Technical Evaluation of Earswitch Phase A (Earswitch)

The goal of this study is to test the Earswitch device in both participants with and without assistive technology (AT) needs. The main aim is to determine how robust the Earswitch is in detecting voluntary eardrum movements (known as ear rumbling) and see how well the Earswitch fits with other assistive technology. To do this, participants will be asked to wear the Earswitch device and voluntarily contract their TT to complete a series of tasks shown on screen and/or explained audibly. Participants will be asked to complete questionnaires to understand their opinions about the effectiveness, usability, and comfort of both their current device (if applicable) and the Earswitch. This feedback will be invaluable in developing the Earswitch.

The overall aim of this study is to evaluate the robustness of the EarSwitch technology (hardware and software) in detecting voluntary eardrum movements in people with assistive needs. This is a key step to ensure a robust design informed by patients' needs. A secondary aim is to gather information from assistive device users about how well they can perform voluntary eardrum movements, as this knowledge will be used to inform what types of interfaces are appropriate for the Earswitch. Healthy participants will also be recruited as a separate cohort, to maximise the amount of data collected which can be used to validate and inform future designs. The methodology will be the same across both cohorts with the same equipment (Earswitch, second screen, and laptop) being used, however the healthy participants will not be able to answer all questions in PART 1. Healthy participants will be invited into a University lab to undertake the study, whereas participants with assistive technology needs will be given the option of a home visit. As the same equipment will be used, only the environment (i.e., home vs lab) will change which is not anticipate to affect the study data. At the start of the study a signal will be agreed with the participant (and carer when required) so that the participant can inform their carer and/or the experimenter if they wish to stop the study for any reason. The proposed study will take place in a single session and includes four main parts: PART 1 - Co-design Questionnaires. The specific aims for this part are to collect data on what assistive technologies people use and gain feedback on current switch technologies to better understand how the Earswitch can be integrated in conjunction with other switch technology. The participants will be asked questions to understand the type of assistive technology (hardware and software) they currently use. This will include their opinions about the effectiveness, usability, and comfort of their current device, measured using the previously validated QUEST2 questionnaire. This process should not take more than 15 minutes, and will be carried out by the researcher, with the support of a carer where necessary. PART 2 - Earswitch insertion. Following this, the participant, with the help of the experimenter or the carer, will select an earbud size which will be inserted into the participant's ear. The participant will then be asked if it is comfortable or whether they would like to try a different size according to the comfort or fit of the earbud. Once the participant is happy with the size of the earbud, the experiment or the carer will install the Earswitch device, which involves putting a small camera into the participant's ear canal and positioning it so that it can see their eardrum. During this procedure, the participant will be asked to let the experimenter or carer know if it is uncomfortable or if they wish to stop. This should be no different to putting a regular ear bud into the participant's ear. Once this is complete, the participant will be asked some questions about the comfort of the device, and how easy they felt the installation process was. Once the Earswitch has been inserted, the experimenter will take a measurement of how far the otoscope is inserted into the ear, by using a ruler to measure the extruding element (with which the depth of the otoscope can be calculated). In addition, the experimenter will take a close-up picture of the Earswitch sitting in the concha to provide visual evidence of the fit (i.e., how the silicone earbud sits in the concha) and to identify any potential pressure points (e.g., on the tragus) that may help explain the device comfort. Part two is scheduled to last 10 minutes. PART 3 - Ear rumbling test. Next, some videos will be recorded by the Earswitch device of the participant's ear drum while they contract the tensor tympani (TT) muscle voluntarily in response to several tasks. The tasks will ask the participant to voluntarily contract the TT muscle in response to different stimuli when requested. These stimuli will be presented to the participant either visually on a screen, audibly through some speakers, or both depending on their sight and hearing. The purpose of this is to see how well the participant can voluntarily contract their TT muscle on demand, and to see how well the Earswitch can recognise this. The tasks will involve the participant contracting their TT muscle voluntarily in response to a presented stimulus, twice in a row, multiple times in a row, and they will be asked to hold the muscle contraction for certain periods of time (1 second, 2 seconds, as long as possible). In addition, the participant's eardrum will be recorded when they perform everyday tasks to make sure that the Earswitch does not accidentally activate when it shouldn't. The everyday tasks will be selected based on the capability of the participant, but will include (at minimum) doing whatever they wish for two minutes, and communicating with their existing communication device fortwo minutes. If possible, the participant will be asked to yawn, move their head, and drink some water. During these tasks the participant will not get any immediate feedback from the system, instead it will be recording the data which will be analysed after they have finished the study. The videos will be recorded using the Earswitch device in the participant's ear, and will therefore only contain anonymous images of their ear drum. This means that no-one will be able to identify the participant from the videos taken as no personally identifiable data will be stored alongside the videos. Finally, the participant will be asked to complete some questionnaires about their experiences with the Earswitch device, including the comfort of the device and how easy it was to use. This part of the study should last about 10minutes.This part of the study should take about 25 minutes. PART 4 - Usability testing (optional). Following this data collection, the participant will be asked if they would like to interact with some interfaces using the Earswitch device. There will be three interfaces which the participant can try which involve text entry using a switch scanning keyboard, a target acquisition task which provides immediate feedback, and a game. The participant can complete these applications in any order they wish, and for as long as they wish. The purpose of this task is to provide an opportunity for the participant to use the Earswitch for real-time interaction using real-world tasks. This will be followed by further questionnaires to understand the participant's perception of interacting with the Earswitch. This part of the study should last about 15-20 minutes. At the end of the study the participant will have the opportunity to provide the research team with their contact details(e.g., an email address or telephone number) in the event that they would like to receive updates about the study results, the development of the Earswitch, or if they would like to participate in future studies involving the Earswitch.

The main research question about the robustness of the Earswitch will be calculated based on the whole participant pool. There is no evidence to suggest that the tensor tympani contraction will physically differ between participants with or without assistive technology needs, and therefore the data is equally valid in determining whether the algorithms accurately detect the tensor tympani contraction.

Participants will wear the Earswitch device and voluntarily contract their TT to complete a series of tasks shown on screen and/or explained audibly. We will also ask participants to complete questionnaires to understand participant opinions about the effectiveness, usability, and comfort of both their current device (if applicable) and the Earswitch.

The Earswitch™ device consists of an earpiece made from silicon or similar bio-compatible material and incorporates a miniature camera. The earpiece sits within the ear canal and is not in direct contact with the eardrum and the camera is not in contact with the ear-canal. The camera detects voluntary movement from the tensor tympani muscle as presented at the eardrum and the Earswitch system translates this movement to an external virtual keypad or similar to allow the patient to communicate nonverbally.

How robust is the Earswitch technology in detecting voluntary eardrum movements? This will be measured by the F-score, Matthews correlation coefficient (MCC), the receiver operating characteristics (ROC) curve, and the Area Under the ROC curve (AUC).

The F-score represents the robustness of the Earswitch in detecting TT contraction and is the harmonic mean of the precision and recall of the binary classifier used to assess whether an ear rumble has taken place. The MCC measures the quality of binary classifications, calculated from the confusion matrix which includes true positives, true negatives, false positives and false negatives. A ROC curve shows the performance of the classification model at all classification thresholds, from which the AUC metric will be calculated providing an aggregate measure of performance across all classification thresholds. AUC ranges from 0-1, with greater values indicating more correct. This gives an overall metric of the model that is invariant to the classification-threshold and can be used to assess different classification models. However, for practical applications of the Earswitch one would want to minimise false positives and hence the individual F-scores and MCC will be calculated.

Quantitative metrics of how well individuals can ear rumble calculated using a custom piece of software which uses the manually labelled data from videos of the participants' ear drum combined with the time logs of the experimental software. These include: (1) Reaction time from when the stimulus was shown until when the tensor tympani contracts, measured in milliseconds, (2) Time between ear rumbles (for double rumbles) measured in milliseconds, (3) Average number of consecutive ear rumbles (for consecutive rumbles) which do not have units (number of ear rumbles), and (4) Average duration of the hold rumbles measured in milliseconds.

: Is there a difference in the accuracy of TT movements between those who require assistive technology and those who do?

Statistical measures (unitless) of a difference between the assistive technology user group and healthy participants who do not require assistive technology for on metrics related to how well users can ear rumble using the outcome measures described for RQ2 (reaction time, predictive time, time between ear rumbles, average number of consecutive ear rumbles, average duration of the hold rumbles). We will use either a paired t-test or Wilcoxon signed-rank test depending on the Shapiro-wilks test of normality.

Quantitative measures on the comfort and ease-of-use when performing voluntary eardrum movements with the EarSwitch. These will be assessed using a custom four Likert-item questionnaire.

Quantitative measures based on subjective questionnaires measuring the comfort and fit of the device. These will be measured using (1) the "Comfort Rating Scales" from Knight et al.1 and (2) the "Bipolar Comfort Rating Scales" from Park and Casali2. In addition, we will take a measure of fit based on (3) depth of otoscope, measured in millimetres, when being worn and when the camera is focused on the eardrum which will be measured using a ruler.

Inclusion Criteria: Able to voluntarily contract the tensor tympani muscle We will be recruiting participants who require assistive technology, as well as people who don't. For the former there will be the requirement that: Individual requires the use of an assistive device in everyday life Exclusion Criteria: Those who do not have the capacity to understand the study and consent (e.g., severe learning disabilities). Inability to communicate either directly to the researcher or through a carer or communication partner and therefore inability to provide consent.

Participant data will be automatically anonymised and therefore individual participant data will not be available to other researchers. Anonymised data will be transferred to Earswitch Ltd and analysed in the future by Earswitch Ltd for the development of more robust algorithms for detecting voluntary tensor tympani movement.

Judge S, Nasr N, Hawley M. A User-Centred Approach Exploring the Potential of a Novel EMG Switch for Control of Assistive Technology. Stud Health Technol Inform. 2017;242:381-384.

Turner MR, Al-Chalabi A. REM sleep physiology and selective neuronal vulnerability in amyotrophic lateral sclerosis. J Neurol Neurosurg Psychiatry. 2020 Jul;91(7):789-790. doi: 10.1136/jnnp-2020-323100. Epub 2020 Apr 30. No abstract available.

https://dl.acm.org/doi/abs/10.1145/3411764.3445205?casa_token=3jX43urd7VgAAAAA:pIas7wDx5dhbivAWZwEuyA8ewJxIdqr2_P0aZQ-dWD_Bk5L6YpkHRcejkT-2Uyynk5ThFGEdwAXy