin injuries go unnoticed, developing into slow-healing ulcers, and in severe cases, even requiring amputation. This severely impacts the patient's physical and mental health and quality of life. Clinical practice shows that accurately capturing and providing early warnings of abnormal physiological signals in the foot during the early stages of ulcer formation can effectively reduce the incidence of ulcers. Therefore, diabetic foot ulcer risk warning technology has become a key research focus in the medical and health field.
[0003] Currently, existing technologies for early warning of diabetic foot ulcer risk can be mainly divided into three categories: The first category is traditional manual monitoring technology, which involves medical staff regularly monitoring the patient's foot temperature and pressure, or the patient observing the condition of their feet. This type of technology relies on manual operation, has low monitoring efficiency and poor real-time performance, and cannot capture the subtle physiological signal differences in the early stage of ulceration, easily leading to delayed warnings. The second category is single-modal sensing early warning technology, which collects foot signals only through one modality of pressure sensing, temperature sensing, or blood oxygenation sensing. This type of technology cannot comprehensively reflect the physiological state of the foot, and the abnormal differences in single-modal signals in the early stage of ulceration are extremely weak, making it difficult to accurately capture with a single sensing method. The first category is multimodal sensing-related early warning technology. Although it integrates multiple sensing modes, it has obvious technical defects. The signal processing algorithm is simple and does not have a specific processing mechanism designed for the weak signal differences in the early stage of ulceration. It is easily affected by environmental temperature, light and other interference, which can lead to signal distortion. The sensor module has poor stability in contact with the skin. The accumulation of sweat and keratin can affect the sensing accuracy. In addition, there is no targeted contact state detection and calibration mechanism. The system has high power consumption and complex structure, making it difficult to achieve long-term foot patch wear. Fourth, the threshold values of the early warning-related parameters are fixed and cannot be adaptively adjusted according to the individual differences of patients at different stages of the disease, resulting in poor adaptability.
[0004] Therefore, a risk warning system and method for diabetic foot ulcers based on multimodal sensing is proposed. Summary of the Invention
[0005] This application aims to at least partially solve one of the technical problems in the aforementioned technologies.
[0006] To achieve the above objectives, the first aspect of this application proposes a diabetic foot ulcer risk early warning system based on multimodal sensing, including a multimodal sensing unit, a preprocessing module, an amplification module, an interference suppression module, and a control module;
[0007] The multimodal sensing unit, preprocessing module, amplification module, interference suppression module, and control module are electrically connected in sequence.
[0008] The multimodal sensing unit is used to collect multimodal physiological signals of the foot. The preprocessing module is used to denoise and extract features from the collected signals. The amplification module is used to amplify the subtle differences in foot physiological signals in the early stage of ulceration. The interference suppression module is used to suppress signal interference caused by ambient temperature and light. The control module is used to output the processed signal.
[0009] This technical solution solves the problems of poor multi-module coordination and signal processing gaps in existing technologies, and realizes closed-loop processing of the entire process from signal acquisition to data output. It not only ensures the effective capture of weak signals in the early stage of ulcers, but also improves the stability of system operation. At the same time, the structure is reasonable and lightweight, suitable for foot wearing scenarios, and makes up for the shortcomings of existing systems that are complex in structure and difficult to wear for a long time.
[0010] In addition, the diabetic foot ulcer risk early warning system based on multimodal sensing proposed in this application may also have the following additional technical features:
[0011] As a further description of the above technical solution:
[0012] The multimodal sensing unit includes a core sensing module and an auxiliary sensing module. The core sensing module is used to acquire foot blood oxygen saturation and temperature signals, while the auxiliary sensing module is a skin impedance sensing module used to acquire foot skin impedance signals and detect the contact state between the skin and the sensing unit in real time. This technical solution overcomes the shortcomings of existing multimodal sensing that only focus on core signals and ignore contact state. By coordinating the core and auxiliary sensors, signal distortion caused by poor contact is avoided, improving the accuracy of multimodal signal acquisition. At the same time, the key parameter types of the core sensor are clearly defined, avoiding the problems of excessive system complexity and increased power consumption caused by redundant sensors, thus balancing accuracy and lightweight design.
[0013] As a further description of the above technical solution:
[0014] The preprocessing module employs a fusion filtering algorithm, which is either a fusion of adaptive median filtering and wavelet packet decomposition, or a fusion of lightweight morphological filtering and adaptive threshold filtering. This fusion filtering algorithm removes high-frequency noise and isolated noise from the signal, extracts low-frequency feature components representing subtle differences in foot physiological signals, and discards redundant high-frequency components. This technical solution addresses the shortcomings of existing single filtering algorithms, which struggle to balance denoising effectiveness with the preservation of subtle signals. The fusion filtering algorithm effectively removes environmental noise and signal noise while fully preserving the subtle differences in foot physiological signals during the early stages of ulceration. It also provides two fusion methods to adapt to different hardware computing power scenarios, balancing accuracy and lightweight design, and overcoming the lack of flexibility in existing preprocessing algorithms.
[0015] As a further description of the above technical solution:
[0016] The amplification module adopts a combined amplification structure, which includes a combination of an adaptive gain adjustment circuit and a differential amplification circuit. The gain range is dynamically adjustable from 10 to 50 times, and the circuit power consumption is ≤1μW.
[0017] The combination of a fixed gain circuit and an adaptive threshold filter circuit, with two preset fixed gain levels, can automatically match the corresponding gain according to the amplitude of the previously sampled signal. Moreover, the algorithm occupies ≤3% of the terminal computing power. This technical solution solves the defects of existing amplification modules, such as fixed gain, easy signal saturation or insufficient amplification of weak signals, and excessive power consumption and computing power. The two implementation methods can flexibly adapt to different hardware resource scenarios. The first method focuses on accurate amplification, while the second method focuses on lightweight design.
[0018] As a further description of the above technical solution:
[0019] The interference suppression module includes a miniature environmental reference sensing unit for collecting ambient temperature and light intensity signals. The module employs a combination of difference cancellation and threshold adaptive shielding to deduct interference from ambient temperature on the foot temperature signal. When ambient light exceeds a set threshold, it automatically shields the interference signal and compensates through interpolation of adjacent valid sampling points. Combined with feature filtering from the preprocessing module, residual interference is eliminated. This technical solution overcomes the shortcomings of existing interference suppression methods, which are often limited in scope and have weak anti-interference capabilities. It comprehensively suppresses interference from ambient temperature and light, ensuring the accuracy of signal detection. The miniature environmental reference sensing unit is small in size and has low power consumption, without increasing system complexity, thus compensating for the shortcomings of existing anti-interference solutions, which are often complex in structure and have poor adaptability.
[0020] The second aspect of this application proposes a method for early warning of diabetic foot ulcer risk based on multimodal sensing, including the following steps:
[0021] 1) Signal acquisition: Multimodal physiological signals of foot blood oxygen saturation, temperature and skin impedance are acquired through a multimodal sensing unit, while ambient temperature and light intensity reference signals are acquired through a miniature environmental reference sensing unit.
[0022] 2) Signal preprocessing: The preprocessing module uses a fusion filtering algorithm to denoise the collected foot physiological signals and extract feature components that characterize the subtle differences in the signals.
[0023] 3) Signal amplification: The amplification module adopts a combined amplification structure to specifically amplify the subtle differences in preprocessed foot physiological signals and avoid signal saturation;
[0024] 4) Interference suppression: By combining the interference suppression module with the environmental reference signal, the amplified signal is subjected to interference suppression processing to eliminate residual environmental interference;
[0025] 5) Signal output: The control module outputs pre-processed, amplified, and interference-suppressed foot physiological signals to provide accurate data support for ulcer risk warning and judgment;
[0026] 6) Parameter calibration: Based on clinically measured weak signal samples in the early stage of ulceration, a preset signal difference threshold range is set. The control module compares the processed signal in step 4) with the preset threshold range in real time, dynamically corrects the filtering threshold of the preprocessing module and the gain parameter of the amplification module, realizes the integration of calibration and interference suppression, and improves the stability of signal detection.
[0027] This technical solution addresses the shortcomings of existing early warning methods, such as cumbersome steps, disjointed logic, and difficulty in capturing weak signals in the early stages of ulceration. The solution features a coherent and targeted approach, with each step addressing core issues in signal acquisition, processing, amplification, and anti-interference. The process is simple and highly operable, achieving accurate data output without complex operations. It is also system-compatible, improving the accuracy and reliability of early warning and overcoming the poor adaptability and lack of practicality of existing methods.
[0028] In addition, the method for early warning of diabetic foot ulcer based on multimodal sensing proposed in this application may also have the following additional technical features:
[0029] As a further description of the above technical solution:
[0030] The multimodal sensing unit described in step 1) detects the skin contact status in real time by collecting skin impedance signals through the auxiliary sensing module. When the skin impedance fluctuation exceeds the set threshold, it automatically adjusts the sampling frequency of the core sensing module and starts a simple denoising algorithm to avoid signal distortion caused by poor contact. This technical solution solves the defects of the existing signal acquisition steps that ignore the contact status and are prone to signal distortion due to poor contact. By monitoring the contact status in real time through the skin impedance signal, it realizes automatic identification and adaptive adjustment of contact abnormalities without manual intervention, which improves the automation and accuracy of signal acquisition. The simple denoising algorithm does not consume too much computing power and balances adjustment effect with lightweight design.
[0031] As a further description of the above technical solution:
[0032] The specific processing steps of the fusion filtering algorithm described in step 2) are as follows:
[0033] If the adaptive median filtering and wavelet packet decomposition algorithm are combined, the high-frequency noise of light is first filtered out by the adaptive median filtering, and then the blood oxygen and temperature signals are decomposed into three layers by wavelet packet decomposition to extract the low-frequency feature components.
[0034] If morphological filtering and adaptive threshold filtering are combined, morphological filtering is first used to remove isolated noise points in the signal, and then adaptive threshold filtering is used to dynamically distinguish weak difference signals from interference signals. This technical solution solves the defects of existing preprocessing steps, such as unclear filtering algorithm operation and difficulty in balancing denoising and weak signal preservation. It clarifies the specific processing flow of the two fused filtering algorithms, is highly operable, and adapts to different hardware scenarios. Through step-by-step processing, it further improves the denoising effect and the accuracy of weak feature extraction, laying the foundation for subsequent signal amplification and interference suppression.
[0035] As a further description of the above technical solution:
[0036] The specific process of interference suppression processing described in step 4) is as follows:
[0037] First, the difference between the foot temperature signal and the ambient temperature reference signal is calculated using the difference cancellation method to deduct the ambient temperature interference.
[0038] Next, it determines whether the ambient light intensity exceeds the set threshold. If it does, it blocks the interference signal and performs interpolation compensation. If it does not exceed the threshold, it outputs the signal directly.
[0039] Finally, by combining the feature components extracted in step 2), interference residues in the signal are further removed. This technical solution solves the defects of existing interference suppression steps, such as single operation and incomplete interference removal. Layered processing can specifically solve different types of environmental interference, and interference removal is more thorough, ensuring signal accuracy. The interpolation compensation step avoids data gaps caused by signal loss. Combined with preprocessing features, the signal is further purified, improving the quality of subsequent signal output and providing more reliable data support for early warning judgment.
[0040] Advantages of this invention:
[0041] According to the multimodal sensing-based risk warning system and method for diabetic foot ulcers in this application, a multimodal sensing collaborative design is adopted, combining core and auxiliary sensors to avoid signal distortion caused by poor contact.
[0042] The combination of fusion filtering and amplification structure can accurately capture the subtle differences in weak signals in the early stage of ulceration, while providing a variety of adaptation schemes to balance accuracy and lightweight design.
[0043] Through a multi-dimensional interference suppression mechanism, it effectively shields against environmental temperature and light interference, thereby improving signal accuracy;
[0044] The methodology is coherent, achieving a complete closed loop from signal acquisition to output. Dynamic parameter calibration has been added to adapt to individual differences among patients at different disease stages.
[0045] The overall system has a reasonable structure and low power consumption, and can be applied to the foot for a long time, which greatly improves the reliability of the early warning system.
[0046] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description
[0047] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:
[0048] Figure 1 This is a schematic block diagram of a diabetic foot ulcer risk warning system based on multimodal sensing according to an embodiment of this application;
[0049] Figure 2 This is a flowchart illustrating a method for early warning of diabetic foot ulcer risk based on multimodal sensing according to an embodiment of this application. Detailed Implementation
[0050] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.
[0051] The following description, in conjunction with the accompanying drawings, illustrates an embodiment of a diabetic foot ulcer risk warning system and method based on multimodal sensing.
[0052] like Figure 1 As shown, the diabetic foot ulcer risk warning system based on multimodal sensing in Embodiment 1 of this application may include a multimodal sensing unit, a preprocessing module, an amplification module, an interference suppression module, and a control module.
[0053] The multimodal sensing unit, preprocessing module, amplification module, interference suppression module, and control module are electrically connected in sequence. The multimodal sensing unit is used to collect multimodal physiological signals of the foot. The preprocessing module is used to denoise and extract features from the collected signals. The amplification module is used to amplify the subtle differences in foot physiological signals in the early stage of ulceration. The interference suppression module is used to suppress signal interference caused by ambient temperature and light. The control module is used to output the processed signal. This embodiment adopts an integrated architecture of sensing, processing, amplification, anti-interference, and control. The multimodal sensing unit collects core physiological signals of the foot. The preprocessing module filters out redundant noise and extracts effective features. Then, the amplification module amplifies the subtle differences in signals in the early stage of ulceration. The interference suppression module eliminates environmental interference. Finally, the control module coordinates the timing of each module to ensure the continuity and accuracy of signal processing and provide high-quality data input for early warning judgment.
[0054] like Figure 1 As shown:
[0055] The multimodal sensing unit includes a core sensing module and an auxiliary sensing module. The core sensing module is used to collect foot blood oxygen saturation and temperature signals. The auxiliary sensing module is a skin impedance sensing module used to collect foot skin impedance signals and detect the contact state between the skin and the sensing unit in real time. In this embodiment, by focusing on the most representative physiological signals of the foot in the early stage of ulceration (blood oxygen reflects the blood circulation status, and temperature reflects the local inflammation trend), the auxiliary sensing module judges the degree of contact between the sensing unit and the skin by the changes in skin impedance signals. For example, the accumulation of sweat and keratin can cause abnormal impedance fluctuations, thereby providing feedback on the contact state and providing a basis for subsequent signal calibration.
[0056] like Figure 1 As shown:
[0057] The preprocessing module employs a fusion filtering algorithm, which is either a fusion of adaptive median filtering and wavelet packet decomposition, or a fusion of lightweight morphological filtering and adaptive threshold filtering. The preprocessing module uses this fusion filtering algorithm to remove high-frequency noise and isolated noise from the signal, extract low-frequency feature components that characterize subtle differences in foot physiological signals, and discard redundant high-frequency components. This embodiment combines the advantages of two single filters through the fusion filtering algorithm. Adaptive median filtering can dynamically adjust the window size to accurately filter out high-frequency noise caused by light interference, while wavelet packet decomposition can perform layered processing of the signal to accurately extract low-frequency features corresponding to subtle differences. Lightweight morphological filtering can quickly remove isolated noise, and adaptive threshold filtering can dynamically distinguish between valid and interference signals.
[0058] like Figure 1 As shown:
[0059] The amplification module adopts a combined amplification structure, including a combination of an adaptive gain adjustment circuit and a differential amplification circuit. The gain range is dynamically adjustable from 10 to 50 times, and the circuit power consumption is ≤1μW. It also includes a combination of a fixed gain circuit and an adaptive threshold filter circuit, with two preset fixed gain levels. The corresponding gain can be automatically matched according to the amplitude of the previously sampled signal, and the algorithm occupies ≤3% of the terminal computing power. In this embodiment, the differential amplification circuit suppresses common-mode interference (such as ambient temperature drift), and the adaptive gain adjustment circuit dynamically adjusts the gain according to the signal amplitude. It amplifies weak differential signals and maintains a low gain for normal signals to avoid signal saturation. The fixed gain circuit simplifies the structure and reduces power consumption. The adaptive threshold filter circuit works with gain matching to ensure that weak signals are not masked by interference signals after amplification, while controlling the computing power consumption.
[0060] like Figure 1 As shown:
[0061] The interference suppression module includes a miniature environmental reference sensing unit for collecting ambient temperature and light intensity signals. The module employs a combination of difference cancellation and threshold adaptive shielding to deduct interference from ambient temperature on the foot temperature signal. When ambient light exceeds a set threshold, interference signals are automatically shielded and compensated through interpolation of adjacent valid sampling points. Combined with feature filtering from the preprocessing module, residual interference is eliminated. In this embodiment, the miniature environmental reference sensing unit collects an environmental reference signal. The difference between the foot temperature signal and the ambient temperature signal is calculated using the difference cancellation method to directly deduct interference caused by ambient temperature drift. By setting a light intensity threshold, it determines whether strong light interference exists. When the threshold is exceeded, interference signals are shielded and interpolated to avoid signal loss. Combined with feature filtering from the preprocessing module, residual interference is further eliminated, achieving multi-dimensional anti-interference.
[0062] like Figure 2As shown in Embodiment 2 of this application, the method for early warning of diabetic foot ulcer risk based on multimodal sensing is described.
[0063] In summary, a diabetic foot ulcer risk early warning system and method based on multimodal sensing according to embodiments of this application includes the following steps:
[0064] 1) Signal acquisition: Multimodal physiological signals of foot blood oxygen saturation, temperature and skin impedance are acquired through a multimodal sensing unit, while ambient temperature and light intensity reference signals are acquired through a miniature environmental reference sensing unit.
[0065] 2) Signal preprocessing: The preprocessing module uses a fusion filtering algorithm to denoise the collected foot physiological signals and extract feature components that characterize the subtle differences in the signals.
[0066] 3) Signal amplification: The amplification module adopts a combined amplification structure to specifically amplify the subtle differences in preprocessed foot physiological signals and avoid signal saturation;
[0067] 4) Interference suppression: By combining the interference suppression module with the environmental reference signal, the amplified signal is subjected to interference suppression processing to eliminate residual environmental interference;
[0068] 5) Signal output: The control module outputs pre-processed, amplified, and interference-suppressed foot physiological signals to provide accurate data support for ulcer risk warning and judgment;
[0069] 6) Parameter calibration: Based on clinically measured weak signal samples in the early stage of ulceration, a preset signal difference threshold range is set. The control module compares the processed signal in step 4) with the preset threshold range in real time, dynamically corrects the filtering threshold of the preprocessing module and the gain parameter of the amplification module, realizes the integration of calibration and interference suppression, and improves the stability of signal detection.
[0070] In this embodiment, a logically progressive process of signal acquisition, preprocessing, amplification, interference suppression, and output is adopted. First, foot physiological signals and environmental reference signals are acquired simultaneously. Then, noise reduction and effective feature extraction are performed through preprocessing. Next, weak difference signals are amplified in a targeted manner. Then, environmental interference is eliminated. Finally, high-quality data is output to ensure that each step provides support for subsequent early warning judgment and achieves accurate capture and transmission of weak signals in the early stage of ulceration.
[0071] like Figure 2 As shown:
[0072] In step 1), the multimodal sensing unit detects the skin contact status in real time by collecting skin impedance signals through the auxiliary sensing module. When the skin impedance fluctuation exceeds the set threshold, the sampling frequency of the core sensing module is automatically adjusted, and a simple denoising algorithm is started to avoid signal distortion caused by poor contact. In this embodiment, the amplitude of the skin impedance signal is positively correlated with the tightness of skin contact. When sweat and keratin accumulate and cause poor contact, the impedance signal will fluctuate abnormally. After exceeding the preset threshold, the system automatically identifies the abnormal contact, increases the sampling frequency of the core sensing module to increase the signal acquisition, and starts a simple denoising algorithm to initially filter out noise caused by poor contact and ensure signal validity.
[0073] like Figure 2 As shown:
[0074] The specific processing steps of the fusion filtering algorithm in step 2) are as follows:
[0075] If an adaptive median filter and wavelet packet decomposition algorithm are fused, the adaptive median filter first removes high-frequency noise from the light, and then the wavelet packet decomposition performs a three-level decomposition on the blood oxygen and temperature signals to extract low-frequency feature components. If a morphological filter and adaptive threshold filter are fused, the morphological filter first removes isolated noise from the signal, and then the adaptive threshold filter dynamically distinguishes weak difference signals from interference signals. In this embodiment, for the two fusion filtering algorithms, the first one first removes noise and then extracts features layer by layer. The adaptive median filter specifically removes high-frequency noise from the light, and the three-level wavelet packet decomposition can accurately locate the low-frequency range corresponding to weak differences, avoiding feature loss. The second one first removes isolated noise and then distinguishes between effective signals and interference. The morphological filter quickly removes noise, and the adaptive threshold filter dynamically adapts to signal fluctuations to ensure that weak difference signals are not misjudged as interference.
[0076] like Figure 2 As shown:
[0077] The specific process of interference suppression in step 4) is as follows:
[0078] First, the foot temperature signal and the ambient temperature reference signal are calculated using the difference cancellation method to subtract ambient temperature interference. Second, it is determined whether the ambient light intensity exceeds a set threshold. If it does, the interference signal is shielded and interpolation compensation is performed. If it does not exceed the threshold, the signal is output directly. Finally, the feature components extracted in step 2) are combined to further remove residual interference in the signal. In this embodiment, the systematic interference of ambient temperature is directly subtracted by the difference cancellation method, and then targeted processing (shielding and compensation) is performed on light interference to avoid signal distortion and loss caused by strong light. Finally, the effective features extracted by preprocessing are compared and screened to remove residual interference signals and ensure signal purity.
[0079] Example 3, further illustrated below with a specific implementation case:
[0080] Multimodal sensing unit: A miniature integrated blood oxygen and temperature sensor (MAX30102) is selected to collect foot blood oxygen saturation and temperature signals. The sampling frequency is preset to 1Hz. The auxiliary sensing module uses a flexible skin impedance sensor (ADS1292) to collect foot skin impedance signals. The sampling frequency is synchronized with the core sensing module. The impedance detection range is 1kΩ-100kΩ. The preset impedance threshold range for good skin contact is 5kΩ-50kΩ. If it exceeds this range, it is judged as abnormal contact.
[0081] Preprocessing module: Employs a fusion of adaptive median filtering and wavelet packet decomposition algorithms. The initial window size for adaptive median filtering is set to 3 sampling points, dynamically adjusted to 2-5 sampling points based on signal fluctuations. Wavelet packet decomposition uses the db4 wavelet basis, performing a 3-level decomposition on the blood oxygen and temperature signals. The 3rd level low-frequency feature components are extracted, while the 1st and 2nd levels of high-frequency redundant components are discarded. The feature extraction formula is as follows:
[0082]
[0083] in For the first Layer Wavelet packet decomposition coefficients ( =3, =1,2,...,8),
[0084] The raw blood oxygen / temperature signal was collected. The decomposition function of the db4 wavelet basis. For sampling point number ( =1,2,...,N, where N is the total number of sampling points).
[0085] Amplification module: A combination of adaptive gain adjustment circuit and differential amplifier circuit is used. The differential amplifier circuit uses the AD8421 chip, with a common-mode rejection ratio ≥140dB. The adaptive gain adjustment circuit uses the PGA280 chip, with a gain range of 10-50 times dynamically adjustable. The gain calculation formula is as follows:
[0086]
[0087] in, Actual output gain (10≤ ≤50), Set as the baseline gain (20x). The values represent the subtle differences in signals detected after preprocessing (blood oxygen difference 3%-5%, temperature difference 0.3-0.8℃). Using the baseline difference value (set to 4% / 0.5℃), the measured power consumption of the circuit is 0.8μW, which meets the design requirement of ≤1μW.
[0088] Interference suppression module: The miniature environmental reference sensing unit uses a DS18B20 temperature sensor (temperature range -10℃ to 85℃, accuracy ±0.1℃) and a TSL2561 light sensor (light range 0-65535 lux). Ambient temperature interference is eliminated using the difference cancellation method, and the calculation formula is as follows:
[0089]
[0090] in, To determine the actual foot temperature after eliminating environmental interference, The foot temperature is collected by the core sensing module. The ambient temperature is collected by the environmental reference sensing unit; the preset light intensity interference threshold is 500 lux. When the detected light intensity is >500 lux, the interference signal is automatically blocked, and linear interpolation compensation is performed using three adjacent valid sampling points. The interpolation formula is as follows:
[0091]
[0092] in The signal value after interpolation compensation. The coordinates of the effective sampling points adjacent to the interference signal. This is the sampling point number corresponding to the interference signal.
[0093] Control module: An STM32L431 microcontroller with a main frequency of 80MHz and a measured computing power utilization of 4.2% is used to coordinate the collaborative work of various modules and output the processed signal data.
[0094] Based on the above system modules, this embodiment selected 50 diabetic patients (disease duration 5-10 years, including 30 with mild neuropathy and 20 with severe ischemia) for testing. The steps are as follows:
[0095] Signal acquisition: The multimodal sensing unit was applied to the forefoot and heel of the patient's foot, areas prone to ulceration. The auxiliary sensing module collected skin impedance signals in real time, the core sensing module collected foot blood oxygen saturation and temperature signals, and the miniature environmental reference sensing unit collected environmental signals simultaneously. Data was collected continuously for 7 days, with 240 sets of data collected each day (1 set every 6 minutes). In actual testing, 3 patients showed abnormal skin contact (impedance > 50kΩ), and the system automatically identified and adjusted the core sensing sampling frequency to 2Hz.
[0096] Signal preprocessing: The raw signals from 50 patients were processed by the preprocessing module. Adaptive median filtering was used to remove high-frequency noise from the light (noise amplitude ≤ 0.05℃ / 0.5%). Wavelet packet decomposition was used to extract three layers of low-frequency feature components. The measured feature extraction accuracy was 98.3%, effectively preserving the slight differences in blood oxygen 3%-5% and temperature 0.3-0.8℃.
[0097] Signal amplification: Based on the difference in the preprocessed signal, the gain is dynamically adjusted by the amplification module. When the blood oxygen difference is 3%, the gain is adjusted to 15 times, and when the difference is 5%, the gain is adjusted to 25 times. When the temperature difference is 0.3℃, the gain is adjusted to 10 times, and when the difference is 0.8℃, the gain is adjusted to 40 times. The amplitude of the amplified signal is 10-50 times that of the original weak signal, with no signal saturation and an amplification error ≤0.2%.
[0098] Interference suppression: Ambient temperature interference was subtracted using the difference cancellation method (ambient temperature 20-28℃, after subtraction, the actual foot temperature error was ≤0.1℃); light interference occurred during the testing of 3 patients (light intensity 600-800 lux), the system automatically shielded the interference signal and performed interpolation compensation, after compensation the signal error was ≤0.03℃ / 0.1%, combined with preprocessing feature screening, the final residual interference was ≤0.02℃ / 0.05%.
[0099] Signal output: The control module outputs processed signal data to support ulcer risk early warning. The measured signal output delay is ≤100ms and the data transmission accuracy is 99.7%.
[0100] Parameter calibration: Based on 100 sets of clinically measured weak signal samples in the early stage of ulceration, the preset blood oxygen difference threshold range is 3%-5%, and the temperature difference threshold range is 0.3-0.8℃. The control module compares the processed signal with the preset threshold range in real time and dynamically corrects the filtering threshold and gain parameters. The correction formula is as follows:
[0101]
[0102] in, These are the corrected parameters (filter threshold / gain). To correct the parameters before, This is a correction factor (set to 0.1). The actual signal value after processing. With the preset threshold midpoint (4% / 0.5℃), the signal detection stability was improved to 99.2% after calibration, the false alarm rate was reduced to 1.8%, and the missed alarm rate was reduced to 0.9%.
[0103] In summary, the multimodal sensing-based diabetic foot ulcer risk warning system and method described in this invention can accurately capture subtle physiological signal differences in the foot during the early stage of ulceration, effectively suppress environmental interference, adapt to different hardware scenarios and diabetic patients at different disease stages, solve many defects of existing technologies, and meet the clinical application requirements in terms of warning accuracy and reliability.
[0104] In the description of this specification, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.
[0105] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.
[0106] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.
Claims
1. A diabetic foot ulcer risk early warning system based on multimodal sensing, characterized in that, It includes a multimodal sensing unit, a preprocessing module, an amplification module, an interference suppression module, and a control module; The multimodal sensing unit, preprocessing module, amplification module, interference suppression module, and control module are electrically connected in sequence. The multimodal sensing unit is used to collect multimodal physiological signals of the foot. The preprocessing module is used to denoise and extract features from the collected signals. The amplification module is used to amplify the subtle differences in foot physiological signals in the early stage of ulceration. The interference suppression module is used to suppress signal interference caused by ambient temperature and light. The control module is used to output the processed signal.
2. The diabetic foot ulcer risk early warning system based on multimodal sensing according to claim 1, characterized in that, The multimodal sensing unit includes a core sensing module and an auxiliary sensing module. The core sensing module is used to collect foot blood oxygen saturation and temperature signals, and the auxiliary sensing module is a skin impedance sensing module used to collect foot skin impedance signals and detect the contact state between the skin and the sensing unit in real time.
3. The diabetic foot ulcer risk early warning system based on multimodal sensing according to claim 1, characterized in that, The preprocessing module employs a fusion filtering algorithm, which is a fusion of adaptive median filtering and wavelet packet decomposition algorithm, or a fusion of lightweight morphological filtering and adaptive threshold filtering. The preprocessing module uses the fusion filtering algorithm to filter out high-frequency noise and isolated noise in the signal, extract low-frequency feature components that characterize the subtle differences in foot physiological signals, and discard redundant high-frequency components.
4. The diabetic foot ulcer risk early warning system based on multimodal sensing according to claim 1, characterized in that, The amplification module adopts a combined amplification structure, which includes a combination of an adaptive gain adjustment circuit and a differential amplification circuit. The gain range is dynamically adjustable from 10 to 50 times, and the circuit power consumption is ≤1μW. The combination of fixed gain circuit and adaptive threshold filter circuit has two preset fixed gain levels, which can automatically match the corresponding gain according to the amplitude of the previous sampled signal, and the algorithm occupies ≤3% of the terminal computing power.
5. The diabetic foot ulcer risk early warning system based on multimodal sensing according to claim 1, characterized in that, The interference suppression module includes a miniature environmental reference sensing unit for collecting ambient temperature and light intensity signals. The interference suppression module uses a combination of difference cancellation and threshold adaptive shielding to deduct the interference of ambient temperature on the foot temperature signal. When the ambient light exceeds the set threshold, the interference signal is automatically shielded and compensated by interpolation of adjacent valid sampling points. Combined with the feature screening of the preprocessing module, the interference residue is removed.
6. The method for early warning of diabetic foot ulcer risk based on multimodal sensing according to any one of claims 1-5, characterized in that, Includes the following steps: 1) Signal acquisition: Multimodal physiological signals of foot blood oxygen saturation, temperature and skin impedance are acquired through a multimodal sensing unit, while ambient temperature and light intensity reference signals are acquired through a miniature environmental reference sensing unit. 2) Signal preprocessing: The preprocessing module uses a fusion filtering algorithm to denoise the collected foot physiological signals and extract feature components that characterize the subtle differences in the signals. 3) Signal amplification: The amplification module adopts a combined amplification structure to specifically amplify the subtle differences in preprocessed foot physiological signals and avoid signal saturation; 4) Interference suppression: By combining the interference suppression module with the environmental reference signal, the amplified signal is subjected to interference suppression processing to eliminate residual environmental interference; 5) Signal output: The control module outputs pre-processed, amplified, and interference-suppressed foot physiological signals to provide accurate data support for ulcer risk warning and judgment; 6) Parameter calibration: Based on clinically measured weak signal samples in the early stage of ulceration, a preset signal difference threshold range is established. The control module compares the processed signal in step 4) with the preset threshold range in real time, dynamically corrects the filtering threshold of the preprocessing module and the gain parameter of the amplification module, realizes the integration of calibration and interference suppression, and improves the stability of signal detection.
7. The method for early warning of diabetic foot ulcer risk based on multimodal sensing according to claim 6, characterized in that, The multimodal sensing unit described in step 1) detects the skin contact status in real time by collecting skin impedance signals through the auxiliary sensing module. When the skin impedance fluctuation exceeds the set threshold, it automatically adjusts the sampling frequency of the core sensing module and starts a simple noise reduction algorithm to avoid signal distortion caused by poor contact.
8. The method for early warning of diabetic foot ulcer risk based on multimodal sensing according to claim 6, characterized in that, The specific processing steps of the fusion filtering algorithm described in step 2) are as follows: If the adaptive median filtering and wavelet packet decomposition algorithm are combined, the high-frequency noise of light is first filtered out by the adaptive median filtering, and then the blood oxygen and temperature signals are decomposed into three layers by wavelet packet decomposition to extract the low-frequency feature components. If morphological filtering and adaptive threshold filtering are combined, isolated noise points in the signal are first removed by morphological filtering, and then weak difference signals and interference signals are dynamically distinguished by adaptive threshold filtering.
9. The method for early warning of diabetic foot ulcer risk based on multimodal sensing according to claim 6, characterized in that, The specific process of interference suppression processing described in step 4) is as follows: First, the difference between the foot temperature signal and the ambient temperature reference signal is calculated using the difference cancellation method to deduct the ambient temperature interference. Next, it determines whether the ambient light intensity exceeds the set threshold. If it does, it blocks the interference signal and performs interpolation compensation. If it does not exceed the threshold, it outputs the signal directly. Finally, by combining the feature components extracted in step 2), interference residues in the signal are further removed.