Acousto-mechanical sensors and their applications

Abstract

In one aspect, the present invention relates to an acousto-mechanical sensor comprising a plurality of electronic components including at least one inertial measurement unit (IMU) for operatively detecting vibration and motion signatures of a physiological process, a microcontroller unit coupled to the at least one IMU for receiving data from and processing the received data, and a wireless communication system for wireless data transmission. The novel acousto-mechanical sensor has a wide range of applications ranging from evaluating the effectiveness of drugs for pruritus-causing conditions to monitoring disease severity and treatment response.

Description

[Technical field]

[0001] STATEMENT OF RIGHTS UNDER FEDERALLY FUNDED RESEARCH This invention was made with government support under Grant No. 1U01FD007001-01 awarded by the U.S. Food and Drug Administration. The Government has certain rights in the invention.

[0002] CROSS-REFERENCE TO RELATED PATENT APPLICATIONS This application claims priority to and the benefit of U.S. Provisional Application No. 63 / 171,483, filed April 6, 2021, which is incorporated herein by reference.

[0003] The present invention relates generally to biosensors, and more specifically to a diagnostic and therapeutic scratch sensor with skin conformable, wireless, dual wearable features enabled by machine learning algorithms to assess specific signatures of scratching, methods for quantifying symptoms of pruritus, and their applications. [Background technology]

[0004] The background description given herein is intended to generally present the context of the present invention. The subject matter described in the Background of the Invention section should not be assumed to be prior art merely as a result of the prior art being mentioned in the Background of the Invention section. Similarly, the problems mentioned in the Background of the Invention section or associated with the subject matter of the Background of the Invention section should not be assumed to have been already recognized in the prior art. The subject matter in the Background of the Invention section merely represents different approaches and may be inventions in themselves. To the extent described in the Background of the Invention section, the work of the currently named inventors, as well as aspects of the description that may not otherwise qualify as prior art at the time of filing, are not admitted expressly or implicitly as prior art against the present invention.

[0005] Pruritus and pain are the two primary examples of nociception in humans, with pruritus eliciting a scratch reflex and pain eliciting a withdrawal reflex. Unfortunately, pruritus is caused by a wide variety of conditions, and despite 1% of all annual outpatient visits being accompanied by pruritus symptoms, pruritus is often overlooked and undertreated in clinical practice, resulting in a significant global disease burden. Although a wide variety of conditions can cause pruritus (e.g., renal failure, liver failure, lymphoma, etc.), atopic dermatitis (AD) is likely the most common. AD is also the most widespread pediatric inflammatory skin disease, affecting 10 million US children with an annual prevalence of 13%. The hallmark of AD is pruritus, which is the primary driver of morbidity leading to chronic sleep disorders in ~60% of affected children. Consequences for millions of children include neurocognitive dysfunction and reduced growth. Quality of life for patients with moderate to severe AD consistently scores the lowest of all chronic diseases.

[0006] In AD and other conditions that cause pruritus, a major unmet need is for objective measures of pruritus to assess the efficacy of new drugs, quantify disease severity, and monitor treatment response. Quantifying pruritus presents challenges due to its subjective nature. Commonly used methods, such as patient-reported visual analog scales (VAS), correlate poorly with visually observed scratching behavior. Such self-report measures are prone to both sensitivity and retention bias. Furthermore, these subjective methods lack validity in young children and adults with cognitive impairment.

[0007] Measuring behaviors associated with scratching provides one way to quantify the severity and frequency of pruritus. Direct visual inspection of scratching on video recordings is the gold standard, but such a scheme is laborious and impractical in clinical practice. Wrist kinetics is a method of recording movement using an accelerometer to estimate scratching by measuring wrist motion. This approach is less than optimal in accuracy, as it cannot distinguish, for example, hand shaking from scratching. Additional aspects of the various characteristics of scratching, including inter-individual variability, present additional unsolved challenges. Thus, current sensors and strategies do not provide sufficient performance or practicality for routine use in clinical trials or patient treatment.

[0008] Thus, a heretofore unmet need exists in the art to address the aforementioned deficiencies and inadequacies. Summary of the Invention [Problem to be solved by the invention]

[0009] One objective of the present invention is to disclose a soft, flexible, wireless sensor capable of capturing the acoustic-mechanical signature of scratching and a method for objectively quantifying the symptoms of pruritus. [Means for solving the problem]

[0010] In one aspect, the present invention relates to an advanced acousto-mechanical (ADAM) sensor, which comprises a plurality of electronic components, including at least one inertial measurement unit (IMU) for operatively detecting vibration and motion signatures of a physiological process, a microcontroller unit (MCU) coupled to the at least one IMU for receiving data from the at least one IMU and processing the received data, and a wireless communication system for wireless data transmission. The acousto-mechanical sensor is configured to measure both low frequency signals below about 100 Hz and high frequency signals above about 100 Hz derived from the at least one IMU. Further, the acousto-mechanical sensor is skin-mounted to enable coupling of the at least one IMU to the skin to capture both high frequency and low frequency signals. The high frequency and low frequency signals represent a physiological response to itch as a scratching or rubbing activity. In one embodiment, the physiological process includes a scratching activity.

[0011] In one embodiment, the plurality of electronic components and the at least one IMU are spatially separated such that the plurality of electronic components are at one end of the ADAM sensor and the at least one IMU is at an opposite end of the ADAM sensor.

[0012] In one embodiment, the ADAM sensor further comprises a flexible printed circuit board (fPCB) for supporting and interconnecting at least one IMU and multiple electronic components.

[0013] In one embodiment, the fPCB employs a patterned layout that includes a plurality of freely deformable, serpentine interconnects designed to mechanically decouple at least one IMU from other components of the ADAM sensor.

[0014] In one embodiment, the plurality of freely deformable, serpentine interconnects are flexible and stretchable and adapted to interconnect at least one IMU and a plurality of electronic components.

[0015] In one embodiment, the ADAM sensor also includes at least one IMU, a plurality of electronic components, and an elastomeric encapsulation layer that at least partially surrounds the fPCB and forms a tissue-facing surface that is operably attached to a living body.

[0016] In one embodiment, the tissue-facing surface of the elastomeric encapsulating layer is configured to conformally match the skin surface of a living body.

[0017] In one embodiment, the encapsulating layer is formed from a thin, low modulus silicone elastomer.

[0018] In one embodiment, the elastomeric encapsulating layer is a biocompatible silicone enclosure.

[0019] In one embodiment, the ADAM sensor further comprises a biocompatible hydrogel adhesive for attaching the ADAM sensor to an anatomical location in a living body.

[0020] In one embodiment, the biocompatible hydrogel adhesive is a medical grade, hypoallergenic adhesive.

[0021] In one embodiment, the biocompatible hydrogel adhesive is adapted to operatively transmit signals from the living body to at least one IMU.

[0022] In one embodiment, the wireless communication system comprises a Bluetooth low energy (BLE) radio.

[0023] In one embodiment, the wireless communication system is configured to transmit the output signal from the ADAM sensor to an external device.

[0024] In one embodiment, the plurality of electronic components further comprises a power module coupled to the at least one IMU for providing power, the wireless communication system, and the MCU.

[0025] In one embodiment, the power module comprises a battery.

[0026] In one embodiment, the battery is a rechargeable battery.

[0027] In one embodiment, the power module further comprises a wireless charging module for wirelessly charging the rechargeable battery.

[0028] In one embodiment, the power module further comprises fault protection elements including short circuit protection components or circuits to prevent malfunction of the battery.

[0029] In one embodiment, the at least one IMU comprises a millimeter-scale three-axis accelerometer configured to detect skin-conducted vibrations and movements associated with scratching activity.

[0030] In one embodiment, the three-axis accelerometer is configured to detect acceleration along the x-axis and y-axis at a first sample rate and acceleration along the z-axis at a second sample rate, where the x-axis and y-axis are oriented parallel to the skin surface on which the ADAM sensor is attached and the z-axis is perpendicular to the skin surface.

[0031] In one embodiment, the first sample rate is in the range of 160-240 Hz and the second sample rate is in the range of 1,200-2,000 Hz.

[0032] In one embodiment, the three-axis accelerometer is configured to have a sampling rate of 1,600 Hz with 16-bit resolution and a dynamic range of ±2 g, where g is the acceleration due to gravity, 9.8 m / s 2 It is.

[0033] In one embodiment, when placed on the dorsum of the hand, the ADAM sensor can detect acousto-mechanical signals associated with scratching activity through a combination of movement and acousto-mechanical signals in a manner that is not affected by ambient noise.

[0034] In one embodiment, when positioned between the metacarpals of the second and third fingers, the ADAM sensor can quantify scratching activity initiated not only from wrist and / or arm movements, but also from the fingers and fingertips, over a wide temporal bandwidth that includes the high frequency vibrational motion associated with scratching itself.

[0035] In one embodiment, scratching activity is characterized by first and second types of signals, the first type of signals corresponding to global movements of the hand and the second type of signals overlapping in time with the first type of signals and resulting from subtle vibration impulses generated by the movement of the fingertip and fingernail against the contact surface.

[0036] In one embodiment, the first type of signal has a characteristic frequency in the range of a few Hz or less, and the second type of signal has a characteristic frequency ranging over several hundred Hz, with the amplitude decaying rapidly with position along the fingers, into the hand, and finally through the wrist and up the arm.

[0037] In one embodiment, the first and second types of signals are distinguished by passing the data through a low pass filter and a high pass filter with a cutoff of 2 Hz.

[0038] In one embodiment, a signature of acousto-mechanical signal features that uniquely characterize scratching activity is extracted from the output signal of the ADAM sensor to train and validate a machine learning based algorithm for scratch detection and to quantify the pruritus symptom.

[0039] In one embodiment, the ADAM sensor is unloaded, stretchable, twistable, and / or bendable.

[0040] In one embodiment, the ADAM sensor has a specific geometric polarity for attachment to an anatomical location in the body, and is flexible and conformable to the skin.

[0041] In one embodiment, a scratching algorithm is embedded on the sensor such that the sensor can process detected data from at least one IMU to obtain signal features that uniquely characterize scratching activity.

[0042] In one embodiment, the ADAM sensor is configured to notify the user and / or professional via audio / visual means on an acousto-mechanical sensor, a mobile device, and / or a smart wearable device. These notifications can be tunable to be strong enough to generate attention to the scratching by the user, or gentler to stop the scratching during sleep but not wake the wearer.

[0043] In one embodiment, the sensor's conformability to the hand allows it to be placed over key bones -- for example, the first and second metatarsals -- which are key to capturing the high frequency vibrations characteristic of scratching.

[0044] In one embodiment, high frequency signal capture enabled by a high frequency IMU allows for the capture of both scratch-specific motion and bone-conducted high frequency vibrations, which allows for the differentiation of scratching motion from true scratches on the skin.

[0045] In one embodiment, the low-profile conformable nature of the sensor allows the acousto-mechanical sensor to be placed on the curved surface of the dorsum of the hand, which allows for capturing scratches regardless of joint articulation, which may be associated with finger movement only, wrist movement only, elbow movement only, shoulder movement only, or a combination thereof.

[0046] In one embodiment, the ADAM sensor is operatively positioned on the back of the hand, fingernail, finger, or wrist to measure scratching movements over time.

[0047] In one embodiment, the ADAM sensor can be used in any disease state where itch is a symptom, as a diagnosis, a therapeutic response indicator, a clinical trial endpoint, or an early warning of disease progression.

[0048] In another aspect, the present invention relates to a method for quantifying a symptom of pruritus, comprising attaching an Advanced Acousto-Mechanical (ADAM) sensor to an anatomical location of a living body, detecting signals associated with scratching and / or non-scratching activity with the ADAM sensor, and analyzing the signals to assess an acousto-mechanical signature of scratching to quantify the symptom of pruritus.

[0049] In one embodiment, analyzing the signal includes preprocessing the signal to remove low-frequency baseline fluctuations caused by slow large-scale motion so that the preprocessed signal contains more specific high-frequency signal components associated with z-axis acceleration, segmenting the preprocessed signal into frames by sliding overlapping n-second windows, extracting a set of features from each frame, and applying a Random Forest (RF) classifier to the extracted set of features for prediction and validation. In one embodiment, the set of features includes x / y / z-axis acceleration fluctuations, high-frequency spectrograms, and / or low-frequency spectrograms.

[0050] In one embodiment, the n second window is a 1 second window.

[0051] In one embodiment, the overlap is in the range of about 50% to 90% overlap, preferably about 50%.

[0052] In one embodiment, analyzing the signals further includes assigning each frame a label of 0, indicating non-scratching activity, if less than about 50% of the data points in a given frame are labeled as scratching activity, and otherwise assigning the frame a label of 1, indicating scratching activity.

[0053] In one embodiment, the RF classifier is trained with a training dataset that includes scratch data from multiple body locations and various types of non-scratch activities collected from a large number of healthy subjects.

[0054] In one embodiment, the RF classifier is optimized by comparing results from leave-one-out cross-validation (LOSO-CV) applied to a training dataset, the results of which form the basis for optimizing the RF classifier.

[0055] In one embodiment, the parameters for optimization include the number of decision trees in the RF classifier and the minimum required scratch (MRS) period.

[0056] In one embodiment, the number of decision trees is in the range of about 10 to 200, preferably about 30.

[0057] In one embodiment, for optimization, the RF classifier returns predictions at the 1-second frame level, and the output is then clustered using Density-Based Spatial Clustering for Noisy Applications (DBSCAN), which concatenates two closely spaced scratch frames as a single scratch event, with the MRS duration defining the minimum number of frames to form a cluster of scratch events.

[0058] In one embodiment, the MRS duration is in the range of about 1.5 seconds to 8.5 seconds, and preferably is at least about 4.5 seconds.

[0059] In one embodiment, the scratching and / or non-scratching activities include scratching the living body's head, arms, stomach, knees, and / or legs, and / or simulating scratching in the air, waving hands, entering text into a phone, typing on a keyboard, and / or clicking a mouse.

[0060] In one embodiment, the anatomical locations are the fingertip, the finger, the dorsum of the hand, and the wrist of a living subject.

[0061] Yet another aspect of the present invention relates to a tangible, non-transitory computer-readable medium storing instructions that, when executed by one or more processors, cause quantifying symptoms of pruritus in the manner disclosed above.

[0062] These and other aspects of the present invention will become apparent from the following description of the preferred embodiments taken in conjunction with the following drawings, in which variations and modifications described can be effected without departing from the spirit and scope of the novel concepts of the present disclosure.

[0063] The accompanying drawings illustrate one or more embodiments of the invention and, together with the written description, serve to explain the principles of the invention. Wherever possible, the same reference numbers are used throughout the drawings to refer to the same or like elements of an embodiment. [Brief description of the drawings]

[0064] [Figure 1A] FIG. 1 shows an overview of an ADAM sensor and signal output according to an embodiment of the present invention, accompanied by photographs of the ADAM sensor in various deformation states highlighting its soft, flexible structure. [Figure 1B]FIG. 13 shows an overview of the ADAM sensor and signal output according to an embodiment of the invention, showing the results of measuring the spatiotemporal patterns of scratching movements and vibration signatures focusing on five different frequency ranges; the experiment involved four acceleration sensors, one each placed on the fingertip, finger, dorsum of the hand, and wrist while scratching by articulating the arm for 10 seconds and then scratching with just the fingers for the last 10 seconds; the color bars represent the average power of the signal per frequency in each frequency band, and the values ​​on the left and right indicate the minimum and maximum of the respective plots; the upper spatial pattern is from scratching by articulating the arm and the lower spatial pattern is from scratching with just the fingers. [Figure 1C] FIG. 1 shows an overview of an ADAM sensor and signal output according to an embodiment of the present invention, illustrating the scratch path and resulting data (raw and filtered data) from the sensor. [Figure 1D] FIG. 1 shows an overview of the ADAM sensor and signal output according to an embodiment of the present invention, with a comparison of signals captured during scratching in five different wearing positions as time series and spectrogram plots, showing that the dorsum of the hand is in an effective wearing position for capturing scratching with just the fingers. [Figure 2A] FIG. 13 represents data collected by an ADAM sensor according to an embodiment of the present invention, showing sample time series data and spectrograms corresponding to scratching activity, and illustrating two modes of scratching performed at two body sites, the dorsum of the hand (DH) and the forearm (FA), and at two intensities, high (H) and low (L). [Figure 2B] FIG. 13 illustrates data collected by an ADAM sensor according to an embodiment of the present invention, showing time series data of scratching activity for five different body parts including the head (hairy skin), arms (normal skin), abdomen (soft skin), knees (skin with bony protuberances), and legs (hard skin). [Figure 2C]FIG. 2C is a diagram representing data collected by an ADAM sensor according to an embodiment of the present invention, showing a spectrogram of each time series data of FIG. 2B, illustrating that the signal resulting from scratching activity has energy in the frequency range of 0 to 800 Hz. [Figure 2D] FIG. 13 depicts data collected by an ADAM sensor in accordance with an embodiment of the present invention, showing time series data for a range of non-scratching activities including simulating moving fingers in the air, waving a hand, composing a text message, typing on a keyboard, and clicking a mouse. [Figure 2E] FIG. 2C is a diagram representing data collected by an ADAM sensor according to an embodiment of the present invention, showing a spectrogram of each time series data of FIG. 2D, illustrating that signals resulting from non-scratching activity have energy exclusively in the range below 200 Hz. [Figure 3A] FIG. 1 shows a comparison of results obtained with an ADAM device on the hand and an Apple Watch with the ItchTracker Mobile Application in accordance with an embodiment of the present invention, showing time series data and spectrograms for various scratching (skin over the knee with bony prominences, soft belly skin, complex surface of the head) and non-scratching activities (waving, air scratching) with 5 seconds of rest between each activity, demonstrating scratching with full wrist joint motion, with "ItchTracker" misclassifying the waving motion as a scratch and failing to detect the head scratch. [Figure 3B]FIG. 10 shows a comparison of results obtained with an ADAM device worn on the hand and an Apple Watch with the ItchTracker Mobile Application according to an embodiment of the present invention, showing time series data and spectrograms for various scratching (skin on the knee with bony prominences, soft belly skin, skin on the head with a complex surface) and non-scratching activities (waving, air scratching) with 5 seconds of rest between each activity, showing scratching by finger articulation only, where the ADAM sensor reliably discriminates skin scratching using finger articulation only, whereas the ItchTracker cannot detect scratches using fingers alone due to its wrist-worn position and inability to record high frequency information. [Figure 4A] FIG. 1 shows an overview of the dataset, signal processing pipeline, and validation method according to an embodiment of the present invention, where algorithm validation data was collected from 10 healthy volunteers. [Figure 4B] FIG. 1 shows an overview of the dataset, signal processing pipeline, and validation method according to an embodiment of the present invention, illustrating the activities performed by each participant and the corresponding activity duration. [Figure 4C] FIG. 1 shows an overview of the dataset, signal processing pipeline, and validation method according to an embodiment of the invention, showing pre-processing subsequently applied to the set by applying a 1 second sliding window, segmenting it into several frames, a set of features was extracted from each frame, a random forest classifier was trained with the feature set extracted from the training frames, and the trained classifier was applied to a test feature set for prediction and validation. [Figure 4D] FIG. 1 shows an overview of the dataset, signal processing pipeline, and validation method according to an embodiment of the invention. Time series data from a validation cohort with a single subject performing a variety of scratching and non-scratching activities, including 5 second rest periods between each activity. [Figure 4E]FIG. 1 shows an overview of the dataset, signal processing pipeline, and validation method according to an embodiment of the present invention. The output probability for a scratch is characterized as "1: scratch" if the probability is above 50%, the blue line is the probability of a scratch for each frame, and the red line shows the final classification result, where 0 and 1 correspond to no scratch and scratch, respectively. [Figure 5A] FIG. 1 shows a summary of validation data and extracted features according to an embodiment of the present invention. Clinical validation was conducted in a natural home environment with mostly pediatric atopic dermatitis patients (median age 10.5 years), where an infrared IR camera was used to record the subjects' scratching behavior and manually scored by two clinical research staff members. [Figure 5B] FIG. 1 shows a summary of validation data and extracted features according to an embodiment of the invention. A total of 11 AD patients were recruited, generating a total of 46 nights of data in the pooled analysis. [Figure 5C] FIG. 1 shows a summary of validation data and extracted features according to an embodiment of the present invention. A total of nine features were used for LOSO and cross-study validation, which are listed in order of feature importance obtained from the random forest classifier. [Figure 6A] FIG. 13 illustrates power dissipation according to an embodiment of the present invention, showing power dissipation from fingertip to wrist determined using four devices and two devices, showing that devices on the fingers and hand do not affect attenuation between the fingertip and wrist. [Figure 6B] FIG. 1C shows power dissipation according to an embodiment of the present invention, showing power dissipation from the back of the hand to the wrist, with attenuation results along the sensor mounting locations using the four sensors in FIG. 1B being consistent with the results in FIG. 1D. [Figure 6C] FIG. 1 illustrates power dissipation according to an embodiment of the present invention, showing the power law frequency and distance dependent acoustic attenuation, P(x+Δx)=P(x)exp(−α0 fηΔx). [Figure 7]A diagram showing how mechanical features enable an ADAM sensor to conformably fit to the back of the hand during a variety of hand movements in accordance with an embodiment of the present invention. [Figure 8A] Normalized time series and spectral characteristics of two scratching modes (scratching with arm articulation and scratching with only fingers) and two body sites (dorsus of hand and forearm) according to an embodiment of the present invention, the spectral characteristics of high intensity and low intensity scratching events are generally the same when the scratching mode and scratching location are the same, and time series data plots, spectrograms, and Fourier transforms of high intensity and low intensity scratching, scratching the dorsum of the hand (DH) by articulating the arm. [Figure 8B] 1 shows normalized time series and spectral characteristics of two scratching modes (scratching with arm articulation and scratching with fingers only) and two body sites (dorsus of hand and forearm) according to an embodiment of the present invention, the spectral characteristics of high intensity and low intensity scratching events are generally the same when the scratching mode and scratching location are the same, and shows time series data plots, spectrograms, and Fourier transforms of high intensity and low intensity scratching, showing scratching of the forearm (FA) by articulating the arm. [Figure 8C] Normalized time series and spectral characteristics of two scratching modes (scratching with arm articulation and scratching with fingers only) and two body sites (dorsus of hand and forearm) according to an embodiment of the present invention; the spectral characteristics of high-intensity and low-intensity scratching events are generally the same when the scratching mode and scratching location are the same; and time series data plots, spectrograms, and Fourier transforms of high-intensity and low-intensity scratching of the DH scratched with fingers only. [Figure 8D]Normalized time series and spectral characteristics of two scratching modes (scratching with arm articulation and scratching with fingers only) and two body sites (dorsus of hand and forearm) according to an embodiment of the present invention; the spectral characteristics of high-intensity and low-intensity scratching events are generally the same when the scratching mode and scratching location are the same; and time series data plots, spectrograms, and Fourier transforms of high-intensity and low-intensity scratching with fingers only scratching the FA. [Figure 9] FIG. 13 illustrates the effect of the number of trees on classification performance, showing that using more than 50 trees provides minimal benefit according to an embodiment of the present invention. [Figure 10] FIG. 13 shows the effect of the minimum scratch duration window on the sensitivity and specificity of the definition of a scratch event according to an embodiment of the present invention, where the best performance is achieved when the minimum required scratch time is 4.5 seconds, and the specificity increases and the F1 score decreases as the minimum required scratch time increases from 1.5 seconds to 8.5 seconds. [Figure 11] FIG. 1 shows a subset of three AD patients wore sensors for over seven days to illustrate long-term scratching behavior in accordance with an embodiment of the present invention; these individuals also started a new topical medication (crisaborole) to assess therapeutic benefit; the scratching duration of the AD patients was obtained based on the scratching predictions of the scratching detection algorithm; the scratching duration of the AD patients varied from patient to patient based on symptom severity, with significant variation observed across multiple nights for a single patient. [Figure 12] FIG. 13 shows time series data collected from an ADAM sensor synchronized with IR video data on an ELAN in accordance with an embodiment of the present invention, where the sensor data is annotated with visual assessment of patient activity in a home natural environment. [Figure 13]FIG. 1 illustrates the probability of confusing non-scratching activities according to an embodiment of the present invention. The predicted probability of each class (scratching or non-scratching) was calculated from the five confusing activities investigated in this study. Scratching in the air, waving, and clicking a mouse were strongly predicted as non-scratching activities, while typing on a keyboard and entering text on a mobile phone had a greater probability of misclassifying as a scratch. [Figure 14] FIG. 13 shows a screenshot of a video illustrating a pediatric patient with AD wearing an ADAM sensor while scratching. A pediatric patient with severe atopic dermatitis is wearing an ADAM sensor while the scratching is recorded on video, and it is noted that the scratching intensity varies within the video, and the algorithm correctly identifies the scratching intensity through the amplitude of the sensor response. [Figure 15] FIG. 13 is a screenshot of a video where z-axis acceleration data sampled at 1,600 Hz has been converted to an audio file, showing that the scratching activity produces a harsh, periodic sound. [Figure 16] FIG. 13 is a screenshot of a video where z-axis acceleration data sampled at 1,600 Hz was converted to an audio file, showing that the sound of the waving motion does not contain any perceptible periodicity or sound compared to the scratching sound. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0065] The present invention will now be described in more detail below with reference to the accompanying drawings, in which exemplary embodiments of the invention are shown. However, the present invention may be embodied in many different forms and should not be construed as being limited to the embodiments set forth herein. Rather, these embodiments are presented so that the present invention will be thorough and complete, and so that the scope of the present invention will be fully conveyed to those skilled in the art. Like reference numerals refer to like elements throughout.

[0066] The terms used herein generally have their ordinary meaning in the art, within the context of the present invention and within the specific context in which each term is used. Some terms used to describe the present invention are explained below or elsewhere in the present specification to provide those skilled in the art with further guidance regarding the description of the present invention. For convenience, some terms may be highlighted, for example, using italics and / or quotation marks. The use of highlighting does not affect the scope and meaning of a term, which is the same in the same context whether or not it is highlighted. It will be understood that the same thing can be said in more than one way. Thus, alternative phrases and synonyms may be used for one or more of the terms discussed herein, and there is no special meaning to whether a term is detailed herein. Synonyms for some terms are provided. The listing of one or more synonyms does not exclude the use of other synonyms. The use of examples anywhere in this specification, including examples of any terms described herein, is merely illustrative and does not limit the scope and meaning of the present invention in any way. Similarly, the present invention is not limited to the various embodiments given herein.

[0067] Those skilled in the art will understand that starting materials, biological materials, reagents, synthesis methods, purification methods, analytical techniques, assay methods, and biological methods other than those specifically exemplified may be used in the practice of the present invention without resorting to undue experimentation. All functional equivalents known in the art of such materials and methods are intended to be included in the present invention. The terms and expressions used are used as expressions of description rather than of limitation, and there is no intention to use such terms and expressions to exclude equivalents of the features shown and described or portions thereof, and it is understood that various modifications are possible within the scope of the claimed invention. Thus, although the present invention has been specifically disclosed by preferred embodiments and optional features, it should be understood that those skilled in the art may resort to modifications and variations of the concepts disclosed herein as a last resort, and such modifications and variations are considered to be within the scope of the present invention as defined by the appended claims.

[0068] Whenever a range is given herein, e.g., a temperature range, a time range, or a composition or concentration range, all intermediate ranges and subranges, as well as all individual values ​​contained within the given range, are intended to be included in the present invention. It will be understood that any subrange or individual value within a range or subrange contained in the description herein may be excluded from the claims of the present invention.

[0069] As used throughout the description and claims that follow, the meaning of "a," or, if omitted ("a," "an," and "the"), will be understood to include plural referents unless the context clearly indicates otherwise. Thus, for example, a reference to a "cell" includes a plurality of such cells and their equivalents known to those of skill in the art. Similarly, the terms "a" ("a" or "an"), "one or more," and "at least one" can be used interchangeably herein. It should also be noted that the terms "comprising," "including," and "having" can be used interchangeably.

[0070] It will be understood that when an element is referred to as "on," "attached," "connected," "coupled," "in contact," or the like, it may be directly on, attached, connected, coupled, or in contact with the other element, or intervening elements may be present. In contrast, when an element is referred to as "directly on," "directly attached," "directly connected," "directly coupled," or "in direct contact with" another element, there are no intervening elements present. Those skilled in the art will also understand that reference to a structure or feature that is disposed "adjacent to" another feature may have portions that are on or under the adjacent feature.

[0071] It will be understood that terms such as first, second, third, etc. may be used to describe various elements, components, regions, layers, and / or sections, but these elements, components, regions, layers, and / or sections should not be limited by these terms. These terms are used only to distinguish one element, component, region, layer, or section from another element, component, region, layer, or section. Thus, a first element, component, region, layer, or section described below may be referred to as a second element, component, region, layer, or section without departing from the teachings of the present invention.

[0072] Additionally, relative terms such as "lower" or "bottom" and "upper" or "top" may be used herein to describe the relationship of one element to another element as depicted in the figures. It will be understood that the relative terms are intended to encompass different orientations of the device in addition to the orientation depicted in the figures. For example, if a device in one of the figures is flipped over, an element described as being "under" the other element would be oriented "upper" of the other element. Thus, the exemplary term "lower" can encompass both an orientation of "lower" and "upper", depending on the particular orientation of the figure. Similarly, if a device in one of the figures is flipped over, an element described as being "under" or "under" the other element would be oriented "upper" of the other element. Thus, the exemplary term "lower" or "underneath" can encompass both an orientation of above and below.

[0073] The terms "comprise" and / or "comprising", or "include" and / or "including", or "have" and / or "having", or "carry" and / or "carrying", or "contain" and / or "contains", or "involving" and / or "involving", "characterized by", and similar phrases should be understood to be open-ended, i.e., will be further understood to mean including but not limited to. When used in this disclosure, they specify the presence of stated features, regions, integers, steps, operations, elements, and / or components, and do not preclude the presence or addition of one or more other features, regions, integers, steps, operations, elements, components, and / or groups thereof.

[0074] Unless otherwise specified, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and the present invention, and it will be further understood that they are not to be interpreted in an idealized or overly formal sense unless expressly defined in this specification.

[0075] As used in this disclosure, "approximately," "about," "approximately," or "substantially" are generally intended to mean within 20 percent, preferably within 10 percent, and more preferably within 5 percent of a given value or range. Numerical quantities set forth herein are approximate, meaning that the language "approximately," "about," "approximately," or "substantially" can be inferred if not expressly stated.

[0076] As used in this disclosure, the phrase "at least one of A, B, and C" should be construed to mean the logical (A or B or C) using a non-exclusive logical OR. As used herein, the term "and / or" includes any and all combinations of the associated listed items.

[0077] As used in this disclosure, the term "wireless communication system" refers to on-board components of sensors, wireless controllers, and other electronic components that provide the ability to receive and transmit signals using at least one of the following communication protocols for wireless communication: Near Field Communication (NFC), Wi-Fi / Internet, Bluetooth, Bluetooth low energy (BLE), and cellular communication protocols. In this manner, output may be provided to an external device, including a cloud-based device, a personal portable device, or a caregiver computer system. Similarly, commands may be sent to the sensor by an external controller, etc., that may or may not be compatible with the external device. Machine learning algorithms may be used to improve the signal analysis, and then command signals may be sent to the medical sensor, including the stimulator of the medical sensor to provide a tactile signal to the user of the medical device to aid in treatment. More generally, these systems may be embedded in a processor, such as a microprocessor, that is located on-board the electronic device of the medical sensor or physically separate from the electronic device. One example of a wireless controller is a Near Field Communication (NFC) chip, including an NFC chip. NFC is a radio technology that allows two-way short-range wireless communication between devices. Another example of a wireless controller is a Bluetooth® chip, or BLE system-on-chip (SoC), which allows devices to communicate over standard radio frequencies rather than through cables, wires, or direct user actions.

[0078] As used in this disclosure, the term "conformable" refers to the ability of a device to undergo macroscopic deformations to maintain intimate contact with a curved surface without causing substantial deformation forces on the underlying surface. Such conformal contact is particularly challenging in the context of soft biological tissues, including skin, which are prone to deformation even under relatively light forces. Such undesirable deformation-generating forces may cause discomfort and irritation to the underlying tissue. Such problems are avoided in the present technology by specially configuring the penetrating member, light source, and supporting substrate, such that the device as a whole is flexible and can accommodate a wide range of complex shapes, thereby ensuring that conformal contact is maintained. Such conformal contact may be further quantified in that there is no significant separation distance between the device and the underlying surface. This helps ensure maximum light irradiation at the desired location and avoids unwanted light leakage. Conformability may also be described in terms of the device being "flexible" and having a bending modulus suitable to provide the desired curvature under external forces. The conformable may be further described in terms of desired mechanical properties such as bending stiffness and Young's modulus such that no voids are present, thereby maximizing light transmission to the underlying tissue.

[0079] The terms "flexible" and "bendable", as used in this disclosure, refer to the ability of a material, structure, device, or device component to be deformed into a curved or bent shape without undergoing a transformation that introduces significant strain, such as a strain that characterizes the breaking point of the material, structure, device, or device component. In an exemplary embodiment, a flexible material, structure, device, or device component may be deformed into a curved shape without introducing a strain of 5% or more in the strain-sensitive region, for some applications, 1% or more, and for other applications, 0.5% or more. As used herein, some, but not necessarily all, flexible structures are also stretchable. A variety of properties are realized in the flexible structures (e.g., device components) of the present invention, including material properties such as low modulus of elasticity, bending stiffness, and flexural stiffness, physical dimensions such as small average thickness (e.g., less than 100 microns, optionally less than 10 microns, and optionally less than 1 micron), and device geometries such as thin films and open or mesh geometries.

[0080] The term "bending stiffness" refers to a mechanical property of a material, device, or layer that describes the resistance of the material, device, or layer to an applied bending moment. Generally, bending stiffness is defined as the product of the Young's modulus and the second moment of area of ​​the material, device, or layer. Materials with non-uniform bending stiffness may optionally be described in terms of a "bulk" or "average" bending stiffness for the entire layer of material.

[0081] The term "elastomer" as used in this disclosure refers to a polymeric material that can be stretched or deformed and return to its original shape without substantial permanent deformation. Elastomers generally undergo substantial elastic deformation. Useful elastomers include those that include polymers, copolymers, composites, or mixtures of polymers and copolymers. An elastomeric layer refers to a layer that includes at least one elastomer. Elastomeric layers may also include dopants and other non-elastomeric materials. Useful elastomers include, but are not limited to, thermoplastic elastomers, styrene materials, olefinic materials, polyolefins, polyurethane thermoplastic elastomers, polyamides, synthetic rubbers, PDMS, polybutadiene, polyisobutylene, poly(styrene butadiene styrene), polyurethanes, polychloroprene, and silicones. Exemplary elastomers include, but are not limited to, silicon-containing polymers such as polysiloxanes, including poly(dimethylsiloxane) (i.e., PDMS and h-PDMS), poly(methylsiloxane), partially alkylated poly(methylsiloxane), poly(alkylmethylsiloxane), and poly(phenylmethylsiloxane), silicon modified elastomers, thermoplastic elastomers, styrene materials, olefinic materials, polyolefins, polyurethane thermoplastic elastomers, polyamides, synthetic rubbers, polyisobutylene, poly(styrene butadiene styrene), polyurethanes, polychloroprene, and silicones. In one embodiment, the flexible polymer is a flexible elastomer.

[0082] "Encapsulate" or "encapsulation," as used in this disclosure, refers to the orientation of one structure such that it is at least partially, and in some cases completely, surrounded by one or more other structures. "Partially encapsulated" refers to the orientation of one structure such that it is partially surrounded by one or more other structures. "Completely encapsulated" refers to the orientation of one structure such that it is completely surrounded by one or more other structures. The present invention includes devices having partially or completely encapsulated electronic devices, device components, and / or inorganic semiconductor components.

[0083] The embodiments of the present invention are illustrated in detail below with reference to the accompanying drawings. The following description is merely exemplary in nature and is in no way intended to limit the invention, its application, or uses. The broad teachings of the present invention may be implemented in various forms. Thus, although the present invention includes specific examples, the true scope of the present invention should not be so limited, as other modifications will become apparent after studying the drawings, the specification, and the following claims. For clarity, the same reference numbers are used in the drawings to identify similar elements. It will be understood that one or more steps in a method may be performed in different orders (or simultaneously) without altering the principles of the present invention.

[0084] Measuring behaviors associated with scratching provides a way to quantify the severity and frequency of pruritus. Direct visual inspection of scratching on video recordings is the gold standard, but such a scheme is laborious and impractical in clinical practice. Wrist kinetic recording is a method of recording movement using an accelerometer, which estimates scratching by measuring wrist motion, which is less than optimal in accuracy because it cannot distinguish, for example, hand shaking from scratching. Additional aspects of the various characteristics of scratching, including inter-individual variability, pose additional unsolved challenges. Thus, existing sensors and strategies do not provide sufficient performance or practicality for routine use in clinical trials or patient treatment.

[0085] The present invention discloses, among other things, a strategy that utilizes a soft, flexible, wireless, low-profile sensor that can capture both vibration and motion signatures of physiological processes with clinical-grade data quality. The device, also referred to as the ADAM (ADvanced Acousto-Mechanic) sensor, softly couples to the skin and allows for the capture of both low-frequency (e.g., motion) and high-frequency (acousto-mechanical) signals from the human body. A built-in rechargeable battery allows for 7 days of continuous operation on a single wireless charge. When placed on the dorsum of the hand, the ADAM sensor can capture the acousto-mechanical signals associated with scratching through a combination of motion and acousto-mechanical signals in a manner that is not affected by any type of ambient sound. In particular, when placed between the metacarpals of the second and third fingers, the ADAM sensor can quantify scratching behaviors initiated not only from wrist and / or arm motions but also from the fingers and fingertips over a wide temporal bandwidth that includes the high-frequency vibration motion associated with scratching itself. Using the sensor data and an associated understanding of the physics of the scratching process, signal features that uniquely characterize scratching activity can be extracted to train and validate machine learning-based algorithms for scratching detection.

[0086] In one aspect of the invention, the ADAM sensor comprises at least one inertial measurement unit (IMU) for operatively detecting vibration and motion signatures of a physiological process, a microcontroller unit (MCU) coupled to the at least one IMU for receiving data from the at least one IMU and processing the received data, and a wireless communication system for wireless data transmission. The acousto-mechanical sensor is configured to measure both low frequency signals below about 100 Hz and high frequency signals above about 100 Hz derived from the at least one IMU. Further, the acousto-mechanical sensor is skin-mounted to enable coupling of the at least one IMU to the skin to capture both high frequency and low frequency signals. The high frequency and low frequency signals represent a physiological response to pruritus as a scratching or rubbing activity. In one embodiment, the physiological process includes a scratching activity.

[0087] In one embodiment, the plurality of electronic components and the at least one IMU are separated such that the plurality of electronic components are at one end of the ADAM sensor and the at least one IMU is at another end of the ADAM sensor.

[0088] In one embodiment, the ADAM sensor is load-free, stretchable, twistable, and / or bendable, hi one embodiment, the ADAM sensor has a specific geometric polarity for attachment to an anatomical location of the body, and is flexible and conformable to the skin.

[0089] In one embodiment, the at least one IMU comprises a millimeter-scale three-axis accelerometer configured to detect skin-conducted vibrations and motion associated with scratching activity. In one embodiment, the three-axis accelerometer is configured to detect acceleration along the x-axis and the y-axis at a first sample rate and the z-axis at a second sample rate, where the x-axis and the y-axis are oriented parallel to the skin surface to which the ADAM sensor is attached and the z-axis is perpendicular to the skin surface. In one embodiment, the first sample rate is in the range of about 160-240 Hz and the second sample rate is in the range of about 1,200-2,000 Hz. In one embodiment, the three-axis accelerometer is configured to have a sampling rate of about 1,600 Hz with 16-bit resolution and a dynamic range of about ±2 g, where g is the acceleration of gravity, 9.8 m / s 2 It is.

[0090] In one embodiment, the ADAM sensor further comprises a flexible printed circuit board (fPCB) for supporting and interconnecting the at least one IMU and the plurality of electronic components. In one embodiment, the fPCB employs a patterned layout including a plurality of freely deformable, serpentine interconnects designed to mechanically decouple the at least one IMU from other components of the ADAM sensor. In one embodiment, the plurality of freely deformable, serpentine interconnects are flexible and stretchable and adapted to interconnect the at least one IMU and the plurality of electronic components.

[0091] In one embodiment, the ADAM sensor also includes an elastomeric encapsulation layer that at least partially surrounds the at least one IMU, the plurality of electronic components, and the fPCB and forms a tissue-facing surface that is operably attached to the living body. In one embodiment, the tissue-facing surface of the elastomeric encapsulation layer is configured to conformally fit to the skin surface of the living body. In one embodiment, the encapsulation layer is formed from a thin, low modulus silicone elastomer. In one embodiment, the elastomeric encapsulation layer is a biocompatible silicone enclosure.

[0092] In one embodiment, the ADAM sensor further comprises a biocompatible hydrogel adhesive for attaching the ADAM sensor to an anatomical location in a living body, hi one embodiment, the biocompatible hydrogel adhesive is a medical grade adhesive.

[0093] In one embodiment, the wireless communication system comprises a Bluetooth low energy (BLE) radio. In one embodiment, the wireless communication system is configured to transmit the output signal from the ADAM sensor to an external device.

[0094] In one embodiment, the plurality of electronic components further comprises a power module coupled to the at least one IMU, the wireless communication system, and the MCU for providing power. In one embodiment, the power module comprises a battery. In one embodiment, the battery is a rechargeable battery. In one embodiment, the power module further comprises a wireless charging module for wirelessly charging the rechargeable battery. In one embodiment, the power module further comprises a fault prevention element including a short circuit protection component or circuit to avoid malfunction of the battery.

[0095] In one embodiment, when placed on the dorsum of the hand, the ADAM sensor can detect acoustic-mechanical signals associated with scratching activity through a combination of movement and acoustic-mechanical signals in a manner that is not affected by ambient noise. In one embodiment, when placed between the metacarpals of the second and third fingers, the ADAM sensor can quantify scratching activity initiated not only from wrist and / or arm movements, but also from the fingers and fingertips, over a wide temporal bandwidth that includes the high frequency vibrational movement associated with scratching itself.

[0096] In one embodiment, scratching activity is characterized by a first and second type of signal, the first type of signal corresponding to the overall movement of the hand, and the second type of signal overlapping in time with the first type of signal and resulting from subtle vibration impulses generated by the movement of the fingertip and fingernail against the contact surface. In one embodiment, the first type of signal has a characteristic frequency in the range of a few Hz or less, and the second type of signal has a characteristic frequency ranging over a range of several hundred Hz, with amplitudes that decay rapidly with position along the fingers, into the hand, and finally through the wrist and into the arm. In one embodiment, the first and second types of signals are identified by passing the data through low-pass and high-pass filters with a cutoff of 2 Hz.

[0097] In one embodiment, signal features that uniquely characterize scratching activity are extracted from the output signal of the ADAM sensor to train and validate machine learning based algorithms for scratch detection and to quantify the pruritus symptom.

[0098] In one embodiment, the ADAM sensor is operatively positioned on the back of the hand, fingernail, finger, or wrist to measure scratching movements over time.

[0099] In one embodiment, the ADAM sensor can be used in any disease state where itch is a symptom, as a diagnosis, a therapeutic response indicator, a clinical trial endpoint, or an early warning of disease progression.

[0100] In another embodiment of the present invention, a method for quantifying a symptom of pruritus includes attaching an advanced acousto-mechanical (ADAM) sensor to an anatomical location of a living body, detecting signals associated with scratching and / or non-scratching activity by the ADAM sensor, and analyzing the signals to assess an acousto-mechanical signature of scratching to quantify the symptom of pruritus.

[0101] In one embodiment, analyzing the signal includes preprocessing the signal to remove low frequency baseline fluctuations caused by slow large scale motions, such that the preprocessed signal contains more specific high frequency signal components associated with z-axis acceleration, segmenting the preprocessed signal into frames by sliding overlapping n-second windows, extracting a set of features from each frame, and applying a Random Forest (RF) classifier to the extracted set of features for prediction and validation. In one embodiment, the n-second window is a 1-second window. In one embodiment, the overlap is in the range of about 50%-90% overlap, preferably about 50%.

[0102] In one embodiment, analyzing the signals further includes assigning each frame a label of 0, indicating non-scratching activity, if less than about 50% of the data points in a given frame are labeled as scratching activity, and otherwise assigning the frame a label of 1, indicating scratching activity.

[0103] In one embodiment, the RF classifier is trained with a training dataset, which includes scratching data from multiple body locations and various types of non-scratching activities collected from a large number of healthy subjects.

[0104] In one embodiment, the RF classifier is optimized by comparing results from leave-one-out cross-validation (LOSO-CV) applied to a training dataset, the results of which form the basis for optimizing the RF classifier.

[0105] In one embodiment, the parameters for optimization include the number of decision trees in the RF classifier and the minimum required scratch (MRS) period.

[0106] In one embodiment, the number of decision trees is in the range of about 10 to 200, preferably about 30.

[0107] In one embodiment, for optimization, the RF classifier returns predictions at the 1-second frame level, and the output is then clustered using Density-Based Spatial Clustering for Noisy Applications (DBSCAN), which concatenates two closely spaced scratch frames as a single scratch event, with the MRS duration defining the minimum number of frames to form a cluster of scratch events.

[0108] In one embodiment, the MRS duration is in the range of about 1.5 seconds to 8.5 seconds, and preferably is at least about 4.5 seconds.

[0109] In one embodiment, the scratching and / or non-scratching activities include scratching the living body's head, arms, stomach, knees, and / or legs, and / or simulating scratching in the air, waving hands, entering text into a phone, typing on a keyboard, and / or clicking a mouse.

[0110] In one embodiment, the anatomical locations are the fingertip, the finger, the dorsum of the hand, and the wrist of a living subject.

[0111] It should be noted that all or part of the steps according to the embodiments of the present invention are implemented by hardware or a program instructing related hardware.

[0112] Yet another aspect of the present invention provides a tangible, non-transitory computer readable medium storing instructions that, when executed by one or more processors, cause the quantification of symptoms of pruritus in the manner disclosed above. The computer executable instructions or program code enable the apparatus or similar systems disclosed above to complete various operations according to the method disclosed above. The storage medium / memory may include, but is not limited to, high speed random access media / memory such as DRAM, SRAM, DDR RAM, or other random access solid state memory devices, and non-volatile memory such as one or more magnetic disk storage devices, optical disk storage devices, flash memory devices, or other non-volatile solid state storage devices.

[0113] According to the present invention, a non-invasive technique can objectively quantify scratching behavior via soft, flexible, wireless sensors that capture the acoustic-mechanical signature of scratching from the dorsum of the hand. Using the sensor data and associated understanding of the physics of the scratching process, signal features that uniquely characterize scratching activity can be extracted to train and validate machine learning-based algorithms for scratch detection.

[0114] Additionally, clinical validation in a cohort of mostly pediatric patients (n=11) with moderate to severe atopic dermatitis included 46 nights of sleep totaling 378.4 hours. Data show approximately 99.0% accuracy (approximately 84.3% sensitivity, approximately 99.3% specificity) versus visual observation. This study suggests a broad range of capabilities relevant for applications ranging from evaluating the efficacy of drugs against pruritus-causing conditions to monitoring disease severity and treatment response.

[0115] These and other aspects of the present invention are further described below. Without intending to limit the scope of the present invention, exemplary instruments, devices, methods, and results related thereto according to the embodiments of the present invention are presented below. Please note that titles or subtitles may be used in the examples for the convenience of the reader, but this does not limit the scope of the present invention in any way. Furthermore, although certain theories are proposed and disclosed herein, whether they are right or wrong, they should not limit the scope of the present invention in any way as long as the present invention is carried out according to the present invention without regard to any particular theory or mode of operation.

[0116] (Example) A skin-conformable wireless sensor for objectively quantifying pruritus symptoms In this exemplary study, we develop a strategy that utilizes a soft, flexible, wireless, low-profile sensor that can capture both vibration and motion signatures of physiological processes with clinical-grade data quality. This device, also referred to in this disclosure as the ADAM (ADvanced Acousto-Mechanic) sensor, softly couples to the skin and allows for the capture of both low-frequency (e.g., motion) and high-frequency (acousto-mechanical) signals from the human body. A built-in rechargeable battery allows for 7 days of continuous operation on a single wireless charge. When placed on the dorsum of the hand, the ADAM sensor can capture the acousto-mechanical signals associated with scratching through a combination of motion and acousto-mechanical signals in a manner that is unaffected by ambient noise. In particular, when placed between the metacarpals of the second and third fingers, the ADAM sensor can quantify scratching behaviors initiated not only from wrist and / or arm motions, but also from the fingers and fingertips, over a wide temporal bandwidth that includes the high-frequency vibration motion associated with scratching itself. Using the sensor data and an associated understanding of the physics of the scratching process, signal features that uniquely characterize scratching activity can be extracted to train and validate machine learning-based algorithms for scratching detection.

[0117] The paper begins with a description of a systematic investigation into the mechanisms of scratching in a set of controlled experiments, followed by validation of the combined use of sensor signals and data analytics approaches in two subject studies: an algorithm development study (n=10 healthy subjects) and a clinical validation study (n=11 mostly pediatric AD patients). The training study involves collecting scratching and non-scratching data under well-defined conditions to develop a machine learning (ML) algorithm that classifies and quantifies scratching behavior. The clinical study validates the sensor and algorithm in a cohort of moderate to severe AD patients. Performance is comparable to a gold standard defined by direct visual observation via manually labeled IR camera recordings.

[0118] Study design Two datasets were collected independently for model development (training) and validation. The training data, collected under a controlled environment, were used as the basis for developing an ML algorithm to detect scratching behavior. The validation dataset was collected from AD patients in a naturalistic home environment (Figure 5A). These data were used to evaluate the generalizability of the ML algorithm. A total of 10 participants (5 males and 5 females) participated in the training data collection, aged between 22 and 27 years (24 ± 1.6 years). The validation data collection involved 11 patients with moderate to severe AD (2 males and 9 females), aged between 4 and 24 years (10 ± 9.1 years).

[0119] Training data were collected from healthy subjects. Participants performed scratching and non-scratching activities, each lasting approximately 35 seconds. Scratching activities were performed over clothing on the thighs, abdomen, upper arms, calves, knees, and shoulders. Each participant then scratched the skin on the inner elbow, outer elbow, forearm, dorsum of the hand, cheek, head, palm, and neck. Scratching was considered a variation of scratching for the purposes of these studies. Each of these scratching activities was performed twice, once while sitting in a chair and once while lying in bed. For non-scratching activities, participants waved their hands, sent a text message to a friend, rolled over in bed, sat idle, typed on a keyboard, and banged on a desk. These non-scratching activities were included as potential confounders to the actual scratching activities. Data captured from each activity was saved as a separate file with a description of each activity annotated as scratching or non-scratching.

[0120] Validation data were collected from AD patients in their home environment as part of a clinical study approved by the Institutional Review Boards at Northwestern University and Ann & Robert H. Lurie Children's Hospital (IRB 2018-2111). Inclusion criteria included patients with mild to severe AD who were older than 2 years of age. Patients with active skin or systemic infections were excluded. Eligible participants were screened at the Northwestern Medicine Department of Dermatology and other pediatric dermatology or allergy departments at Ann & Robert H. Lurie Children's Hospital. A total of 11 subjects were recruited for validation data collection (Figure 5B). Written informed consent was first obtained for subjects who agreed to participate. Clinical research personnel then completed the Eczema Area and Severity Index (EASI) assessment and Investigator Global Assessment (IGA) for each patient. After completion of the EASI and IGA, each patient was trained on the procedure for operation of the ADAM sensor and overnight data collection. During training, patients were provided with specific instructions to place the ADAM sensor on the dorsum of the hand and secure it with a Coban self-adhesive wrap. Patients then received a package containing the ADAM sensor, an iPhone, a tripod, an IR camera, a charging device, an instruction manual, and a checklist. During validation data collection, each patient's nighttime activities were recorded with an IR video camera. The ADAM sensor data was compared to the IR video data for data annotation. This annotation process was performed using the Data Visualization and Annotation Software (Figure 12).

[0121] Sensor Operation and Performance The ADAM sensor is a small, soft, stretchy, wireless device that can be attached to the curved surface of the dorsum of the hand using a thin adhesive and positioned in various orientations across a range of possible anatomical locations (Figure 1A). In some embodiments, the ADAM sensor includes a BLE radio, electronics, and a rechargeable battery at one end of the sensor, and a millimeter-scale 3-axis accelerometer at the other end with a sampling rate of 1,600 Hz with 16-bit resolution and a dynamic range of ±2 g, where g is the acceleration due to gravity, and is 9.8 m / s 2 The work reported here focuses primarily on acceleration measured in a direction perpendicular to the skin's surface. A thin, low modulus silicone elastomer forms a skin-compatible encapsulating package as a watertight enclosure for a flexible printed circuit board (fPCB) that supports and interconnects these various components. The fPCB employs a patterned layout that includes a collection of freely deformable serpentine interconnects and is specifically designed by using computational modeling to mechanically decouple the accelerometer from the other components of the device. This design optimizes the device's ability to accurately and unconstrainedly record subtle movements of the skin's surface. Figure 1A shows an image highlighting these "soft" mechanical attributes and the overall "patch" form factor.

[0122] The ADAM sensor's high bandwidth triaxial accelerometer captures the skin-conducted vibrations as well as the overall motion associated with the scratching behavior. The ADAM sensor captured acceleration along the x- and y-axes at a rate of 200 Hz, while the z-axis acceleration was sampled at a rate of 1,600 Hz. The x- and y-axes were oriented parallel to the skin surface, and the z-axis was oriented perpendicular to the skin surface. High frequency components are attenuated as the disturbance propagates from its origin at the fingertip along the finger to the hand and finally to the wrist. The location of the ADAM sensor in this manner determines the nature and quality of the signal. Experiments to investigate the dependency on the wearing position involved a set of repeated scratching activities (10 seconds of scratching with arm articulation and 10 seconds of finger scratching, each separated by 5 seconds of idle) with different placements and orientations of the sensor. Based on signal quality and wearing considerations, the ideal location is between the first and second metacarpal fingers.

[0123] Scratching activity generates two types of signals. The first type of signal corresponds to the gross movement of the hand, with characteristic frequencies in the range of a few Hz or less. The second type of signal overlaps in time with the first type of signal and results from subtle vibration impulses generated by the movement of the fingertip and nail against the contact surface. Here, the frequency ranges from hundreds of Hz and decays rapidly with position along the finger, into the hand, and finally through the wrist and into the arm. Detailed multi-accelerometer measurements hint at the physics of this second type of signal, albeit in the context of a haptic interface. Similar direct measurements of acceleration at different positions from the fingertip to the wrist, as shown in Figure 1B, reveal the spectral characteristics of the signal decay for scratching behavior. The measurements involve wireless accelerometers placed on the fingertip, midpoint along the finger, dorsum of the hand, and wrist, as in the right frame shown in Figure 1B (also shown in Figure 6A). The test involves articulating the arm for 10 seconds, then capturing the signals resulting from scratching the surface of the skin by scratching with the fingers only for 10 seconds. The peak-to-peak amplitude of acceleration at the surface of the skin attenuates with increasing distance along the surface of the fingers and hand to the wrist. The color bars of the spatial patterns in Figure 1B represent the average power of the signal per frequency in each frequency band. The numbers on the left and right indicate the maximum and minimum values ​​for each of the respective plots. For scratching by articulating the arm (first row of spatial patterns), the power attenuates from a starting point of 20 dB / Hz at the fingertips to -32 dB / Hz at the wrist in the frequency range of 10 to 100 Hz. The attenuation increases exponentially with frequency. Specifically, the signal decreases from -19 dB / Hz to -37 dB / Hz and from -20 dB / Hz to -41 dB / Hz for the frequency ranges of 100 to 200 Hz and 200 to 400 Hz, respectively. For both types of scratches, the approximate power losses from fingertip to wrist are 2 dB in the frequency range of 0.1 to 10 Hz, 12 dB in the frequency range of 10 to 100 Hz, 20 dB in the frequency ranges of 100 to 200 Hz and 200 to 400 Hz, and 16 dB in the frequency range of 400 to 800 Hz.As expected, the attenuation at low frequencies is much less than the attenuation at high frequencies. The power of the signal decreases exponentially with increasing propagation distance and with increasing frequency. The power law frequency and distance dependent acoustic attenuation is given by Eq. P(x+Δx)=P(x)exp(-α0f η Δx) where P is the signal power, f is the frequency, Δx is the wave propagation distance, α0 is the damping coefficient, and η is the damping constant, which is a frequency-dependent property of viscoelastic materials obtained by fitting experimental data. In the experiment in Figure 1B, η has a value of 0.38 for vibration propagating from the fingertip to the back of the hand (Figure 6C). These results suggest that the very important high-frequency information associated with scratching is most effectively captured by (1) placing the ADAM sensor in close proximity to the fingertip and (2) operating the accelerometer in high-bandwidth mode.

[0124] Existing scratch sensors use relatively low bandwidth accelerometers (typically 50 Hz or less) and associated electronics in a rigid enclosure that couples to the wrist using a band. These existing devices are able to capture the first type of signal but are unable to record the second type of signal with sufficient fidelity, consistent with the propagation results described above. This limitation is due to (1) the non-ideal wearing position of the wrist, which has a small amplitude of high frequency signals, (2) insufficient measurement bandwidth to capture high frequency information, and (3) a loose mechanical coupling to the skin. In contrast, the ADAM sensor quantifies both types of signals with high accuracy due to its close proximity to the fingertip / nail, high bandwidth operation, and close mechanical interface with the skin. The result is an excellent ability to identify scratching events with high specificity and sensitivity. Figure 1C shows representative data generated by scratching the arm with the index finger of the opposing hand, as illustrated in the left frame. The motion involves (1) translating the hand, wrist, and arm with the fingers in a fixed position for 10 seconds, and then (2) articulating only the fingers for 10 seconds. The two types of signals described above can be discerned by passing the data through low-pass and high-pass filters with a cutoff of 2 Hz. The first type of signal represents hand motion and may serve as the basis for detecting scratching events, but when used in isolation, without the inclusion of the second type of signal, may produce false positive results from hand shaking, essential tremor, and other sources of motion unrelated to scratching. As exemplified by the red arrows described above and shown in Figure 1C, important high-frequency information emanates from the fingertips and is transmitted through the soft tissue and metatarsals to the sensor location. These features include fingertip sliding, and in some cases, energy over a wide range of frequencies that are time-synchronized with hand motion. Penetrating the body results in significant attenuation, especially at frequencies higher than 100 Hz, but over distances of several centimeters, the complex waveform is well preserved.

[0125] These effects are manifested in the characteristics of signals captured by ADAM sensors placed on the dorsum of the hand and wrist in both horizontal and vertical orientations, as well as on a smartwatch (such as an Apple Watch Series 4) secured to the wrist. Measurements involve first articulating the finger with a scratching motion in the air for 2.5 seconds, then scratching the surface of the skin with the same motion for 2.5 seconds. The time series and spectrogram displays of the data in Figure 1D clearly show the high frequency components associated with scratching the skin against the air in all cases. The spectrograms in Figure 1D use a Hamming window with a frame size of 0.2 seconds and an overlap duration of 0.19 seconds. These discriminatory features extend in frequency up to 600 Hz, with some energy extending further up to 800 Hz, but the most significant amplitudes are in the range up to 200 Hz. The results show only a modest dependence on wearing position, provided the sensor is in the radial half of the dorsum of the hand. As previously described and highlighted by the red dots in Figure 1D, orientation has little effect simply because the accelerometer is in front of the ADAM sensor. At the wrist, high frequency content appears only as a weak feature in the data, consistent with previous attenuation results shown in Figure 1B (Figure 6B). When worn on the back of the hand where signal strength is higher, the soft physical properties of the ADAM sensor and the medical grade hypoallergenic disposable adhesive support a robust yet compliant and non-irritating interface for many everyday activities involving a variety of hand and wrist movements, as shown in Figure 7.

[0126] Characterize the scratch Scratching activity refers to a collection of movements that press the fingertip against a surface, move the fingertip, and cause friction. The friction between the fingertip and the surface generates vibrations that propagate along the surface. Scratch signals may differ from each other because the vibrations between the fingertip and the surface depend on many factors, including but not limited to scratching speed, scratching strength, and scratching mode (e.g., using only the finger or moving the whole arm). Experiments to explore the time and frequency characteristics associated with these factors included scratching the head, arm, knee, and leg, as well as various non-scratching events such as waving, texting, typing on a keyboard, clicking a computer mouse, and moving a finger in the air. Figure 13 shows the predicted labeling for these confounding factors. The system shows the best performance (above 0.7) in distinguishing between scratching a finger in the air, waving a hand, and clicking a mouse.

[0127] Scratching responds to both internal and external stimuli and varies from person to person. As investigated in the tests summarized above and shown in Figures 1B-1D, scratching can be performed solely through elbow, wrist, or finger joint movements. In the mode involving the entire upper arm, the fingers press against the surface and rest, while the movement of the forearm articulated at the elbow joint drives the scratching movement. On the other hand, the finger scratching mode mainly involves localized movements of the fingers and no significant movement of the forearm or wrist. A mixture of these two modes is also possible. Depending on the scratching location and the severity of the itch, individuals use different scratching modes interchangeably. For any type of scratching movement, the intensity is also important. The force at the contact point determines the friction between the fingertip and the surface and therefore, together with the frequency and speed, can be considered as a metric of the intensity of the scratching event. These factors contribute to differences in the signal captured by the ADAM sensor. A systematic study of different types of scratching activities is an important precursor to the development of a generalizable scratch detection algorithm.

[0128] A control experiment for this purpose involved two body positions, the dorsum of the hand (DH) and the forearm (FA) (Figure 2A), with an ADAM sensor attached to the dorsum of the left hand, as shown in Figure 1D, where the two positions are scratched at high intensity followed by low intensity with only the arm articulated and then only the fingers articulated. The solid blue line in Figure 2A shows the time-dependent z-axis acceleration from the oscillatory motion of the surface of the skin, in a direction perpendicular to the plane of the scratching motion. The peak-to-peak amplitude is, as expected, larger for the high intensity scratching compared to the low intensity case. For example, the amplitude of the signal for the high intensity FA scratching (45-55 s) is 1.5 g, which is more than twice as large as for the low intensity case (65-75 s, 0.7 g). For an otherwise somehow similar scenario but with only the fingers articulated, the high (125-135 s) and low (145-155 s) amplitudes are 0.4 g and 0.15 g, respectively. Results for arm-only articulation (0-80 s) show larger amplitudes compared to fingers (80-160 s) for each scratch location (HD, FA) and intensity level. The frequency domain content associated with these signals appears as spectrograms (Figure 2A) plotted using the same window and frame sizes as Figure 1C. In all cases, even at low intensities with fingers alone, high frequency content, especially up to ~200 Hz, features prominently in the data. As in Figure 8, the normalized spectral characteristics of high and low intensity scratching events are broadly the same.

[0129] Scratching activities in five different body positions (head, arm, belly, knee, and leg), as well as non-scratching activities (simulating finger scratching in the air, waving, composing a text on a mobile phone, typing on a keyboard, and clicking a mouse), highlight important differences in the signals and the importance of high-frequency components in a range of practical scenarios (Figures 2B and 2D). Notably, data for all scratching activities have significant energy above 200 Hz, regardless of position (Figure 2C). As expected based on the previous explanation, spectrograms of most non-scratching activities show minimal energy in this frequency range (Figure 2E). Composing a text message, typing on a keyboard, and clicking a computer mouse are exceptions, with data showing contributions from 100 Hz to 200 Hz attributable to the impulse components of these movements. Nevertheless, the spectral content across all frequencies, the temporal characteristics of the signal, and low-frequency information can help distinguish such activities from scratching. In other words, although the high frequency components of the recorded signal are important indicators of scratching, these must be used in conjunction with other features of the data to achieve high selectivity for a specific, relatively small, set of confounding activities. Furthermore, measuring nighttime scratching alleviates the need to distinguish it from waking activities such as typing or composing text messages. As illustrated, Figures 14-15 highlight one form of manual signal analysis that involves conversion of raw data into an audio file that can be interpreted in a manner similar to sounds from a stethoscope. Various activities such as those described above can be instantly distinguished from scratching in this manner. These combined considerations motivate a machine learning approach to data analysis, where a variety of features, including but not limited to high frequency components, contribute to a classification scheme for unparalleled combined levels of sensitivity and specificity.

[0130] Hand-worn vs. wrist-worn scratch sensing As mentioned above, scratching behavior can occur via joint motions of the fingers, wrist, elbow, or shoulder. An important limitation of wrist-worn sensors is the inability to measure the movements associated with scratching of the fingers alone and the difficulty in distinguishing between hand shaking and scratching. Controlled studies reveal the performance of the ADAM sensor compared to a wrist-fixed system (e.g., Apple Watch Series 4) with an integrated mobile application to quantify scratching.

[0131] As shown in the left frame of Figure 3A, the Apple Watch Series 4 was placed on the left wrist and the ADAM sensor was placed on the back of the left hand. This configuration was intended to synchronize the data collected from both devices for the purpose of direct comparison. The tight coupling of the Apple Watch Series 4 to the back of the left hand was verified by ensuring that the smartwatch could not freely rotate around the wrist. Similarly, the tight attachment of the ADAM sensor was verified by bending and straightening the left hand to ensure that the ADAM sensor was in contact with the skin. After placing the sensors, a set of scratching and non-scratching activities shown in the left frames of Figures 3A-B were performed in sequence. One of the scratching activities involved the movement of the whole arm / hand, while the other scratching method used only the fingers. In addition, a simple hand-waving activity was performed to determine the ability to distinguish simple hand movements from scratching activities.

[0132] This comparison focuses on two modes of scratching, one mode with arm joint motion and the other with finger only scratching. To investigate scratching activity with arm joint motion, we included knee scratching, hand shaking, belly scratching, and head scratching continuously for 10 seconds with 5 seconds pause between, as shown in Figure 3A, where the blue line represents the time series of raw data and the red line shows the final binary classification result, where 0 and 1 correspond to non-scratching and scratching, respectively. The black dashed line in Figure 3A highlights the classification result from the app ItchTracker, where 0 represents non-scratching activity and 0.5 represents scratching activity. ItchTracker captures knee scratching and belly scratching, but misclassifies hand shaking and head scratching. The former arises from a heavy reliance on hand motion for scratch detection. The latter probably arises because head scratching only involves a short range of wrist motion. The ADAM sensor performs well across all of these and other scenarios.

[0133] Additional experiments investigate a scratching mode that involves only knuckle motion (Figure 3B). Similar to the previous comparison of scratching using the arm, these four activities include scratching the knee, simulating scratching in the air, scratching the stomach, and scratching the head. In these activities, ItchTracker does not detect any instances of scratching due to the lack of wrist motion. On the other hand, the ADAM sensor detects all instances and accurately distinguishes scratches even from pseudo-scratches (Figure 3B). The simulated scratching in the air motion provides an additional test of the ADAM system. A significant difference between scratching and pseudo-scratching is the presence of a frequency signature higher than 100 Hz. The ADAM system is capable of capturing frequencies as high as 800 Hz, whereas most wrist-fixed accelerometers, such as the Apple Watch, only record frequencies as high as 50 Hz (i.e., 100 Hz sampling rate).

[0134] Data Analysis and Validation In this exemplary study, we focused on z-axis acceleration data since the important information about scratching mainly pertains to out-of-plane skin vibrations. The raw z-axis acceleration data was processed through a signal processing pipeline as illustrated in Figure 4C. In the first step, the raw data was passed to a preprocessing block to remove baseline fluctuations in the data caused by slow and large-scale movements. Thus, the low-frequency baseline fluctuations were removed from the signal processing pipeline, and more intrinsic high-frequency signal components could be better emphasized. The preprocessed z-axis acceleration data was then segmented into several frames by sliding a 1-s window with overlap. For training, 90% overlap was used, and for testing, 50% overlap was used. Each frame was assigned a label of 0, indicating non-scratching activity, if less than 50% of the data points in a given frame were labeled as scratching activity. Otherwise, the frame was assigned a label of 1, indicating scratching activity. From each frame, a set of features, shown in Figure 5C, was extracted, and these features were used to train and validate a random forest (RF) classifier.

[0135] Both random forest (RF) classifiers and leave-one-out cross-validation (LOSO-CV) are used in the exemplary study for algorithm validation. Cross-validation (CV) is frequently used in many machine learning applications because it provides a simple and effective way to evaluate classification models. In CV, the number of folds is a parameter that specifies how the data is divided into training and test sets. LOSO-CV is used to evaluate the performance of the model because it is known to work well on real-world data. After the model is optimized on the training dataset, we evaluated the model performance in cross-study validation, where the trained model is applied to a clinical validation dataset collected in nature. Cross-study validation validates the generalizability of the algorithm since the training and validation datasets were collected independently.

[0136] The RF classifier allows for the automatic detection of scratching activities. The training dataset includes scratching data from 15 different body positions and 6 types of non-scratching activities collected from 10 healthy subjects, as shown in Figure 4A-B. The results of LOSO-CV applied to the training dataset form the basis for the optimization of the RF classifier. In this process, a dataset from a single participant serves as the test dataset, while the remaining data allow training. Repeating this process across all participants allows each participant's data to be used for testing only once. The classification performance corresponds to the average from all iterations of validation. Unlike the k-fold CV approach, LOSO-CV prevents the classifier from learning confounding relationships in the subject's dataset with unrealistically high classification accuracy.

[0137] The RF model can be optimized by comparing the results of LOSO-CV. Parameters for this optimization include the number of trees in the RF classifier and the minimum required scratch duration.

[0138] The RF classifier is a meta-estimator that fits a number of decision tree classifiers on different subsamples of the dataset. The choice of the number of trees in the forest strikes the right balance between model complexity and performance. Increasing the number increases training time and even the potential for overfitting. Therefore, the optimization focuses on examining different numbers of estimators, from 10 to 200 trees, and comparing the performance of LOSO-CV. The classification performance reaches a maximum with around 30 decision trees.

[0139] The minimum required scratch (MRS) period is another optimization parameter. In the signal processing pipeline, as shown in Figure 4C, the RF classifier returns a frame-level prediction for 1 second. The output is then clustered using density-based spatial clustering for noisy applications (DBSCAN). The goal of DBSCAN is to link two closely-spaced scratch frames as a single scratch event.

[0140] In DBSCAN, two important parameters eps and min_samples need to be optimized, which define the clustering rules. The parameter eps specifies the maximum step size that a cluster can take to determine whether a sample (a frame in this study) belongs to a cluster. For example, an eps of 5 frames means that two frames that are 5 frames or less away from each other can be merged as a single cluster. In terms of temporal distance, this corresponds to 1 second frame size and 3 seconds with 50% frame overlap in the sliding window analysis. The parameter min_samples determines the minimum samples required to form a cluster. For example, a min_samples of 5 frames means that a cluster contains 5 or more samples (5 frames in this application). A min_samples of 5 frames means that for the system to detect a series of frames as scratching activity, there must be at least 5 consecutive scratch frames, which corresponds to 3 seconds assuming a 50% overlap between the frames. To find the optimal values, the values ​​of eps and min_samples were varied from 2 frames (1.5 seconds) to 16 frames (8.5 seconds) in 1-frame increments for both the training and validation datasets, as shown in Figure 10. These results suggest that an eps of 10 frames (5.5 seconds) and a min_samples of 8 frames (4.5 seconds) achieve the best performance. This means that a minimum of 4.5 seconds of consecutive scratching is required to form a scratching activity. The algorithm can detect two consecutive scratching frames separated by less than 5.5 seconds of non-scratching activity as the same cluster. In simple terms, a person must scratch continuously for at least 4.5 seconds for the algorithm to cluster a scratch as a scratching event, and a person is allowed to engage in non-scratching activity for up to 5.5 seconds without the algorithm separating the scratch from a previously identified scratching event, as shown in Figure 10.When min_samples was reduced from 8 frames (4.5 seconds) to 3 frames (1.5 seconds), the F1 score remained at 90%, but the specificity dropped to 80%, as shown in Figure 10. This analysis indicates that there is a trade-off between scratch sensitivity and specificity.

[0141] MRS defines the minimum number of frames that form a cluster of scratch events and corresponds to the parameter min_samples. As shown in Figure 10, MRS values ​​from 1.5 to 8.5 seconds result in different classification performances using LOSO-CV. Increasing the MRS duration up to 4.5 seconds improves performance, then remains constant for values ​​up to 8.5 seconds. Improved performance can be achieved with further increases in MRS, but with diminishing returns. In the following analysis, MRS is defined as at least 4.5 seconds.

[0142] The optimized RF classifier is applied to data collected from a single subject performing various scratching (head, arms, abdomen, knees, and legs) and non-scratching activities (simulating scratching in the air, waving hands, composing a text on a phone, typing on a keyboard, and clicking a mouse). The corresponding time series z-axis acceleration data is in Figure 4D and the associated classification results are in Figure 4E.

[0143] With the optimized RF classifier (30 trees) and MRS duration (4.5 seconds), LOSO-CV can be applied to the healthy subject training dataset to evaluate the performance. The results show that the average precision is 94.4% and the average sensitivity is 87.8%. The overall accuracy is 89.1%. Table 1 summarizes the overall performance.

[0144] The RF classifier trained on the algorithm validation dataset can be used directly on the patient dataset from the clinical validation study. The evaluation reveals a mean accuracy of 82.5%, a mean sensitivity of 84.3%, and an overall accuracy of 99.0%, with a sensitivity of 84.3% and a specificity of 99.3%, as shown in Table 1. The clinical study included manually labeled datasets from all 46 nights with the ADAM sensor worn on each subject's dominant hand.

[0145] [Table 1]

[0146] Clearly, our classification results are a significant improvement over most of the prior studies on automatic scratch detection. Table 2 summarizes related studies that were rigorously validated with video recording data. Figure 14 is a screenshot of a video illustrating an AD pediatric patient wearing an ADAM sensor scratching. The IR camera shows the scratching behavior that is then coded. Note that the scratching intensity varies from the beginning to the end of the scratching event, reflected as a strict scratching motion in the IR video and as a large amplitude signal in the ADAM sensor data.

[0147] [Table 2]

[0148] The objective classifier optimized in this study was the RF classifier, but the performance of other classifiers such as logistic regression and gradient boosting were calculated for comparison using the training data and are listed in Table 3. The logistic regression and gradient boosting classifiers showed low sensitivity. The gradient boosting classifier had low sensitivity (67%), but it showed the highest specificity. We observed the best sensitivity, F1 and accuracy with the RF classifier, while the best specificity was observed with the gradient boosting classifier.

[0149] [Table 3]

[0150] Inter-rater reliability In the clinical validation study, infrared (IR) videos were used as ground truth. Annotation was performed based on visual inspection and human judgment. To assess the inter-rater reliability of the annotations, annotation was performed again by different researchers (EA, HC) on data from three patients in the validation dataset. Labels from the two researchers were compared using Fleiss's kappa coefficient. The average kappa was 0.88. According to Landis and Koch, annotations with such kappa values ​​are considered to be in good agreement.

[0151] Essay Pruritus is one of the most common clinical disease symptoms affecting both children and adults across cutaneous and non-cutaneous pathologies. Despite the prevalence of pruritus and its serious impact on quality of life, there are no accurate and convenient methods to fully characterize pruritus and the scratching behavior associated with pruritus for either clinical practice or home use. Techniques that can accurately and reliably assess pruritus serve as vital tools in developing drugs, quantifying pruritus severity, and monitoring treatment responses.

[0152] Video monitoring with manual labeling is considered the gold standard as it allows for accurate identification and quantification of scratching behavior. However, this technique is laborious and difficult to automate. These drawbacks, combined with various associated privacy concerns, make this approach impractical for routine use. Wrist-fixed accelerometers represent a common tool for collecting objective data in a wide range of clinical studies, with sufficient sensitivity but low specificity. Some studies have questioned the clinical applicability of behavioral recording data, due in part to poor correlation with AD disease severity and quality of life. New approaches using smartwatches in combination with machine learning algorithms are superior to simple behavioral recording-based methods. However, these systems are unable to detect finger-only scratching, and hand waving and related gestures are misclassified as scratching (Figure 3A-B). As an example of an unconventional approach, Noro et al. developed a wristwatch-type sound detection system that measures scratching sounds through bone conduction. While accuracy can approach that of video analytics, microphone-based systems carry significant privacy concerns and suffer from performance degradation due to ambient noise.

[0153] The ADAM sensor and scratching algorithm reported here offer important advantages. First, the soft flexibility of the device allows it to conformably fit a wide range of patient hands, and has been validated to remain soft, but firmly in contact during a range of natural activities, from children as young as 4 years old to fully grown adults. This aspect of the ADAM sensor makes it particularly suitable for investigating long-term changes in pathological symptoms such as nocturnal pruritus. For example, with the ADAM sensor, we observed significant night-to-night variations in scratching in a subset of AD patients who wore the sensor for extended periods (10 days or more). This means that it will likely be necessary to average scratching duration over a more sustained period of wear rather than overnight (Figure 11). Second, the system captures both low- and high-frequency acoustic-mechanical signatures associated with scratching. The “acoustic” signal generated by scratching can be effectively captured without the need for a microphone, ensuring patient privacy and allowing performance in noisy environments without compromising signal quality. Finally, the system is waterproof, wireless, rechargeable, and has a continuous operating efficiency of 7 days, which increases ease of use and reduces user burden in both clinical practice and research studies.

[0154] However, several limitations must be noted. First, the clinical validation study reported here involves only n=11 subjects, due in part to the high degree of manual effort required to assess IR camera capture times. However, this represents one of the largest cohorts studied to date on scratching, given the time intensity required for data labeling (Table 2). Future work will include additional patient testing with even younger patients than those studied here, across a broad range of clinical conditions beyond AD. Nevertheless, the findings here clearly demonstrate that training healthy subjects to perform voluntary scratching activities can be successfully applied to clinical data in an exclusively pediatric population affected by AD. These results, derived from unique distinguishing features in the data from the ADAM device, also suggest a remarkable level of generalizability and even the opportunity to develop algorithms tailored to patient populations. Second, all scratching measurements described here are focused exclusively on the subjects' dominant hand. Adding sensors on the contralateral hand would provide additional information. Previous studies have shown no difference in scratching tendency between dominant and nondominant hands, suggesting that a single sensor would not cause significant information loss. Third, the clinical validation study has focused primarily on AD patients only. Future studies will explore application to other conditions where pruritus is a primary symptom. As a final comment, AD affects sleep to a significant degree, especially in children. The ADAM sensor itself can repeat sleep stages versus polysomnography, the gold standard method for assessing sleep. An immediate opportunity for additional research lies in using two time-synchronized ADAM sensors to assess both scratching and sleep concomitantly.

[0155] conclusion This exemplary study reports a unique, hand-worn, soft, flexible sensor and data analysis pipeline for detecting scratching activity using high-frequency triaxial accelerometer data that captures features associated with both movement and acoustic-mechanical signals. From a training study conducted with 10 participants, scratching activity detection has an overall accuracy of 89.1%, sensitivity of 87.8%, and specificity of 88.1%. A clinical validation study with 11 participants reveals an overall accuracy of 99.0%, sensitivity of 84.3%, and specificity of 99.3%, even in a home natural environment. The small, flexible properties and reusability of the ADAM sensor are key features that enable its deployment in both research and clinical care settings. The promising results presented herein highlight its utility as a powerful methodological tool with the potential to accurately capture scratching activity not only in patients with atopic dermatitis, but also in patients suffering from systemic pruritus disorders.

[0156] The invention is, among other things, a non-invasive technique that objectively quantifies scratching behavior via a soft, flexible, wireless sensor that captures the acoustic-mechanical signature of scratching from the dorsum of the hand. Machine learning algorithms validated with data collected from healthy subjects (n=10) show superior performance compared to smartwatch-based approaches. Clinical validation in a cohort of mostly pediatric patients (n=11) suffering from moderate to severe atopic dermatitis included 46 nights of sleep totaling approximately 378.4 hours. Data show approximately 99.0% accuracy (approximately 84.3% sensitivity, approximately 99.3% specificity) versus visual observation. This study suggests a wide range of capabilities relevant for applications ranging from evaluating the efficacy of drugs for pruritus-causing conditions to monitoring disease severity and treatment response.

[0157] The foregoing description of exemplary embodiments of the invention has been presented for purposes of illustration and description only, and is not intended to be exhaustive or to limit the invention to the precise form disclosed. Many modifications and variations are possible in light of the above teachings.

[0158] The embodiments have been chosen and described in order to explain the principles of the technology and its practical application, thereby enabling those skilled in the art to utilize the invention and its various embodiments with various modifications as suited to the particular application contemplated. Alternative embodiments will become apparent to those skilled in the art to which the invention pertains without departing from its spirit and scope. Thus, the scope of the invention is defined by the appended claims rather than by the foregoing description and the exemplary embodiments described therein.

[0159] Several references, which may include patents, patent applications, and various publications, are cited and described in the description of the present invention. The citation and / or discussion of such references is provided solely for clarity of the description of the present invention, and is not an admission that such references are "prior art" to the inventions described herein. All references cited and discussed herein are incorporated by reference in their entirety and to the same extent as if each reference was individually incorporated by reference.

[0160] (References) [1] G. Yosipovitch, MW Greaves and M. Schmelz, Itch. Lancet. 361, 690-694 (2003). [2] A. Bathe, E. Weisshaar, U. Matteme, Chronic pruritus-more than a symptom: a qualitative investigation into patients’ subjective illness perceptions. J Adv. Nurs. 69, 316-326 (2013). [3] M. Shive, E. Linos, T. Berger, M. Wehner, M.-M. Chren, Itch as a patient-reported symptom in ambulatory care visits in the United States. J. Am. Acad. Dermatol. 69, 550- 556 (2013). [4] T. E. Shaw, G. P. Currie, C. W. Koudelka, E. L. Simpson, Eczema prevalence in the United States: data from the 2003 National Survey of Children's Health. J. Invest. Dermatol. 131, 67-73 (2011). [5] S. L. Chamlin, C. L. Mattson, I. J. Frieden, M. L. Williams, A. J. Mancini, D. Celia, M - M. Chren, The price of pruritus: sleep disturbance and cosleeping in atopic dermatitis. Arch. Pediatr. Adolesc. Med. 159, 745-750 (2005). [6] K. Hon, T. Leung, K. Wong, C. Chow, A. Chuh, P. Ng, Does age or gender influence quality of life in children with atopic dermatitis? Clin. Exp. Dermatol. 33, 705-709 (2008). [7] D. Camfferman, J. D. Kennedy, M. Gold, A. J. Martin, K. Lushington, Eczema and sleep and its relationship to daytime functioning in children. Sleep Med. Rev. 14, 359-369 (2010). [8] R. Sack, J. Hanifin, Scratching below the surface of sleep and itch. Sleep Med. Rev. 14, 349-350 (2010). [9] A. B. Fishbein, O. Vitaterna, I. M. Haugh, A. A. Bavishi, P. C. Zee, F. W. Turek, S. H. Sheldon, J. I. Silverberg, A. S. Paller, Nocturnal eczema: review of sleep and circadian rhythms in children with atopic dermatitis and future research directions. J. Allergy Clin. Immunol. 136, 1170-1177 (2015).

[10] A. B. Fishbein, K. Mueller, L. Kruse, P. Boor, S. Sheldon, P. Zee, A. S. Paller, Sleep disturbance in children with moderate / severe atopic dermatitis: a case-control study. J. Am. Acad. Dermatol. 78, 336-341 (2018).

[11] J. I. Silverberg, N. K. Garg, A. S. Paller, A. B. Fishbein, P. C. Zee, Sleep disturbances in adults with eczema are associated with impaired overall health: a US population-based study. J. Invest. Dermatol. 135, 56-66 (2015).

[12] M. K. LeBourgeois, D. C. Dean, S. C. Deoni, M. Kohler, S. Kurth, A simple sleep EEG marker in childhood predicts brain myelin 3.5 years later. Neuroimage. 199, 342-350 (2019).

[13] G. Kiebert, S. V. Sorensen, D. Revicki, S. C. Fagan, J. J. Doyle, J. Cohen, D. Fivenson, "Atopic dermatitis is associated with a decrement in health-related quality of life. Int. J. Dermatol. 41, 151-158 (2002).

[14] C. S. Murray, J. L. Rees, Are subjective accounts of itch to be relied on? The lack of relation between visual analogue itch scores and actigraphic measures of scratch. Acta Derm. Vener. 91, 18-23 (2011).

[15] A. Moreau, P. Anderer, M. Ross, A. Cemy, T. H. Almazan, B. Peterson, Detection of nocturnal scratching movements in patients with atopic dermatitis using accelerometers and recurrent neural networks. IEEEJ. Biomed. Health Inform. 22, 1011-1018 (2017).

[16] K. Lee, X. Ni, J. Y. Lee, H. Arafa, J. P. David, S. Xu, R. Avila, M. Irie, J. H. Lee, R. L. Easterlin, Mechano-acoustic sensing of physiological processes and body motions via a soft wireless device placed at the suprasternal notch. Nat. Biomed. Eng. 4, 148-158 (2020).

[17] Y. Shao, V. Hayward, Y. Visell, Spatial patterns of cutaneous vibration during whole- hand haptic interactions. PNAS, 113, 4188-4193 (2016).

[18] A. Ikoma, T. Ebata, L. Chantalat, K. Takemura, F. Mizzi, M. Poncet, D. Leclercq, Measurement of nocturnal scratching in patients with pruritus using a smartwatch: initial clinical studies with the itch tracker app. Acta Derm. Verier. 99, 268-273 (2019).

[19] J. Lee, D. Cho, J. Kim, E. Im, J. Bak, K. H. Lee, K. H. Lee, J. Kim, Itchtector: a wearable-based mobile system for managing itching conditions. Proc. CHI Conf. Hum. Factor. Comput. Syst. 893-905 (2017).

[20] K. Benjamin, K. Waterston, M. Russell, O. Schofield, B. Diffey, J. L. Rees, The development of an objective method for measuring scratch in children with atopic dermatitis suitable for clinical use. J. Am. Acad. Dermatol. 50, 33-40 (2004).

[21] M. Wiertlewski, J. Lozada, V. Hayward, The spatial spectrum of tangential skin displacement can encode tactual texture. IEEE Trans. Robot. 27, 461-472 (2011).

[22] M. Wiertlewski, V. Hayward, Mechanical behavior of the fingertip in the range of frequencies and displacements relevant to touch. J. Biomech. 45, 1869-1874 (2012).

[23] B. Delhaye, V. Hayward, P. Lefevre, J.-L. Thonnard, Texture-induced vibrations in the forearm during tactile exploration. Front. Behav. Neurosci. 6, 37 (2012).

[24] S. Saeb, L. Lonini, A. Jayaraman, D. C. Mohr , K. P. Kording, The need to approximate the use-case in clinical machine learning. Gigascience. 6, gix019 (2017).

[25] T. Ebata, H. Aizawa, R. Kamide, An infrared video camera system to observe nocturnal scratching in atopic dermatitis patients. J. Dermatol. 23, 153-155 (1996).

[26] T. Ebata, S. Iwasaki, R. Kamide, M. Niimura, Use of a wrist activity monitor for the measurement of nocturnal scratching in patients with atopic dermatitis. Br. J. Dermatol. 144, 305-309 (2001).

[27] Y. Noro, Y. Omoto, K. Umeda, F. Tanaka, Y. Shiratsuka, T. Yamada, K. Isoda, K. Matsubara, K. Yamanaka, E. C. Gabazza, Novel acoustic evaluation system for scratching behavior in itching dermatitis: rapid and accurate analysis for nocturnal scratching of atopic dermatitis patients. J. Dermatol. 41, 233-238 (2014).

[28] O. Crasbom, H. Sloetjes, E. Auer, P. Wittenburg, Combining video and numeric data in the analysis of sign languages with the ELAN annotation software. 2nd workshop on the representation and processing of sign languages: lexicographic matters and didactic scenarios. 82-87 (2006).

[29] S. Arlot, A. Celisse, A survey of cross-validation procedures for model selection. Statistics surveys. 4, 40-79 (2010).

[30] R. Kohavi, A study of cross-validation and bootstrap for accuracy estimation and model selection. IJCAI. 14, 1137-1145 (1995).

[31] J. L. Fleiss, J. Cohen, B. S. Everitt, Large sample standard errors of kappa and weighted kappa. Psychol. Bull. 72, 323 (1969).

[32] J. R. Landis and G. G. Koch, The measurement of observer agreement for categorical data. Biometrics. 159-174 (1977).

Claims

1. 1. An acousto-mechanical sensor, comprising: At least one inertial measurement unit (IMU) for operatively detecting vibration and movement signatures of physiological processes including scratching or rubbing activity; a microcontroller unit (MCU) coupled to the at least one IMU for receiving data from the at least one IMU and processing the received data; and a wireless communication system for wireless data transmission; the acousto-mechanical sensor is configured to measure both low frequency signals derived from the at least one IMU below about 100 Hz and high frequency signals above about 100 Hz; the acousto-mechanical sensor is skin-mounted to enable coupling of the at least one IMU to the skin to capture both high frequency and low frequency signals; Acousto-mechanical sensors in which high- and low-frequency signals represent the physiological response to itch as scratching or rubbing activity.

2. 2. The acousto-mechanical sensor of claim 1, wherein the plurality of electronic components and the at least one IMU are separated such that the plurality of electronic components are at one end of the acousto-mechanical sensor and the at least one IMU is at an opposite end of the acousto-mechanical sensor.

3. a flexible printed circuit board (fPCB) for supporting and interconnecting the at least one IMU and the plurality of electronic components; 2. The acousto-mechanical sensor of claim 1, wherein the fPCB has a patterned layout including a plurality of freely deformable, serpentine interconnects designed to mechanically decouple the at least one IMU from other components of the acousto-mechanical sensor, the plurality of freely deformable, serpentine interconnects being flexible and stretchable and adapted to interconnect the at least one IMU and the plurality of electronic components.

4. an elastomeric encapsulation layer at least partially surrounding the at least one IMU, the plurality of electronic components, and the fPCB and forming a tissue-facing surface for operative attachment to a living body, the tissue-facing surface of the elastomeric encapsulation layer being configured to conformally conform to a skin surface of the living body; 4. The acousto-mechanical sensor of claim 3, wherein said encapsulation layer is formed from a thin, low modulus silicone elastomer, said elastomeric encapsulation layer being a biocompatible silicone enclosure.

5. a biocompatible hydrogel adhesive for attaching the acousto-mechanical sensor to an anatomical location of a living body; the biocompatible hydrogel adhesive is a medical grade hypoallergenic adhesive; 2. The acousto-mechanical sensor of claim 1, wherein the biocompatible hydrogel adhesive is adapted to operatively conduct signals from the living body to the at least one IMU.

6. The acousto-mechanical sensor of claim 1 , wherein the wireless communication system is configured to transmit an output signal from the acousto-mechanical sensor to an external device, the wireless communication system comprising a Bluetooth low energy (BLE) radio.

7. 2. The acoustic mechanical sensor of claim 1, wherein the plurality of electronic components further comprises a power module coupled to at least one IMU, the wireless communication system, and the MCU for supplying power, the power module comprising a battery, the battery being a rechargeable battery, the power module further comprising a wireless charging module for wirelessly charging the rechargeable battery, and a fault prevention element including a short circuit protection component or circuit to avoid malfunction of the battery.

8. The acousto-mechanical sensor of claim 1 , wherein the at least one IMU comprises a millimeter-scale three-axis accelerometer configured to detect tissue-conducted vibrations and motion associated with scratching activity.

9. the three-axis accelerometer is configured to detect acceleration along an x-axis and a y-axis at a first sample rate and a z-axis at a second sample rate, the x-axis and y-axis being oriented parallel to a skin surface to which the acousto-mechanical sensor is attached and the z-axis being perpendicular to the skin surface, the first sample rate being in a range of approximately 160-240 Hz and the second sample rate being in a range of approximately 1,200-2,000 Hz; Alternatively, the acousto-mechanical sensor of claim 8 , wherein the three-axis accelerometer is configured to have a sampling rate of about 1,600 Hz with 16-bit resolution and a dynamic range of about ±2 g, where g is the acceleration of gravity, 9.8 m / s 2 .

10. When positioned on the dorsum of the hand, the acousto-mechanical sensor is capable of detecting acousto-mechanical signals associated with the scratching activity via a combination of movement and acousto-mechanical signals in a manner that is not affected by ambient noise; or between the metacarpal bones of the second and third fingers, the acousto-mechanical sensor is capable of quantifying the scratching activity initiated not only from wrist and / or arm movements but also from the fingers and fingertips over a wide temporal bandwidth including the high frequency vibrational motion associated with scratching itself.

11. the scratching activity is characterized by first and second types of signals, the first type of signals corresponding to global movements of the hand and the second type of signals overlapping in time with the first type of signals and resulting from subtle vibration impulses generated by movements of the fingertips and fingernails against a contact surface; the first type of signal has a characteristic frequency in the range of a few Hz or less, and the second type of signal has a characteristic frequency over a range of several hundred Hz, and the amplitude decays rapidly with position along the fingers, into the hand, and finally through the wrist and onto the arm; Alternatively, the first and second types of signals are identified by passing the data through a low pass filter and a high pass filter having a cutoff of about 2 Hz.

12. The acousto-mechanical sensor of claim 8, wherein signal features uniquely characterizing the scratching activity are extracted from the output signal of the acousto-mechanical sensor to train and validate machine learning based algorithms for scratch detection and to quantify symptoms of pruritus.

13. configured to execute a scratching algorithm on the detected data from the at least one IMU to obtain signal features that uniquely characterize the scratching activity; or configured to notify a user and / or an expert via tactile / audio / visual means of the acoustic mechanical sensor, a mobile device, and / or a smart wearable device.

14. having a specific geometric polarity for attachment to an anatomical location of the body, and being flexible and conformable to the skin; Or the acousto-mechanical sensor of claim 1 has a low-profile conformable nature that allows the acousto-mechanical sensor to be positioned on the curved surface of the back of the hand and capture scratches regardless of joint articulation, and the scratches are associated with finger movement only, wrist movement only, elbow movement only, shoulder movement only, or a combination thereof.

15. Operatively positioned over key bones, including the metatarsals of the first and second toes, to capture high frequency vibrations specific to scratching; or operatively placed on the dorsum of the hand, fingernail, finger, or wrist to measure scratching movements over any period of time; or in any disease state in which pruritus is a symptom, as a diagnostic, treatment response indicator, clinical trial endpoint, or early warning sign of disease progression; Or an acousto-mechanical sensor as described in claim 1 that can capture both scratch-specific motion and bone-conducted high frequency vibrations, thereby distinguishing scratch motion from true scratching on the skin.

Images