Soft Material-Enabled, Active Wireless, Thin-Film Bioelectronics for Quantitative Diagnostics of Cervical Dystonia

Abstract

Recent advances in flexible materials, nanomanufacturing, and system integration have provided a great opportunity to develop wearable flexible hybrid electronics for human healthcare, diagnostics, and therapeutics. However, existing medical devices still rely on rigid electronics with many wires and separate components, which hinders wireless, comfortable, continuous monitoring of health-related human motions. Here, we introduce advanced materials and system integration technologies that enable a soft, active wireless, thin-film bioelectronics. The low-modulus, highly flexible wearable electronic system incorporates a nanomembrane wireless circuit and functional chip components, enclosed by a soft elastomeric membrane. The bioelectronic system offers a gentle, seamless mounting on the skin, while offering a comfortable, highly sensitive and accurate detection of head movements. We utilize the wireless wearable hybrid system for quantitative diagnostics of cervical dystonia (CD) that is characterized by involuntary abnormal head postures and repetitive head movements, sometimes with neck muscle pain. A set of analytical and experimental studies shows a soft system packaging, hard-soft materials integration, and quantitative assessment of physiological signals detected by the SKINTRONICS. In vivo demonstration, involving ten human subjects, captures the device feasibility for use in CD measurement.

Graphical Abstract

We introduce advanced materials and system integration technologies that enable a soft, active wireless, thin-film bioelectronics. The low-modulus, highly flexible wearable electronic system incorporates a nanomembrane wireless circuit and functional chip components, enclosed by a soft elastomeric membrane. The bioelectronic system offers a gentle, seamless mounting on the skin, while offering a comfortable, highly sensitive and accurate detection of head movements.

In vivo pilot study with human subjects captures the device feasibility for a quantitative evaluation of cervical dystonia.

1. Introduction

Cervical dystonia (CD) is a chronic neurological disorder, with an incidence of 16 per 100,000 people and currently more than 300,000 people in the United States.^{[1, 2]} The disorder is characterized by involuntary contraction of neck and shoulder muscles, leading to repetitive movements and abnormal postures of the head including torticollis (horizontal turning), laterocollis (tilt of the head), retrocollis (extension of the head), and anterocollis (flexion of the head).^[3–5] CD patients suffer from substantial decrease in normal activities of daily living such as cooking, dressing, eating, and driving.^{[6, 7]} While botulinum toxin (BoNT) injection into neck muscles has widely been used to relieve pain and abnormal movements, reaching the ideal dose and muscle pattern can be challenging.^[8–11] The traditional methods for assessment of CD involve clinical rating scales such as the Toronto Western Spasmodic Torticollis Rating-2 (TWSTRS-2).^{[12, 13]} However, these scales are dependent upon the level of expertise of the clinician.^{[1, 11]} Therefore, it is critical to develop a novel quantitative assessment system to estimate the severity and duration of abnormal postures in different planes of movements for CD patients.

Here, we introduce a wearable bioelectronic device that could address the current challenges in quantitative diagnosis of CD. The device has a highly sensitive accelerometer and wireless system for accurately recording of total head movements in the yaw, pitch, and roll planes in a soft, ultrathin, miniaturized, comfortable platform. Prior works using commercial wearable sensors show capabilities of motion recording of daily activities at various body locations.^[14–17] However, these devices based on rigid sensors/circuits and multiple electronic components require a cumbersome device mounting with adhesives and straps on the body, which interrupts accurate measures of body movements.^[18–20] Thus, they are not applicable for CD detection, requiring a sensitive detection of subtle motions.

On the other hand, the newly developed, skin-like bioelectronics (referred to as ‘SKINTRONICS’) in this work offers a quantification of multiple CD symptoms and severity levels. The soft-membrane construction allows a gentle and seamless mounting on the forehead of a patient with negligible effects on natural head motions while supplying sufficient and reliable digital signals throughout the diagnostic process of CD. This digital assessment from the SKINTRONICS also provides a quantitative detection of retrocollis, anterocollis and dystonic tremor that are not currently included in a clinical manual evaluation of TWSTRS-2 because of its poor inter-rater reliability and low item-to-total correlation. To validate the device accuracy, a direct comparison between SKINTRONICS-based assessment and TWSTRS-2 based on manual evaluation is provided from a clinical study with 5 patients and 5 healthy subjects. Collective results indicate the new class of soft technology platform for a wireless, accurate monitoring of abnormal head movements may provide a more effective and safe measurement for improved CD healthcare.

2. Results and Discussion

2.1. Overview of the system architecture

Figure 1a illustrates a highly flexible, wireless SKINTRONICS that includes multi-layers of chip components, interconnects, dielectric layers, base PI layer, and soft encapsulants. The integrated circuit, powered by a miniaturized lithium polymer battery, consists of a highly sensitive accelerometer, Bluetooth module, ceramic antenna, and power management feature in a soft elastomeric enclosure, which is used for a fully portable, wireless detection of head movements. Details of the circuit design and chip components can be found in Figure S1 (Supporting Information) and Table S1 (Supporting Information). The ultralight (< 10 g), low-profile (< 2 mm) device offers a comfortable, seamless wearability on the skin without the use of wires or a separate adhesives (Figure 1b). Figure 1c captures a long-range wireless capability of the fabricated SKINTRONICS, gently mounted on the forehead, which communicates with a mobile device. A home-made Android application enables a real-time monitoring of a head movement in 3-axis acceleration, along with an automatic data saving for post analysis. A study of CD must detect abnormality based on head postures in x, y, and z that are related to roll, pitch, and yaw, respectively. Different types of such movements are detected by the 3-axis accelerometer in the SKINTRONICS on the forehead (Figure 1d). In clinical terminology, the symptom associated with head rotation to the right or left (yaw) is called torticollis and is the most common in CD. Subjects who have a problem with laterally tilting the head (roll) have laterocollis. Retrocollis and anterocollis are related to extension and flexion of the head (pitch).^[21] In subjects with CD, any combination of these movement abnormalities may occur. Figure 1e captures an example of the device sensitivity in detection of a head movement. There is no significant deflection from the baseline when the head is still. In contrast, the bottom graph indicates five repetitive postures with fixed pitch, corresponding to a head flexion in x and z axes. Collectively, the analysis of acceleration data measured by the wireless SKINTRONICS offers quantification of abnormal head movements from CD patients.

Figure 1.

Overview of SKINTRONICS architecture. (a) Schematic illustration of a multi-layered, soft material-enabled SKINTRONICS. (b) Photos showing an extreme flexibility of the device with different radii and on a finger. (c) Photo of capturing a long-range (> 10 m), active wireless detection of cervical dystonia via Bluetooth. (d) 3D diagram of the basic forms of cervical dystonia. Rotation around x, y, and z are related to roll, pitch, and yaw, respectively. Head rotation to the right or left in z axis (Yaw): Torticollis, Head lateral shift in x axis (Roll): Laterocollis, Head extension backward and forward in y (Pitch): Retrocollis (backward) and Anterocollis (forward). (e) Examples of a measured acceleration data from the device without movement (top graph) and with movement (bottom graph).

2.2. Material characterization of SKINTRONICS

Figure 2 shows the characterization results of materials used in the SKINTRONICS. One of the key features for a wearable device is to maintain a seamless contact to the skin that experiences multi-modal, dynamic movements. We studied a peeling force (adhesion) and Young’s modulus of four types of materials for the skin mounting, including polydimethylsiloxane (PDMS, Dow Sylgard 184), Solaris (Smooth-On), Ecoflex (Smooth-On), and Snap gel patch (Biopac). As shown in Figure 2a, a membrane’s peeling force from the human skin is measured by a digital force gauge (M5–5, Mark-10). The calculated adhesive energy for four materials (Figure 2b) shows that the commercial gel patch has the highest energy compared to other silicone elastomers; among them, Ecoflex has the highest adhesion. Another material’s property, Young’s modulus, was measured by monitoring stress variation according to applied strain (Figure 2c and 2d); the gel patch has the biggest modulus compared to others. The quantitative values of adhesive energy and Young’s modulus are summarized in Figure 2e. Based on this study, we chose the low-modulus Ecoflex as the ideal substrate and encapsulant to integrate the flexible electronics package since it could offer enough adhesion, while reducing the mechanical loading to the skin interface. Even though the commercial gel patch has the highest adhesion, it is too aggressive to exfoliate the stratum corneum layer on the skin, which causes skin rash and pain. ^[22] In addition, the high-modulus material causes discomfort when mounted on the soft skin, while the Ecoflex encapsulant improves the mechanical bendability upon deformation by protecting the active components.^[23] In addition, we investigated the fabricated multi-layer SKINTRONICS via scanning electron microscopy (SEM). Figure 2f captures the cross-sectional structure around vertical interconnect access (VIA) connecting two Cu layers (chip interconnect layer and ground plane layer) of the SKINTRONICS. The vertical, high-resolution SEM image shows a seamless connection between two layers by coating the 2^nd Cu membrane along with the etched middle PI wall (Figure 2g), resulting in a fully functional, flexible SKINTRONICS.

Figure 2.

Material characterization of the SKINTRONICS. (a) Photo showing a quantitative measurement of an adhesion energy of skin-wearable materials via a force meter. (b) Result of measured adhesion energy of four types of membrane materials. (c) Photo capturing a tensile loading for a modulus measurement. (d) Stress-strain curve to calculated Young’s modulus. (e) Young’s moduli, and peel force of 4 different samples. (f-g) Cross-sectional SEM images of a fabricated SKINTRONICS to investigate a seamless connection of VIA with two Cu layers.

2.3. Validation of the mechanical flexibility and performance of the SKINTRONICS

Figure 3 summarizes the result of the mechanical reliability of the flexible device. A device on a soft membrane is wrapped around a curved glass edge to apply a cyclic bending (180 degrees at the radius of curvature of 1.5 mm) up to 100 cycles (Figure 3a–c). The device performance during the test is monitored by recording the acceleration in y- and z-axes (Figure 3d), which could detect any mechanical failure of the device in signal loss. The results in Figure 3e–f show that the SKINTRONICS has negligible change in acceleration, even with 100 bending cycles, indicating the mechanical flexibility in the applications on the skin with time-dynamic deformation. In addition, we validated the device performance via a direct comparison with a state-of-the-art, wearable motion detector (The Opal, APDM Wearable Technologies). Figure 4 shows the one-to-one comparison of the signal detection during a subject head motion. The corresponding acceleration data, measured from the subject’s forehead, are shown in Figure 4a–b. A regression plot in Figure 4c demonstrates that the SKINTRONICS has a similar performance to the state-of-the-art motion detector as measured in a coefficient of determination of R² = 0.9750. For a continuous device validation, both devices were used in a clinical study to measure a series of head motions. Overall, this experimental study captures the novelty of the soft electronic system in the mechanical reliability and sensitivity, along with a simple, comfortable mounting on the forehead without a strap.

Figure 3.

Mechanical behaviors of SKINTRONICS. (a-c) A sequential photo of the highly flexible SKINTRONICS wrapped around a curved glass edge, showing 0^o, 90^o, and 180^o bending (radius of curvature: 1.5 mm). (d) Acceleration data for 100 cycles of bending. (e,f) Acceleration data from a head movement in y and z axes before the first cycle (e) and after 100 cycles (f).

Figure 4.

Validation of the device performance compared with a commercial device. (a) A representative acceleration data from a commercial hair band-embedded device. (b) Direct comparison of the acceleration data from the same movement, measured by the SKINTRONICS. (c) Correlation of data obtained from the commercial device vs. SKINTRONICS.

2.4. Signal processing for classification of CD symptoms

We designed a clinical study protocol that measured various head motions, including holding, turning, tilting, and extension/flexion (details of the protocol in Table S2, Supporting Information). Each posture was repeated five times, except holding that maintains the head still for 60 seconds. The acquired signal was filtered by Butterworth bandpass filter at 0.5 – 5 Hz for distinguishable features of target signals. Figure 5 summarizes the signal processing sequence from the measured acceleration to the distance change. The final travel distance of a head motion is calculated by following equations:

$$v_x(\text{t})={\int }_0^{t1}{a}_x(t)dt,v_y(\text{t})={\int }_0^{t1}{a}_y(t)dt,v_z(\text{t})={\int }_0^{t1}{a}_z(t)dt,$$ (1)

$$v(t)={(v_x^2+v_y^2+v_z^2)}^{0.5}$$ (2)

$$Distance={\int }_{t1}^{t2}v(t)dt$$ (3)

where t₁ and t₂ are the beginning and end of the peak.

Figure 5.

Signal processing sequence of raw acceleration data. (a) 3-axis acceleration data after Butterworth filtering. (b) Velocity data after integration of (a). (c) Root-mean-square (RMS) speed from the vector in (b); the motion is repeated five times and the magnified graph (right) shows the repeating motion calculated by the area of the RMS speed.

The filtered acceleration data (Figure 5a) is converted to velocity by using an integral calculus (Eq. 1) in all directions (Figure 5b). Note that magnified graphs on the right in Figure 5a–b are the step for flexion of head without the movement in y-axis. The velocity plot is then converted to scalar values of root-mean-square speed (Figure 5c) by taking the magnitude (Eq. 2). Lastly, the area of RMS speed in Figure 5c provides the information on a head’s travel distance that is directly associated with head movement abnormalities in CD. For example, the distance for a head flexion is calculated as gray peaks in Figure 5c (right), while the white zones indicate return head motions to the neutral position. To classify four directions of torticollis, laterocollis, retrocollis, and anterocollis, we separately calculated the distance of each motion.

2.5. Comparison of head motions between healthy subjects and CD patients

CD refers to a neurological disorder, caused by excessive pulling and twisting of the neck muscles, which results in limited head movements.^{[24, 25]} For quantitative detection of CD, we designed a clinical protocol (Table S3, Supporting Information) and developed the signal analysis method with 10 subjects (5 heathy people and 5 CD subjects). Figure 6 summarizes the quantitative results from different head motions, associated with four directions of movement. Figure 6a–b illustrate the discrepancy of the movement range between healthy subjects and patients, following the instruction in the protocol. Figure 6c supports that patients (red bars) have limited capability in head travel distances (details in Table S4, Supporting Information). The distance data of 5 patients were extracted from the velocity data in Figure S2 (Supporting Information). Four CD symptoms were measured, including 1) torticollis (movement around z axis); turning the head to right or left, 2) laterocollis (movement around x axis); tilting the head to right or left, 3) retrocollis (movement around y axis); head up, and 4) anterocollis (movement around x axis); head down.

Figure 6.

Comparison of different head postures between healthy subjects and patients. (a, b) The human testing protocol involves four types of motions to determine four symptoms and severity; the travel distance of each motion can be easily differentiated between a healthy subject (a) and a CD patient (b). (c) Summary of the comparison of the averaged travel distance and standard deviation (error bars; n=5) for four symptoms.

2.6. In vivo demonstration of the device performance in CD quantification

We demonstrated the device performance via in vivo clinical pilot study. As a result, the SKINTRONICS offered a quantitative, digital scaling of CD symptoms and severity (Figure 7). By following the clinical rating scale (TWSTRS-2), we colorized the severity levels. Based on the recorded acceleration data, subjects are categorized in torticollis (Figure 7a), laterocollis (Figure 7b), and retrocollis/anterocollis (Figure 7c) via the distance data (details in Table S4, Supporting Information). Figure 7d shows the severity of dystonic tremor, which is measured as an uncontrollable rhythmic shaking or oscillation of the head. The existing clinical scaling is not able to quantify tremor due to low accuracy and the lack of standardized definitions.^[26] We utilized a digital scaling method to express signals in a frequency domain via fast Fourier transformation of the baseline motion. The corresponding output is shown in Figure S3 (Supporting Information). While no peak is detected from the healthy subject (Figure S3a, Supporting Information), the oscillation is clearly found at the frequency of 4 – 5 Hz from the patient #1, 2, and 5 (Figure S3b, Supporting Information). The severity of tremor is classified by calculating the area of Gaussian fit.

Figure 7.

Quantitative evaluation of the distance-based CD severity from five patients. Abnormal CD symptoms includes torticollis (a), laterocollis (b), retrocollis-backward / anterocollis-forward (c), and dystonic tremor (d). Five different colors represent the severity levels from normal to slight, mild, moderate, and severe.

Table 1 summarizes the severity scaling of five patients in five CD symptoms, while comparing the outcomes between the SKINTRONICS-based digital assessment and clinical manual evaluation. The severity scale factor ranges from 4 highest to 0 lowest. Overall, the digital assessment, provided by the SKINTRONICS, shows a consistent, quantitative scaling of severity levels in all symptoms. More importantly, the device could successfully detect retrocollis/anterocollis and dystonic tremor that cannot be evaluated by the existing manual evaluation in TWSTRS-2. Two real-time recorded video clips show a clear visual difference of the head motions from a healthy subject (Video S1, Supporting Information) and patient (Video S2, Supporting Information) who are following the given testing protocol; this patient was diagnosed with all symptoms including torticollis, laterocollis, retrocollis/anterocollis, and dystonic tremor.

Table 1.

Comparison of the severity levels (4 highest – 0 lowest) between quantitative digital assessment from the SKINTRONICS and clinical manual evaluation from a clinician.

Patient No.	Torticollis		Laterocollis		Retrocollis/Anterocollis		Dystonic tremor
	Digital assessment	Clinical manual evaluation	Digital assessment	Clinical manual evaluation	Digital assessment	Clinical manual evaluation*	Digital assessment	Clinical manual evaluation*
#1	1	1	0	1	0	-	3	-
#2	3	1	3	1	1	-	2	-
#3	1	1	3	3	1	-	0	-
#4	0	0	1	1	0	-	0	-
#5	4	3	3	3	2	-	2	-

^*not measurable in the existing manual evaluation sheet.

3. Conclusion

We have introduced the first demonstration of a fully portable, wireless, wearable SKINTRONICS for quantification of CD. The skin-friendly system enables quantitative digital scaling of five CD symptoms and severity levels from five patients, which shows a more consistent rating compared to a manual evaluation by a highly trained clinician. A set of experimental study demonstrates the mechanical reliability of the highly flexible and wearable device, along with the device sensitivity in the measurement of 3-axis acceleration on the forehead. Collectively, the wearable bioelectronics system, incorporating a flexible membrane circuit, Bluetooth-based wireless telemetry, and soft material packaging, will serve as a non-invasive, accurate tool for quantification of CD and potentially also dystonia in other body regions.

4. Experimental Section

Fabrication of the SKINTRONICS:

The overall fabrication process of the SKINTRONICS followed the combined approaches of photolithography, materials transfer printing, and hard-soft materials integration.^{[22, 27–29]} The overview of the SKINTRONICS is shown in Fig. 1. A set of conventional microfabrication steps was utilized to construct a multi-layered metal/polymer composite, incorporating polyimide (PI) - copper (Cu) – PI – Cu - PI with VIAs connecting two Cu layers. Those layers were patterned on a 100 nm-thick sacrificial layer of polymethylmethacrylate (PMMA) on a Si wafer. 600 nm-thick, 1^st Cu layer was deposited and patterned by sputtering, photolithography, and wet etching on a supporting layer of PI. The middle dielectric layer of PI (~ 7 µm in thickness) was spin-coated and etched to create VIAs. Sputtering of 1 µm-thick, 2^nd Cu and spin-coating of PI (1.5 µm in thickness) were followed before transfer printing. Chip components were soldered on the completed thin-film structure. Afterwards, the electronics were transferred onto a 500 µm-thick elastomer and completely encapsulated. Details of the entire fabrication steps are shown in Supplementary Figure S4, Note S1, and Table S5.

Clinical rating scale with TWSTRS-2:

In this work, we compared the outputs (quantitative scaling of CD) of the SKINTRONICS with the clinical rating scale, TWSTRS-2.^[12] The clinical measurement classifies the severity of CD as the following five levels, including normal (none), slight (less than 25% full range), mild (25% to less than 50% full range), moderate (50% to less than 75% full range), and severe (75% or greater of full range). Details of the protocol for a clinical study in this work and the example of rating items based on TWSTRS-2 are listed in Table S2 and S5, respectively.

In vivo study with human subjects:

A pilot study in this work involved 10 subjects with ages from 18 to 60. The study was conducted by following the approved IRB #00024699 ‘Clinical Studies of Dystonia and Related Disorders’ at Emory University School of Medicine. Prior to the study, all subjects agreed with the study procedures and provided signed consent forms.