System and method for non-contact epidermal sensing

Abstract

The present disclosure provides systems and methods for characterizing a wound. For example, a method includes transmitting a mmWave signal to a target tissue; receiving a reflected signal from the target tissue, the reflected signal corresponding to the mmWave signal and including a target signal: repeating the transmitting and receiving steps at multiple x, y locations in a 2D plane to scan the target tissue and at multiple times t at each location; and calculating a wound size of the target based on moisture values determined from the target signals.

Description

SYSTEM AND METHOD FOR NON-CONTACT EPIDERMAL SENSINGCross-Reference to Related Applications

[0001] This application claims priority to U.S. Provisional Application No. 63 / 604,890, filed on November 30, 2023, now pending, the disclosure of which is incorporated herein by reference.Field of the Disclosure

[0002] The present disclosure relates to biometric sensors, and more particularly to noncontact sensors using mmWave radar.Background of the Disclosure

[0003] Nowadays, human epidermal sensing has diverse applications across healthcare, personal care, sports, and various industries. It provides valuable data for maintaining health, comfort, and well-being. Compared with the existing epidermal sensing methods, radiofrequency (RF) sensors can penetrate clothing on the epidermal and sense the epidermal properties in a non-contact way, which makes the sensing process more convenient and user-friendly. The millimeter-wave (mmWave) sensor, as one of the RF sensors, can further provide a much more accurate sensing ability for the high working frequency.Brief Summary of the Disclosure

[0004] The present disclosure is directed to a mmW ave-based human epidermal sensing system (dubbed “mmSkin”). It utilizes the impact of epidermal moisture difference on the mmWave signal strength and extracts multi-functional skin features based on the mmWave signal strength difference, which are further used to realize wound size and moisture value sensing. In addition, to eliminate environmental interference and focus on sensing the target, some embodiments of the system use an environmental-denoise mmWave imaging algorithm to clearly split the signal reflected from the target and the environment.

[0005] A mmWave-based system and method is disclosed to enhance human epidermal sensing. Experimental findings highlight mmSkin’s proficiency in detecting variations in water content through RF signal analysis and its utility in estimating wound size and moisture value. Embodiments of the system may incorporate an innovative environmental-denoise mmWaveimaging algorithm, which effectively differentiates the target signals from environmental noise. The practicality of mmSkin has been substantiated through extensive testing, showcasing its reliable performance even when utilized over dressed wounds. The mmSkin system is a valuable tool for non-invasive skin health monitoring and contributes to the advancement of mmWave sensor technology in epidermal diagnostics.Description of the Drawings

[0006] For a fuller understanding of the nature and objects of the disclosure, reference should be made to the following detailed description taken in conjunction with the accompanying drawings.

[0007] Figure 1 : mmSkin system collects and processes the mmWave signal reflected from the human epidermal covered by the dressing. Water content on the human epidermal influences its response to the mmWave signal, it outputs the wound size based on the signal difference between the wound area and healthy epidermal and epidermal moisture value based on the signal amplitude.

[0008] Figure 2: Human skin water distribution model.

[0009] Figure 3: Framework of the mmSkin system.

[0010] Figure 4: One frame structure of the mmWave radar.

[0011] Figure 5: mmSkin system moves the mmWave radar in the 2D plane and scans the target, the collected raw mmWave signals are processed to generate the mmWave image. Each pixel in the mmWave image is a combination of all 2D scanning signals.

[0012] Figure 6: (a) Traditional mmWave imaging algorithm selects the same distance in the range bin for each scanning location, which makes the selected data in some locations miss the target signal, (b) Our optimized mmWave imaging algorithm selects the distance in the range bin for each scanning location based on its real distance to the target.

[0013] Figure 7: mmSkin System Prototype.

[0014] Figure 8: 2D scanning kit framework.

[0015] Figure 9: Wound phantom with different area size and depth.

[0016] Figure 10: Wound and Moisture phantoms without occlusion and covered occlusion in different depth and fabric types.

[0017] Figure 11 : Three in vitro phantoms for moisture value estimation and their corresponding mmWave image results.

[0018] Figure 12: A diagram of a system according to another embodiment of the present disclosure.

[0019] Figure 13: A chart showing a method according to another embodiment of the present disclosure.Detailed Description of the Disclosure

[0020] With reference to Figure 13, in an aspect, the present disclosure may be embodied as a method 100 for characterizing a wound. The method 100 includes transmitting 103 a mmWave signal to a target tissue and receiving 106 a reflected signal from the target tissue. The reflected signal corresponds to the transmitted mmWave signal. The reflected signal includes a target signal and an environment signal (i.e., noise).

[0021] The transmitting and receiving steps are repeated 109 at multiple x, y locations in a 2D plane to scan the target tissue and at multiple times t at each location. The method 100 may further include denoising 112 the reflected signal to obtain the target signal. For example, denoising the reflected signal may include splitting the reflected signal into the target signal and the environment signal using synthetic aperture radar. This approach may enhance the signal-to- noise ratio (SNR) of the target by eliminating background noise. Further details of denoising and all steps of the method are provided below under the heading "‘Further Discussion.”

[0022] The method 100 may include reconstructing 115 an image from the target signals. For example, the image may be reconstructed by performing 118 a Fast Fourier Transform (FFT) on the t samples to obtain a range bin profile R (x,y, kxy), where k ∈ K is a range index in the range bin profile. At each x, y location, a target-related range index kxyis calculated 124 based on the distance 121 from the location to the target. A two-dimensional (2D) FFT is calculated for Phase compensation is performed by multiplying 127 the 2D FFT result bywhere z represents the distance from the target to the 2D scan plane, ωxand are the space frequencies of the 2D plane. A 2D inverse FFT (IFFT) is calculated 130 to obtain the reconstructed image r(x, y).

[0023] The method 100 may include segmenting 133 the reconstructed image to identify the wound based on moisture values of the target signals. A wound size of the target tissue may be calculated 136 based on moisture values determined from the target signals. An average moisture value of the wound may be determined 139. In some embodiments, the method includes transforming the reconstructed image to a moisture value image based on a physics- informed correlation model.

[0024] With reference to Figure 12, in another aspect, the present disclosure may be embodied as a system 10 for characterizing a wound 90. The system 10 includes a radar sensor 12 configured to transmit RF energy between 30 GHz and 300 GHz and receive reflected signals. A processor 14 is in electronic communication with the radar sensor 12. The processor is configured to perform any of the methods disclosed herein. For example, the processor may be configured to: cause the radar sensor to transmit a frequency -modulated continuous wave (FMCW) signal towards a target; receive a reflected signal through the radar sensor; repeat the steps of transmitting a FMCW signal and receiving a reflected signal for each location of a plurality of locations on the target; and identify a wound size of the target based on moisture values determined from the reflected signals.

[0025] In some embodiments, the system 10 further includes a stage 16 on which the radar sensor 12 is mounted. The stage is configured to translate the radar sensor (e.g., move the radar sensor to multiple locations, for example, in a 2D plane). In some embodiments, the system 10 includes a housing 18. The radar sensor 12 and the processor 20 may be contained within the housing 18 for portable (e.g., handheld, etc.) use.

[0026] The term processor is intended to be interpreted broadly. For example, in some embodiments, the processor includes one or more modules and / or components. Each module / component executed by the processor can be any combination of hardware-based module / component (e.g., graphics processing unit (GPU), a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), a digital signal processor (DSP)), software-based module (e.g., a module of computer code stored in the memory and / or in thedatabase, and / or executed at the processor), and / or a combination of hardware- and software- based modules. Each module / component executed by the processor is capable of performing one or more specific functions / operations as described herein. In some instances, the modules / components included and executed in the processor can be, for example, a process, application, virtual machine, artificial intelligence, neural network, and / or some other hardware or software module / component. The processor can be any suitable processor configured to run and / or execute those modules / components. The processor can be any suitable processing device configured to run and / or execute a set of instructions or code. For example, the processor can be a general-purpose processor, a central processing unit (CPU), an accelerated processing unit (APU), a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), a digital signal processor (DSP), graphics processing unit (GPU), microprocessor, controller, microcontroller, and / or the like.Further Discussion

[0027] Further non-limiting details and example embodiments of the present disclosure are provided in the following discussion.1. INTRODUCTION

[0028] Epidermal sensing is the use of sensors and electronic devices that are designed to monitor a wide range of physiological parameters on the human epidermal. These sensors can detect data such as body temperature, humidity, blood perfusion, sebum production, and even environmental variables like UV exposure and pollution levels. The data collected by epidermal sensors can be used for various applications, including healthcare monitoring, fitness tracking, disease diagnosis, sports performance analysis, and human-computer interaction. They are particularly useful in continuous health monitoring, allowing individuals and healthcare professionals to track vital signs and health metrics in real time.

[0029] The realm of epidermal sensing encompasses two primary methodologies: contact-based sensing (utilizing technologies like EEG, ultrasound, and photoacoustic methods) and non-contact-based sensing (employing devices such as thermal and NIR cameras). Contactbased approaches, while effective, have notable drawbacks. They can be discomforting for the patient and often entail energy restrictions, necessitating periodic replacements. Invasive sensors, a subset of contact-based sensors, require insertion into the epidermal, potentially causing painand increasing the risk of infection, particularly if not administered with care, or if the patient is sensitive. On the other hand, non-contact-based methods offer a more convenient solution, enabling epidermal sensing without direct physical contact. However, these sensors are sensitive to environmental factors like temperature and light conditions, leading to potential variations in accuracy and reliability. In scenarios involving open wounds, both contact and non-contact methods require the dressing or gauze to be opened for sensing, potentially causing harm to the wound site.

[0030] The radiofrequency (RF)-based sensor, as a non-contact method, is highly responsive to electromagnetic effects. The electromagnetic properties of human epidermal are intricately linked with its water content, causing alterations in the water content to impact RF signals differently. Moreover, water content significantly influences key epidermal factors like epidermal moisture and wound conditions. Consequently, RF measurements offer a versatile solution for epidermal sensing, overcoming the limitations associated with other methods.

[0031] The present disclosure provides a mmW ave-based system for epidermal sensing. mmWave systems are RF systems operating at frequencies of 30 GHz to 300 GHz, inclusive. In comparison to other RF-based techniques, a mmWave sensor offers several advantages. Firstly, it operates at high frequencies, providing millimeter-level resolution ideal for thin object sensing, such as human epidermal. Secondly, its compact size allows easy integration into medical devices. Thirdly, it operates on low power, posing no threat to human health. Lastly, Commercial Off-The-Shelf (COTS) mmWave radar is cost-effective, making it accessible for various scenarios.

[0032] Figure 1 shows the relationship between mmWave signal reflectivity and water content in the human epidermal. Embodiments of the present system utilize this correlation for multi-functional human epidermal sensing. To mitigate environmental interference, a synthetic aperture radar method may be employed, enabling high spatial resolution imaging, and allowing the target signal to be distinguished from environmental noise. Experimental results demonstrate the system’s effectiveness, achieving wound size estimation accuracy for about 1 cm2mean absolute area error. In basic scenarios, the system was shown to maintain around 5% absolute moisture estimation error within the 0-100% moisture range. In occluded scenarios, the error margin increases to about 10%.

[0033] The contributions are listed as follows:• Non-Contact Epidermal Moisture Sensing, even with wound dressing: Based on the insight that water content can influence the mmWave signal reflectivity, the presently- disclosed “mmSkin” technology provides the first wireless epidermal moisture sensing system that can be used in dressing conditions.• Multi-Functional Sensing: The mmWave signal exhibits strong correlations with several epidermal features, including spatial distribution and dielectricity. Various embodiments of the present disclosure include a multi-functional extraction algorithm designed to effectively extract epidermal features from the mmWave signal for wound size and epidermal moisture estimation.• RF-Epidermal Sensing software, including denoising and feature reconstruction: Various embodiments of the present disclosure provide an environmental-denoising mmWave imaging algorithm aimed at isolating the target signal from environmental interference. In contrast to traditional mmWave imaging techniques, the presently-disclosed algorithm utilizes a higher ratio of target-related signals for the target imaging, enhancing its robustness against environmental interference.• To rigorously assess the capabilities of the mmSkin system, a series of phantoms that emulate the intricacies of epidermal wounds and moisture variations were engineered. A comprehensive battery of experiments was conducted on these phantoms to ascertain the system’s accuracy in determining the wound size and moisture level, both in the presence and absence of occlusive materials. Subsequently, the evaluation was extended to actual epidermal, further validating the system’s proficiency in moisture level estimation.2. BACKGROUND2.1 Human Skin Model

[0034] Figure 2 offers a detailed depiction of the skin's layers, emphasizing components integral to the skin’s hydration and moisture retention. The varying shades of blue in the illustration symbolize the water content within the skin layers. A deeper shade of blue typically represents a higher concentration of water. Beginning with the outermost layer, the epidermis, essential components like the cuticle and ceramides stand out. The ceramides, in particular, play a pivotal role in water retention, preventing trans-epidermal water loss (TEWL) by forming aprotective barrier. Comeocytes, also present in the epidermis, help maintain the skin’s natural moisture by binding water to their surface. Deeper within the skin lies the dermis, which showcases the moisture-binding hyaluronic acid. Hyaluronic acid has an unparalleled ability to retain water, holding up to 1,000 times its weight in moisture. This ensures that the skin remains plump, hydrated, and youthful. The presence of proteins in the dermis, while not directly responsible for water retention, aids in maintaining the skin’s structural integrity, ensuring it can effectively hold and distribute moisture. Lastly, the blood vessels, represented by the red line, provide the necessary hydration and nutrients to the skin. They ensure a continuous supply of water and essential elements, supporting overall skin health and hydration. Finally, the third layer depicted in Figure 2 is the subcutaneous tissue, also known as the hypodermis or subcutis. This deepest layer of skin primarily comprises fat cells (adipocytes) and connective tissue. Its primary functions are to provide insulation, store energy, and cushion the body against trauma.

[0035] Moisture level within the human skin is a principal biomarker indicative of its health and condition. The hydration state of the epidermal, the outermost layer of the skin, is particularly informative as it affects the skin’s elasticity, pliability, and mechanical resistance. Adequate moisture not only imparts a healthy appearance but also serves as a barrier against environmental stressors, pathogens, and the permeation of substances. Conversely, diminished hydration can lead to various dermatological concerns, such as dryness, itchiness, scaling, and even the exacerbation of chronic skin conditions like eczema and psoriasis. Therefore, maintaining an optimal moisture balance on the human epidermal is essential for preserving the skin’s integrity and function.2.2 Epidermal Sensing Technologies2.2.1 Wound Size Sensing

[0036] When significant moisture changes occur in the human epidermal, it can precipitate wound formation. These changes result in a distinct spatial variation between the wound and healthy epidermal. By leveraging this spatial disparity, clinicians can estimate wound size more accurately. This estimation is crucial for assessing the wound’s severity and the effectiveness of the healing process, making moisture management a key factor in wound care. Traditional methods for wound size sensing include:

[0037] 1) Manual Measurement: Using tools such as rulers, calipers, or tape measures to determine the length, width, and sometimes depth of a wound. Although simple to operate, this approach is highly subjective and prone to inaccuracies. Also, the tool may contact the wound and be intrusive or painful for the patient.

[0038] 2) Photographic Method: Taking photographs of the human epidermal and using software to analyze and calculate the abnormal area. This method can be more consistent but depends on the quality of the images and is unable to penetrate the clothing.

[0039] 3) Advanced Imaging Methods: Techniques like Optical Coherence Tomography(OCT), Magnetic Resonance Imaging (MRI), or ultrasound can provide detailed cross-sectional images of the epidermal, allowing for precise volumetric and morphological analysis of abnormal areas. However, these technologies are often limited by their high cost, need for specialized operators, extended procedure times, and issues with accessibility and patient comfort.2.2.2 Moisture Level Sensing

[0040] Besides sensing the spatial features of the human epidermal system, moisture value sensing in the context of epidermal health is also essential for various diagnostic and therapeutic purposes. The traditional methods for assessing epidermal moisture are based on different principles:

[0041] 1) Comeometry: This method uses a comeometer device to measure the electrical capacitance of the human epidermal, which increases with higher water content. It is a commonly used technique for evaluating the hydration level of the human epidermal.

[0042] 2) Transepidermal Water Loss (TEWL): TEWL measurements gauge the amount of water that evaporates from the epidermal, indirectly providing data on epidermal barrier function and hydration. This method requires a controlled environment to minimize the influence of external factors like humidity and temperature.

[0043] 3) Impedance Measurement: By applying a small electrical current to the epidermal and measuring the resistance, this method infers moisture content since water conducts electricity better than dry epidermal tissue.

[0044] 4) Spectroscopic Methods: Techniques such as Raman spectroscopy can be used to assess hydration by analyzing the vibrational modes of water molecules in the epidermal.

[0045] These conventional methods, while useful, have their limitations such as the need for direct epidermal contact, potential variability due to environmental conditions, and the requirement for specialized equipment.

[0046] The present mmWave sensor technology stands out as it offers a non-contact solution capable of simultaneously sensing both wound size and moisture value. This approach eliminates the need for direct epidermal contact, thus minimizing discomfort and the risk of infection, especially in sensitive or healing areas. It works by emitting mmWave signals that reflect off the epidermal, the properties of these reflections can indicate changes in epidermal moisture due to the dielectric properties of water and can also capture the topography of the epidermal, allowing for accurate size measurements of abnormal moisture areas. This dual capability, along with its potential to penetrate through clothing and dressings, makes mmWave sensors particularly advantageous for comprehensive epidermal assessment in a variety of clinical and consumer health settings.3. MMSKIN SYSTEM DESIGN3.1 mmSkin Overview

[0047] As shown in Figure 3, the framework of the present mmSkin system may include four steps: Range-FFT, mmWave imaging, Image segmentation, and Multi-functional feature extraction. By using the present mmSkin system, the spatial feature and the dielectricity feature can be obtained for wound size and moisture value estimation, respectively. In the following section, we introduce the details of each step in the mmSkin system.3.2 Radio Frequency Selection for Epidermal Sensing

[0048] When analyzing the potential of different frequency ranges for epidermal sensing with wound dressing (e.g., gauze, bandage and garment), there are a spectrum of capabilities. Starting with microwave frequencies, which lie below mmWave, they offer a moderate resolution given their longer wavelengths, especially in the UHF and L-band regions. Their penetration abilities through the fabric and certain materials are reasonably good, but they might not capture the nuanced moisture changes with high precision. In contrast, the mmWavefrequencies, spanning from 30 GHz to 300 GHz, inherently possess high spatial resolution because of their shorter wavelengths. This allows them to adeptly detect minute variations in epidermal moisture. Their penetration is moderate, sufficient to traverse the fabric occlusion but with a focus on the human epidermal. Beyond mmWave, frequencies like sub-terahertz and terahertz have even shorter wavelengths, providing superb resolution. However, their primary limitation lies in penetration. These high frequencies tend to be absorbed more by materials and might not penetrate the fabric as effectively as mmWave signal, making them potentially less ideal for the occluded scenarios. In essence, mmWave frequencies strike a balance, making them a prime candidate for epidermal sensing under fabric occlusion conditions.3.3 Scanning Modality and Parameter Configuration

[0049] As shown in Figure 4, a mmWave sensor may be configured to transmit a Frequency-Modulated Continuous-Wave (FMCW) signal, and the frequency increases linearly over a cyclical period of time (i.e., a chirp). Generally, a transmitted mmWave signal at time t can be modeled as follows:where P0is the initial transmitted signal energy power, φtx(f) is the transmitted signal phase at time t. The received signal at time t can be written as follows:where φrx(t) is the received signal phase at time t. After the transmitted signal gets reflected by the target, the received signal will suffer from energy loss in the transmission path, which is denoted as G. The mmWave radar may mix the transmitted signal ftx(t) with the received signal frx(t) by multiplication to provide a mixed mmWave signal f (t) as follows:

[0050] The channel gain G is mainly determined by the refraction coefficient χ and transmission path loss α. In wireless human skin sensing, the radar emits a mmWave signal to the human skin surface and receives a reflected signal, the signal is transmitted in the air media all the time, thus transmission path loss α depends on the relative permittivity of air, which is very small and can be neglected, i.e.. α ≈ 1. Additionally, the refraction coefficient χ is themetric that measures the ratio of the mmWave signal being reflected by the boundary of two materials. According to the Fresnel Equations, the reflection coefficient is defined as χ = where the mmWave signal transmits from the first medium with the refractive indexto the second medium with the refractive index n2. In wireless human skin sensing, the first medium is the air with a refractive index n1≈ 1, and the second medium is the human skin, while the refractive index value n2is highly related to the water content. The relationship between the mmWave signal channel gain G and the reflection coefficient χ is described as follows:while nskinis the refractive index of the human skin. Substituting Equation 4 into Equation 3, we can get the following formula:

[0051] According to Equation 5, the amplitude of received mmWave signal f(t) is only related to the initial transmitted signal energy power P0and the refractive index of the human skin nskin. P0is a constant that can be acquired from the sensor manual, thus it is available to derive nskinfrom the received mmWave signal amplitude, which is the parameter that is highly related to the human epidermal moisture.

[0052] Due to the multi-path effect, the received mmWave signal is the combination of the reflected mmWave signal from the target and the environment. Some embodiments of the present disclosure use a Synthetic Aperture Radar (SAR) technique to split the mmWave signal. SAR techniques require the sensor to move and construct a much larger antenna aperture than the real size of the radar antenna As shown in Figure 5, the mmWave sensor is moved and data is collected at each location uniformly distributed in the 2D plane, dx and dy are the step distance in X axis and Y axis. Additionally, the mmWave sensor collects T samples at each location. Therefore, the mmWave signal can be modeled as f(x, y, t), where x and y represent the position of x • dx and y • dy in the X axis and Y axis, t represent the sample index in the T samples.3.4 Environmental-Denoise mmWave Imaging3.4.1 Environmental-Denoise

[0053] The mmWave signal collected at a location f (x, y, t) is the overlay of reflected signals from different objects in the environment. In this case, Fast Fourier Transform (FFT) may be performed on the sample dimension to transform the original signal f(x, y, t) into range bin profile R (x,y, k), which is named " Range-FFT.” Variant k ∈ K corresponds to the range index in the range bin profile. Range-FFT enables the mmWave sensor to focus on the specific distance of the target d, while d can be calculated to the range index k based on Equation 6. Slope is the rate at which the frequency changes with respect to time, fsis the sampling rate, c is the light speed and T is the number of samples.

[0054] Considering the movement of the mmWave sensor, the distance to the target d can be changed with x and y. As shown in Figure 6, compared to Location 1, the mmWave radar in Location 2 has Δd more distance. When performing Range-FFT, the range index for the two locations may be different, i.e.. However, the mmWave imaging algorithm using theSAR technology assumes the target to have the same distance to each mmWave sensor location. For example, in Figure 6(a), a traditional algorithm assumes Location 2 has the same distance to the target as Location 1, then chooses the same range index after Range-FFT. Due to the real distance of Location 2 being greater than the distance to Location 1, the traditional algorithm causes the mmWave radar to focus on the environment noise rather than the target in Location 2. In the presently-disclosed modified mmWave imaging algorithm, as shown in Figure 6(b), the target is selected based on the real distance from each location to the target, which causes the mmWave radar to maintain focus on the target throughout all locations. The transformed mmWave signal is obtained and denoted as while and1 / are the number of locations in X axis and Y axis.3.4.2 mm Wave Image Reconstruction

[0055] Next, a previous work gives a theory correlation between the mmW ave signal and the mmWave image, which is shown in Equation 7. Based on the equation, SAR imaging algorithm 1 may be performed to reconstruct the target.where FT2Daanndd are a two-dimensional (2D) Fourier Transform and Fourier InverseTransform, respectively; z' represents the distance from the target to the mmWave scanning plane; x and y are indexes in X and Y axes; r(x,y) is the reflectivity of the target at the location of (x · dx, y · dy, z'); and ωxand ωyare the space frequencies of the 2D image plane in Figure 5. As described in Equation 5, the reflectivity is highly related to the refractive index of the target.3.5 Multi-Function Feature Extraction3.5.1 Spatial Feature Extraction

[0056] By using the presently disclosed mmWave imaging algorithm, a 2D mmWave image may be reconstructed with both the target and surrounding environment information. To analyze the target, a segmentation model may be used to separate the target within the image. In a non-limiting example, a Segment Anything Model (SAM) is used in this example to crop the target from the mmWave image for its good performance on image segmentation. Specifically, when inputting a 2D image r (x, y) to a SAM, a masked imagewith the same size as r(x, y) will be obtained, while each pixel m(x, y) has only two values: 0 or 1. If the pixel (x, y) is not marked as the target area, then the pixel value m(x, y) = 0, otherwise m(x, y) = 1.3.5.2 Moisture Feature Extraction

[0057] According to Section 3.4, the mmWave image value at each location is subject to the corresponding target refractive index, thus the change in the image value indicates the change of the refractive index. In this section, the refractive index is further used to estimate the moisture value.

[0058] As shown in Equation 8, the refractive index n is the root of the relative permittivity ε:

[0059] In epidermal moisture sensing, the human epidermal can be regarded as a mixture, which is composed of water and another component. Based on a dielectric model of heterogeneous mixtures, the correlation between the Volume of Water Content (VWC) V and the relative permittivity of the human epidermal ε can be modeled in Equation 9 as follows:

[0060] While the VWC represents the volume ratio of the water to the target, it is also the metric of the moisture value. εwaterand εotherare the relative permittivity of the water and another component in the target, respectively. εwaterand εotherare the intrinsic property of the material, which should both be the constant. Therefore, the moisture value V is determined by the relative permittivity of the human epidermal ε. Based on Equations 5, 7, 8, and 9, there is a correlation between the mmWave image value and the moisture value.

[0061] As described in Section 3.5.1, the 2D mmWave image will be labeled with the target area. Next, the mean value of the segmented target area in the mmWave image may be calculated to estimate the moisture value of the target, which may be based on Equation 10 as follows:4. EXPERIMENTAL EMBODIMENT AND EVALUATION4.1 mmSkin System Integration4.1.1 2D scanning kit

[0062] As shown in Figure 7, a mmW ave sensor module was mounted on two mutually perpendicular linear motion guides (i.e., a 2D motion stage) and moved at a uniform speed in the 2D plane. In this way, the sensor scanned the target and to create a mmWave image through using the mmWave imaging algorithm described in Section 3.4. In addition, the starting time of the mmWave sensor and the 2D motion stage may be synchronized, to correspond the frame index of the mmWave sensor to its location based on the known moving speed and frame sampling rate. As shown in Figure 8, a 2D scanning kit was used to enable the mmWave sensor and the 2D motion stage to start working simultaneously. The 2D scanning kit components are described as follows:

[0063] (a) Trigger signal generator: A trigger signal generator ((RIGOL DG4062)) generated stable square waves as the trigger signal to hardware trigger the mmWave sensor to start data sampling and the 2D motion stage to move the mmWave sensor in 2D plane.

[0064] (b) Control signal generator: A control signal generator (ELEGOO UNO R3) stimulated the electrical signals as the control signal to guide the movement of the mmWave sensor on the 2D motion stage.

[0065] (c) Signal encoder: A signal encoder was composed of two parts, i.e., a digital stepper driver (DM542T) and a power supply (YHG X-360-12). The digital stepper driver encoded the raw control signal from the control signal generator to the more complex motion control signal with direction, speed, and movement duration information. The power supply provided the energy for the digital stepper driver to stimulate the complex signal.

[0066] (d) 2D motion stage: A 2D motion stage was composed of two parts, i.e., a motor and a 2D linear rail. The motor (BIPOLAR NEMA 17 STEPPER MOTOR) converted the encoded electrical motion signal into mechanical energy, which was used to precisely drive the movement of the 2D linear rail (THOMSON LINEAR MOTION SYSTEM).

[0067] (e) mmWave sensor module: A mmWave sensor module was composed of two parts, a mmWave sensor (TI AWR1642BOOST) and an evaluation board (DCA1000EVM). A transmitter antenna in the mmWave sensor emitted a Frequency-Modulated Continuous Wave (FMCW), the FMCW signal was reflected by the target and surrounding environment, the receive antenna in the mmWave radar received the reflected signal and the mmWave sensor forwarded the data to a processor (e.g., a computer) through the DCA1000EVM board. Its sensing range was from 5 cm to 10m.

[0068] (f) PC: The computer (Lenovo ThinkPad P15 Laptop with I7-10750H 2.6GHz processor) received and stored the collected data from the mmWave sensor for further processing.

[0069] By using the 2D scanning kit, the synchronization error between the mmWave radar and the 2D motion stage can be within 1μs, depending on the crystal oscillation frequency of the trigger signal generator (i.e., 16MHz in the experimental setup), which is very small and makes the collected data points have negligible location deviation during scanning.4.1.2 System Parameter Setup

[0070] In the experimental embodiment, the 2D motion stage was controlled to move the mmWave sensor at uniform speed v = 4 mm / s along the X-axis linear motion guide and v = 6 mm / s along the Y-axis linear motion guide. Additionally, only one Tx and Rx were used in the mmWave sensor for data collection. Each FMCW frame period was set to 0.07s, and only has one chirp, the duty cycle of the chirp in the whole frame period was less than 0. 1%, thus the chirp duration was short and the mmWave sensor could be regarded as static at its location during the data collection. In the following experiments, the 2D mmWave scanning array hadand location points on the X axis and Y axis, the step in X axis dx = 0.49 mmand the step in Y axis dy = 7.2 mm, thus the 2D scanning range was196 mm and The distance of the target to the scanning planewas set as 0.2 m. In one chirp, the working frequency of the mmWave sensor was set from 77 GHz to 81 GHz, the bandwidth was 4 GHz, and the slope of the frequency change B was 46.493 MHz / μs, while the sensor uniformly sampled T = 512 times during the chirp period.4.2 Evaluation Plan and Benchmark4.2.1 Wound Phantom Fabrication

[0071] To evaluate the performance of the mmSkin system on wound size and moisture value estimation, two phantoms were designed, respectively. As shown in Figure 9, for the wound size estimation, a mixture of 96% by mass of water and 4% by mass of agar gel was used as the wound size phantom. Based on the human skin water content model in Section 2.1, the mixture modeled the high water content characteristic of human skin. During production, the mixture was put into different molds with different shapes and depths, in order to create different wound size phantoms. As shown in Figure 10, for the moisture value estimation, a sponge was used as the testing phantom. The sponge had the advantage of flexible water content adjustment and uniform water distribution. Thus, it was an useful phantom to use in evaluating system performance with the moisture value changes.4.2.2 In vitro Test Benchmark

[0072] Pork meat is commonly utilized as an in vitro sample to approximate human skin in various biomedical studies, especially in epidermal moisture estimation. The choice of porkmeat is anchored in its structural and compositional similarities to human skin. Both human skin and pork skin share comparable thickness, texture, collagen content, and fat distribution.Furthermore, the hydration properties and water-holding capacities of pork are close parallels to those of human skin. These analogies make pork a viable and easily accessible medium for researchers to conduct preliminary tests and calibrations without the ethical constraints associated with human testing. To test the performance of the mmSkin system, as shown in Figure 11, pork meat was used to conduct the in vitro experiments for the epidermal moisture estimation of the mmSkin system.4.3 Performance Metrics

[0073] Area error (AE): The area error was used to estimate the discrepancy between the measured size of a wound, as determined by the mmSkin system, and its actual size. As shown in Equation 11, Areaactualis the actual area value of the wound, dx × dy is the actual area value of a pixel in the mmWave image, and is the number of pixels thatbelong to the wound. Thus, is the estimated wound area value inthe mmWave image.

[0074] Mean absolute error (MAE): The mean absolute error is a widely used statistical measure that quantifies the accuracy of predictions in modeling or forecasting. Specifically, MAE calculates the average magnitude of errors between predicted and observed values, without considering the direction of the errors. It is computed by taking the absolute differences between predicted and actual values, summing all these absolute differences, and then dividing by the number of predictions made. MAE is anon-negative value, which indicated the accuracy of the mmSkin system on the moisture value estimation.5. WOUND ASSESSMENT FUNCTIONS5.1 Wound Size Estimation

[0075] In the fundamental assessment of the mmSkin system’s capability for wound size estimation, wound phantoms of varying sizes and depths were employed as depicted in Figure 9.These phantoms served as controlled models to closely mimic actual wounds in terms of area and volume, facilitating the evaluation of the system’s accuracy. Four wound phantoms with different area sizes and depths were imaged and the area value described in Section 3.5.1 was calculated, two with an area of 28.274 cm2and two with 15.904 cm2, but with differing depths of 2 cm, 1.5 cm, and 1 cm respectively. The estimated sizes derived from the mmWave images were then compared against the actual sizes of the phantoms. Table 1 shows that the present system showcased a high degree of accuracy, with size estimation errors within approximately 1 cm2. This margin of error is quite minimal, indicating that the mmSkin system is robust in assessing wound dimensions, which is an important aspect in medical diagnostics and treatment planning.Table 1: mmSkin system wound size estimation accuracy in the basic scenario.5.2 Moisture Estimation

[0076] The analysis of the mmSkin system’s capability for moisture level estimation revealed a sophisticated understanding of the relationship between mmWave signal reflectivity and the subject’s moisture content. The experimental design involved the use of a sponge phantom as the subject, chosen for its high water absorption properties. The experiments were conducted to capture a comprehensive data set across the entire spectrum of possible moisture values, from completely dry (0%) to highly saturated (80%). A total of 17 evenly distributed moisture value samples were generated by systematically wetting the sponge and allowing for a 5 -min uniform distribution of the added water, which was accurately measured using a digital scale with a 0.1g resolution.

[0077] It was observed from the collected data that the mmWave signal reflectivity was in a strong positive linear correlation with the moisture level of the sponge, thus a linear correlation was constructed to predict the moisture level based on the mmWave signalreflectivity. The collected 17 samples were split into two groups: one for establishing the linear correlation and the other for validating the system’s predictive accuracy. The process of splitting and validation was repeated ten times to remove the randomness. As shown in Table 3, the mmSkin system demonstrated high predictive accuracy, with a mean absolute error of just 5.6% when estimating moisture values without the influence of any occlusive materials, across the full range of 0 to 80%. This accuracy indicates that the system is highly reliable for assessing epidermal moisture values, which can be an important measure in both medical diagnostics and skincare regimes.6. EXPERIMENTS IN REAL-WORLD SETTINGS6.1 Wound Assessment with Fabric Occlusions

[0078] Occluded Wound Size Estimation: The mmSkin system’s prowess in wound size estimation was evaluated under various conditions, with a focus on scenarios involving fabric occlusions. An experimental embodiment of the system was tested by placing different types of fabric over the wound phantoms, simulating real-world circumstances where skin maybe covered by clothing.Table 2: mmSkin system wound size estimation accuracy under the fabric occlusion.Table 3: mmSkin system moisture value estimation accuracy under with and without fabric occlusion (Moisture value ranges from 0% to 100%).

[0079] The examination was twofold: it assessed the impact of both the thickness of the fabric layers and the nature of the fabric material. In the realm of textiles, cotton represented natural fibers, while polyester signified synthetic options. Hence, for a comprehensive analysis, the system’s accuracy was measured against occlusions made of a single layer of cotton gauze, a double layer of the same, and a single layer of polyester cloth.

[0080] The results, as documented in Table 2, are telling. The presence of fabric, whether it be the single-layered natural cotton gauze that yielded an impressively minimal area error of 0.072 cm2, the slightly more intrusive double-layered cotton with an area error of 0.649 cm2, or the synthetic polyester which resulted in a 0.265 cm2error, does not significantly detract from the mmSkin system’s ability to accurately estimate wound size. This resilience to various types and depths of occlusion demonstrates the system’s robustness and its potential applicability in a real-world, non-invasive skin assessment context, maintaining an error margin under 1 cm2. This remarkable performance holds promise for the system’s use in clinical settings, where it could reliably function despite the inevitable presence of clothing or bandages over wounds.

[0081] Occluded Moisture Estimation: The evaluation of the mmSkin system’s performance in estimating moisture levels under varying conditions of fabric occlusion offers insightful findings. By introducing occlusions atop the phantom used in the mmWave data acquisition process and implementing a thin waterproof membrane to prevent moisture transfer to the occlusion material, we were able to maintain the integrity of our experiment. For each type of fabric occlusion, we gathered between 10 to 15 individual data samples to solidify our statistical analysis. Aligning with the methodology detailed in Section 5.2, we segregated these samples into two distinct sets. One set was utilized to establish a linear correlation, while the other was employed to validate the predictive accuracy of moisture values based on mmWave signal reflectivity.

[0082] The results presented in Table 3 are indicative of the system’s robustness, albeit with diminished performance when fabric occlusion is present. It is noteworthy that the mean absolute error in moisture value prediction increased from 5.6% without occlusion to 9.9% with a single layer of cotton, and further to 13% when a second layer was introduced. The single layer of polyester resulted in a 12% mean absolute error, showcasing a slight advantage over the two- layer cotton setup.

[0083] Despite these increases in MAE metrics, the system maintained an error rate below the predefined threshold of acceptability. This demonstrates that the mmSkin system retains a considerable degree of accuracy and reliability in moisture estimation, even when challenged with fabric occlusions that simulate real-world scenarios where skin might be covered by clothing.6.2 In vitro Moisture Estimation

[0084] When quantifying moisture content on the real epidermal, we encounter a challenge: water evaporates quickly, leading to fluctuating moisture values that defy precise control during imaging. To circumvent this issue, we cannot employ water to adjust the moisture value of the epidermal as we do with inanimate objects. Instead, we apply an ultrasound coupling agent (Aquasonic Clear Ultrasound Gel) to the epidermal. This substance, predominantly waterbased and non-volatile, ensures a stable moisture value throughout the data collection.

[0085] To accurately evaluate the moisture values, we harness the capabilities of Shortwave Infrared (SWIR) light. SWIR light, spanning the 0.9 - 1.7 μm wavelength range, has a limited penetration capacity but is notably absorbed by water molecules. Consequently, the epidermal with a higher moisture content will absorb more SWIR light, resulting in a decrease in reflected light intensity and, thus, a lower pixel value on the SWIR camera image. The SWIR camera, esteemed for its high resolution, proves to be an excellent instrument for the precise estimation of epidermal moisture values.

[0086] The WIDy SenS 640 SWIR camera is utilized to image three pork meat surfaces (shown in Figure 11) across a spectrum of moisture values. Post image acquisition, we apply the SAM model in Section 3.5 to discern the meat area within the SWIR images and calculate the mean pixel value. This pixel mean value becomes a pivotal reference: we adjust the water content in a sponge until the SWIR image of the sponge mirrors the mean pixel value of the meat surface. This calibrated sponge moisture level is then assumed to be equivalent to that of the pork meat surface and is employed as the ground truth for subsequent mmWave sensor moisture estimations.Table 4: mmSkin system moisture value estimation accuracy on the pork meat (Moisture value ranges from 0% to 100%).

[0087] For each pork meat sample, we generated 10 separate instances with a consistent distribution of moisture values. We then divided these instances into two distinct sets: one for establishing a linear correlation (the fitting set) and the other for verifying this correlation (the testing set). Table 4 demonstrates that the present system sustained a mean absolute error margin of approximately 5% in its moisture estimations, providing empirical evidence of the mmSkin system’s efficacy when applied to realistic skin scenarios. This highlights the system’s potential utility for non-invasive epidermal moisture assessments in medical and cosmetic applications.7. DISCUSSION7.1 mmSkin Scanning Speedup

[0088] By modifying the translator stage type and substantially increasing its movement speed from 7 mm / s to 50 mm / s, the scanning duration can be notably reduced. This substantial speed enhancement alone would have streamlined the process considerably. Moreover, the introduction of MMWCAS-RF-EVM board, equipped with an advanced 82 virtual antenna array, has augmented the scanning efficiency even further. By integrating these advancements, reducing the scanning time from a lengthy 14 minutes down to one minute or less represents a monumental stride in efficiency. Such improvements not only optimize the scanning process but can also greatly enhance user experience, operational throughput, and the overall efficacy of applications where time is a critical factor.7.2 mmWave Image Resolution Improvement

[0089] The mmSkin system’s capability to image objects is a foundational achievement, but understanding the system’s image resolution is paramount for real -world applications.Resolution can make a significant difference in the accuracy and clarity of the images produced, impacting the system’s overall performance and applicability. By planning extensive experiments to evaluate the resolution, we can get a comprehensive understanding of the system’s strengths and potential limitations. Furthermore, investigating the various factors thatcan influence the resolution will offer insights into optimizing the system for specific applications. By tailoring the system parameters based on these findings, the mmSkin system can be more effectively customized for different use cases, ensuring both efficiency and precision.8. RELATED WORK8.1 Moisture Measurement via RF Sensor

[0090] In the field of moisture measurement, RF sensors are differentiated primarily by their interaction with the target material: contact and non-contact types. Contact RF moisture sensors interact directly with the material being measured. These include Time-Domain Reflectometry (TDR), Frequency Domain Reflectometry (FDR), and Capacitance sensors, among others. TDR sensors measure the time delay of a reflected signal to determine moisture content, providing high accuracy but requiring physical insertion into the material, which may be disruptive. FDR sensors analyze the frequency changes in an electromagnetic wave when interacting with different moisture values and offer the benefits of rapid measurement cycles and ease of use. However, they may be less accurate in materials with varying densities or compositions. Capacitance sensors, another type, function by measuring the change in capacitance caused by the dielectric constant variations in the material due to moisture. They are generally less expensive and can be used for a wide range of materials but can be affected by temperature and may require calibration for specific applications.

[0091] Non-contact RF moisture sensors measure moisture without physical interaction with the material. This category includes microwave, radar-based systems and Near-Infrared (NIR) reflectometry. Microwave sensors operate on the principle of the dielectric constant, capitalizing on water’s significant dielectric disparity with most materials, which enables deep penetration and robust performance in diverse environments, though they can be cumbersome and sensitive to material composition. NIR sensors, conversely, offer rapid surface moisture assessments through selective NIR light absorption by water molecules, albeit limited to surfacelevel detection and susceptible to surface characteristics interference. Additionally, the poor penetration ability of the NIR light limits its availability in various applications. Radar-based systems offer non-invasive moisture detection through electromagnetic wave reflection analysis. Within this category, mmWave sensors emerge as a superior choice, especially in the 30 GHz to 300 GHz spectrum, combining the penetration strength of microwave systems with the precision of NIR technology. The compact nature and enhanced resolution of mmWave sensors allow foraccurate moisture content mapping, providing a sophisticated and versatile tool for a multitude of industrial applications where precise moisture detection is crucial.8.2 mmWave Imaging Technologies

[0092] Millimeter-wave (mmWave) imaging capitalizes on the short wavelengths within the millimeter spectrum, facilitating high-resolution imaging suitable for diverse applications. Multiple Input Multiple Output (MIMO) imaging leverages antenna arrays for both transmission and reception, creating detailed images that can differentiate between various targets and attributes concurrently. Meanwhile, beamforming, a sophisticated signal processing method employed in antenna arrays, concentrates the antenna’s energy in targeted directions, thereby enhancing signal strength and spatial resolution. Contrasting with these methods, Synthetic Aperture Radar (SAR) imaging constructs expansive, highly detailed images by moving a solitary antenna across a predetermined trajectory, proving instrumental in extensive radar systems. Unlike MIMO and beamforming, whose imaging resolutions are tightly linked to the antenna array’s dimensions, SAR circumvents this limitation, enabling high-resolution imaging without necessitating larger, more costly sensors. This is a pivotal reason our mmSkin system adopts SAR technology, aiming to deliver precision without increasing the system’s size and cost.

[0093] Although the present disclosure has been described with respect to one or more particular embodiments, it will be understood that other embodiments of the present disclosure may be made without departing from the spirit and scope of the present disclosure.

Claims

What is claimed is:

1. A method for characterizing a wound, comprising: transmitting a mmWave signal to a target tissue; receiving a reflected signal from the target tissue, the reflected signal corresponding to the mmWave signal and including a target signal; repeating the transmitting and receiving steps at multiple x, y locations in a 2D plane to scan the target tissue and at multiple times t at each location; calculating a wound size of the target based on moisture values determined from the target signals.

2. The method of claim 1, further comprising denoising the reflected signal to obtain the target signal.

3. The method of claim 2, wherein denoising the reflected signal includes splitting the reflected signal into the target signal and an environment signal using synthetic aperture radar.

4. The method of claim 1, further comprising reconstructing an image from the target signals.

5. The method of claim 4, wherein reconstructing the image from the target signals comprises: performing a Fourier Transform (FFT) on the t samples to obtain a range bin profileR (x, y, k), where k ∈ K is a range index in the range bin profile; for each x, y location: calculate a distance from each location to the target; and calculate a target-related range index kxy; calculate a 2D FFT onperform phase compensation by multiplying the 2D FFT result by wherez' represents a distance from the target tissue to the 2D plane, ωxand ωyare space frequencies of the 2D plane; and calculate a 2D inverse FFT to obtain the reconstructed image r(x,y).

6. The method of claim 4, further comprising transferring the reconstructed image to a moisture value image based on a physics-informed correlation model.

7. The method of claim 6, further comprising segmenting the moisture value image to identify the wound based on the moisture difference between the wounded and healthy area in the target tissue.

8. The method of claim 1, further comprising determining an average moisture value of the target.

9. A system for characterizing a wound, comprising: a radar sensor configured to transmit RF energy between 30 GHz and 300 GHz and receive reflected signals; a processor in electronic communication with the radar sensor, the processor programmed to: cause the radar sensor to transmit a frequency-modulated continuous wave (FMCW) signal towards a target; receive a reflected signal through the radar sensor; repeat the steps of transmitting a FMCW signal and receiving a reflected signal for each location of a plurality of locations on the target; and identify a wound size of the target based on moisture values determined from the reflected signals.

10. The system of claim 9, further comprising a stage on which the radar sensor is mounted, the stage configured to translate the radar sensor.

11. The system of claim 10, wherein the stage is configured to translate the radar sensor in two dimensions.

12. The system of claim 9, wherein the radar sensor comprises an array of antennas, each antenna in the array is positioned at a unique 2D location on the sensor board plane.

13. The system of claim 9, wherein the processor is further programmed to combine the reflected signal received at each location to produce a synthetic antenna aperture.

14. The system of claim 9, wherein the processor is further programmed to denoise the reflected signals.

15. The system of claim 9, wherein the processor is further programmed to reconstruct an mmWave image of the target.

16. The system of claim 15, wherein the processor is further programmed to segment the reconstructed image to identify a wound of the target.

17. The system of claim 9, further comprising determining an epidermal moisture level of the target.

18. The system of claim 8, further comprising a housing wherein the radar sensor and processor are contained within the housing for handheld use.

Images