A Psychometric Platform to Collect Somatosensory Sensations for Neuroprosthetic Use

Somatosensory neuroprostheses exploit invasive and non-invasive feedback technologies to restore sensorimotor functions lost to disease or trauma. These devices use electrical stimulation to communicate sensory information to the brain. A sensation characterization procedure is thus necessary to determine the appropriate stimulation parameters and to establish a clear personalized map of the sensations that can be restored. Several questionnaires have been described in the literature to collect the quality, type, location, and intensity of the evoked sensations, but there is still no standard psychometric platform. Here, we propose a new psychometric system containing previously validated questionnaires on evoked sensations, which can be applied to any kind of somatosensory neuroprosthesis. The platform collects stimulation parameters used to elicit sensations and records subjects' percepts in terms of sensation location, type, quality, perceptual threshold, and intensity. It further collects data using standardized assessment questionnaires and scales, performs measurements over time, and collects phantom limb pain syndrome data. The psychometric platform is user-friendly and provides clinicians with all the information needed to assess the sensory feedback. The psychometric platform was validated with three trans-radial amputees. The platform was used to assess intraneural sensory feedback provided through implanted peripheral nerve interfaces. The proposed platform could act as a new standardized assessment toolbox to homogenize the reporting of results obtained with different technologies in the field of somatosensory neuroprosthetics.


INTRODUCTION
Somatosensory neuroprostheses are highly innovative devices (1). Several research groups have investigated the ability to restore sensory feedback in patients with upper or lower limb amputation, tetraplegia, or paraplegia using invasive (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14) and non-invasive (15)(16)(17)(18)(19) interfaces with the peripheral (PNS) and the central nervous systems (CNS) (Figure 1). The main aim of these FIGURE 1 | Neuroprosthetic applications. Neurotechnologies for restoring somatosensations have been developed for peripheral (PNS) or central nervous systems (CNS). The stimulation technique used to restore sensory feedback can be invasive (surgically implanted and in intimate contact with the nervous tissue) or non-invasive (applied on the skin surface). Delivering a stimulation to the brain or peripheral nerves provides benefits such as the control of robotics, smart prosthetics, or other assistive technologies.
technologies is to elicit somatotopic-referred sensations emanating from the affected limb, creating a personalized map of the these sensations which could be used as a sensory feedback aimed at improving the patients' quality of life (20,21). All these approaches use neural stimulation to evoke sensations stemming from contact with sensory peripheral nerves or the neural interfaces are placed directly on the somatosensory cortex.
The external stimulation of the residual (still functional proximal to the lesion) nervous structures guarantees to evoke an artificial sensation that can also be controlled by modulating the stimulation parameters (22)(23)(24).
Since there is intersubject variability due to the different nerve structures, implantation levels, and innervation (25)(26)(27), together with the subjective perception of the elicited sensations, a "sensation characterization" procedure is necessary to obtain a uniform sensation mapping (Figure 2). The goal of this procedure is to collect all the stimulation parameters corresponding to the evoked sensations characterized by the intensity, quality, location, and type in order to have a personalized sensation map. The mapping phase is crucial to implement an effective real-time assistive system, e.g., bidirectional hand or leg prostheses, eliciting homologous referred sensations emanating from the phantom limb (somatotopic) for therapeutic or functional purposes. In fact, the personalized sensation map is often translated into a robotic arm or hand in order to elicit sensations during object manipulation tasks aimed at increasing patient motor control performance (4,6,28,29). When the patient is controlling a robotic arm, and touches a surface with the second robotic digit, the sensation perceived should be in the same location (index), with the safety and exact intensity (mapped with the pressure force of the robotic finger) and the type should be in line with finger pressure (i.e., no electricity or warmness). The personalized sensation map should thus be as detailed as possible.
In addition, to provide a rich and reliable artificial sensory feedback to the patient, it is necessary to accurately test multiple sets of stimulation parameters and re-test them even over multiple days during long-term applications (14,(30)(31)(32)(33). This characterization procedure is fundamental to re-create an artificial sensory feedback that could be effectively exploitable in neuroprosthetic applications by the users. To this aim, it is crucial to collect the data in an effective and efficient way using a platform in which the information are exhaustive and standardized. This would guarantee an easy comparison of the results with different technology or algorithm.
Several psychometric questionnaires exist regarding the quality and type of the sensations evoked (15,(34)(35)(36). However, they do not appear to be easy-to-use or fast for recording and integrating all the properties of the elicited sensations with detailed standard questionnaires and which could be used for several types of sensory feedback.
The psychometric platform presented in this study provides a uniform way of characterizing and quantifying the artificial sensory feedback systems used for invasive and non-invasive, peripheral, and central sensory feedback, in order to efficiently compare, optimize, and evaluate all the different approaches even over time. Our platform records the stimulation parameters, quality, type, intensity, and location of the evoked sensations. All the sensation data are collected from questionnaires already presented in the literature.
The platform also provides a user-friendly graphical user interface with a touch screen for the patient's answers that not only enables the patient to describe the percept in detail but also provides clinicians with all the main information on the evoked sensation. The platform accepts new questionnaire definitions as text and is easy to understand and implement. This means that researchers can add new questionnaires, such as phantom limb pain (PLP) (37,38), in order to collect information on new treatments. This psychometric platform was tested on three trans-radial amputees who had four intrafascicular electrodes (39) implanted in their median and ulnar nerves for 6 months each. The patients responded using the psychometric platform when they received electrical stimulation by the electrical contacts of the neural interfaces. The software was used by clinicians and engineers to collect the data. This has proven to be more convenient than writing down hundreds of answers in weekly trials over 18 months.
In this study, we describe the usability of this new platform. We believe that our new psychometric platform will facilitate and unify the characterization of percepts and the comparison of the effects when applying different neural stimulation techniques or using different devices.

Software Platform
The psychometric platform is made up of a mobile application for compiling questionnaires (which we have called Easy Quest), two desktop tools (Easy Quest Create and Easy Quest Evaluate), and a desktop application to control the neurostimulator.
The Easy Quest mobile app is described in depth in the following sections.
Easy Quest Create shows a simple graphical user interface in which the experimenter can create a list of questions from a set of predefined types. The content can be customized.
Easy Quest Evaluate is devised for the rapid evaluation of a set of answers; the software reads the archive file exported by Easy Quest and exports a comma-separated values (CSV) file. The choice of CSV format of the results makes further analyses easier, as it is compatible with MATLAB (MathWorks, Inc., Natick, Massachusetts, United States) and Microsoft Excel (Microsoft Corporation, Redmond, Washington, United States).
The desktop application for the actual neurostimulation is not described here, because its design is strongly dependent on the type of experiment and neurostimulation device (communication protocols, stimulator commands, and architecture); however it is mentioned as part of the experimental setup.

Somatosensory Questionnaires
Somatosensory descriptors were selected from the literature and clinical settings also including questionnaires that have already been used in neuroprosthetic studies. Several options describing the type, quality, intensity, and the location are presented in order to characterize the somatosensory percepts being evoked during the stimulation. To describe the quality of sensations, we used a scale presented by Lenz et al. (34) and used also by Valle et al. (9). For the sensation type, we adapted the questionnaire proposed by Kim et al. (36) based on our experience with several upper limb patients stimulated with invasive (4,9,23,24,29,31,(40)(41)(42)(43)(44)(45)(46)(47) and non-invasive technologies (17,48). We also considered other studies on sensations elicited using peripheral (5,10,14,22,(49)(50)(51)(52)(53)(54)(55) or central (6,8,13) neural stimulation approaches. For the intensity, we used a Visual Analog Scale (VAS) (37) already presented by Tan et al. (56). Lastly, the perceived sensation locations were shown directly on a schematic representation of the human hand. It is further possible to select the feet, arms, or legs (11,12) with several possible spots (Figure 2). In this way, the patient can accurately indicate the affected areas.
We added several questionnaires in order to collect information on PLP: VAS (37) and neuropathic pain symptom inventory (NPSI) (38). It is also possible to add or modify the existing questionnaires in order to adapt the platform to the needs and specifications of the clinical trial.

Use Cases
Two main use cases for the app were identified (Figure 3). In the first, the user fills in a questionnaire and saves the results on the device, defined as the "local fill-in" (LF). In the second, an external software prompts the app to show a questionnaire and to send back the results, defined as "on demand fill-in" (ODF). The two cases (Figures 3A,B) involve the same procedure in the part where the user is asked to fill in the answers.
The main difference, besides the location where the results are stored, is how the procedure starts: in the first case, the user choses a questionnaire by selecting it from the main menu, in the second, the app waits for an external command, usually from the network, instructing the software to show a specific questionnaire.
The application can set recurrent reminders for specific questionnaires, enabling the experimenter to plan the follow-up for home use by the patient, and these reminders prompt the user to fill in the questions in LF mode.
A third use case ( Figure 3C) explains the workflow from the perspective of the experimenter, who uses the companion software to define new questionnaires at the beginning of the experimentation and to display the results at the end.
An explanatory example of the platform is displayed in Supplementary Video 1.

Software Architecture
The software was developed in Dart, an object-oriented programming language developed by Google in order to address server-side, web, and mobile platforms. The mobile SDK, Flutter, compiles the code in fast native apps for Android and iOS devices.
The app is developed following the Model View Controller (MVC) pattern, and a simple Object-Relational Mapping (ORM) is implemented to store the models in an SQLite database in the device's memory. The ORM is accessed through classes which show APIs where serialized objects can be stored and retrieved.
To implement the ODF, a simple Hypertext Transfer Protocol (HTTP) server runs in the background thus the app can, when requested, wait for remote commands from the network. While doing so, the app shows a numerical code, which must be notified to the experimenter to secure the remote connection.
An interface with the mail app is used to send the completed questionnaires as a CSV file by e-mail.
Another provider class parses the questionnaires defined in JavaScript Object Notation (JSON) format, making it possible to create and add new questionnaires to working devices, On demand fill-in (ODF), in this case, the app waits for an external command from a controller app containing information on the questionnaire to be shown; the fill-in procedure is the same but nothing is stored within the device, instead results are sent back to the controller. (C) The procedure seen from the experimenter's point of view, here, the role of the other software programs of the platform (Easy Quest Create, Easy Quest Evaluate) is explained.
without code interventions and recompiling the whole app. The import service can parse a compressed file containing a set of questionnaires and also a collection of images referred to in the questions. There are five questions accepted by the parser: (1) open, which prompts the user for a string (2); radio, which asks the user to choose one option from a set (3); multiple choice; (4) slider, where users have to select a number or a label with a slider; and (5) image touch, where the user selects a set of touchable areas displayed on top of a given background image.
The app enables multiple users to access the same device while keeping the results separate.
The system architecture is shown in Figure 4, along with the external software highlighting its relationship with the app modules.
The app UI/UX is designed in accordance with Material, an open source system of guidelines developed by Google. The view layer written for the app exploits all the available space, presenting the questionnaire as a list of questions on small devices and as a grid on larger screens.

Quality and Usability Assessment
During the clinical trial, we collected feedback information from patients, clinicians, and engineers who used the platform presented in this study in three clinical trials (N = 12). The investigations regarded the development and assessment of bidirectional hand prostheses for upper limb amputees with a neural sensory feedback delivered by implantable electrodes (9,23,31,41,42). After 6 months of use, we asked participants to answer different quality and usability questions using: questionnaires for user interface satisfaction (QUIS) (57), system usability scales (SUS) (58,59), Nielsen's attributes of usability (NAU) (60), and after-scenario questionnaires (ASQ) (61). We collected and analyzed all the information using validated and standardized questionnaires (Figure 5).

Data Storage
Alongside the answers to the psychometric questionnaires, a MATLAB script running in experimenter's computer stored date and time for further analysis about the use of the platform. All the relevant measurements are stored as.mat files, as all the data processing have been performed inside the MATLAB software.

Somatosensory Questionnaires for Sensation Characterization
To efficiently characterize the sensations emanating from (invasive or non-invasive) electrical (central or peripheral) stimulation, a user-friendly platform is needed with a set of somatosensory-related questionnaires. This helps to reduce the long time required to collect all the electrically evoked sensation data.
To assess the properties of the sensations being evoked by stimulating peripheral nerves using a neural interface in trans-radial amputees, we used the psychometric platform presented here. We performed a procedure called "sensation characterization" with all the patients involved in the clinical investigation (Figure 2). For each electrically active site used to stimulate the nerve, the neural stimulation was delivered, and the patient was asked to report the sensations he/she felt. This mapping phase enabled us to identify the sensation properties for all the stimulation channels of the implanted electrodes by varying the stimulation parameters and building a personalized map of the sensations. The stimulation parameters varied in terms of frequency (1-1,000 Hz), pulse-width (1-120 µs), and amplitude (1-1,000 µA), as well as stimulation  train duration (discrete or continuous). We collected the sensation intensity, quality, type, and location of the patient's perceived sensations.
The intensity was used to find the perceptual thresholds for each stimulation channel (4,23,24,31), together with the range of stimulation (between threshold and below pain level). Using a VAS scale in the range from 0 to 10 also enables us to identify perceptual magnitude levels (5,23,32,52,62).
The quality of the sensory feedback was assessed in order to test different stimulation strategies and approaches (9,23), since this quality is considered to be an important factor for prosthesis acceptance (53,63). To quantify the perception quality and naturalness, we used a scale (34) from 1 (totally unnatural) to 5 (totally natural).
The type of sensation was collected in order to understand the type of fibers being recruited during the stimulation and to identify the best channels for restoring homologous sensations while using the bidirectional prosthesis. We used 20 descriptors ( Table 1) considering all the important aspects. In this platform, the patient could also report a new sensation or add comments in an empty text box when a correct descriptor for the elicited sensation was lacking.
The sensation location was reported using a picture of the limb of interest (foot, arm, leg, or hand) with several highlighted spots (20 for foot, 24 for leg, 48 for arm, and 45 for hand) (Figure 2). The zones with a higher density of receptors had more selectable spots. This information is useful to understand the electrode stimulation selectivity (analyzing the spreading of the zone) and the layout of the fibers inside the nerve. In addition while the bidirectional prosthesis was being used, the location map was needed to stimulate the correct active sites eliciting the somatotopic sensation during the prosthesis hand/finger contact with objects (4).
Finally, several questions can be used to assess phantom limb pain levels before and after a pain treatment with electrical stimulation (31). We decided to use two different questionnaires (VAS and NPSI) to characterize the location, quality, and intensity of the pain (11,31).

Software Usability
The usability testing of the app was performed on an Android phone (a Nexus 6p), designed by Huawei and running Android 8.
The app loading time is <2 s, needing only the time to open the local database, and after the login screen, the user can access all the main functions in no more than two taps.
The home page shows a list of all the available questionnaires, the user can tap on each one to see the questions and fill in the answers, which are stored in the internal database.
From the lateral menu (drawer), the ODF mode can be accessed in only one tap, after which the app will wait for a network command containing the identifier of the questionnaire to be shown.
Minimal user interaction is needed to complete a questionnaire, usually all the questions need just one tap, except for the multiple choice and clickable area ones. The average time to fill in a sensation characterization questionnaire is 10 s.
The export page lets the user write all the stored data in a CSV archive file and opens the default mail to send to the experimenters for further analysis, facilitating and speeding up the data-gathering phase.
A specific section of the app lets the user choose which questionnaire should be visible in the home page, personalizing the user interface for a specific use.
Other pages are designed for secondary tasks, such as previewing stored answers and editing settings.

Psychometric System Validation
In order to assess the usability and quality of this novel psychometric platform to collect somatosensory percepts, several questionnaires were filled in by different kinds of users. Three patients, six engineers, and three clinicians evaluated the system by answering four questionnaires after using the platform in clinical applications (Figure 5). Analyzing the results, the overall reactions to the system were very positive. The average score was 7.1 ± 0.3. Considering the user interface satisfaction (QUIS), the rating achieved was 6.6 ± 0.8. In both these questionnaires, the maximum achievable score was 9.
During the clinical trial, the psychophysical platform stored 1,913 measurements.

Usage Metrics
The time passed between a measurement and the subsequent have been calculated and plotted in Figure 5. Records were aggregated by day and outliers due to technical problems and breaks for the subject were ignored, the mean time between two measurements is 74 ± 31.8 s. The number of measurements per day have also being considered an indicator of the overall use of the platform, the mean use overall was 128 ± 39.6 measurements per day. Each measurement consisted of a.mat file of roughly 1.14 kB, the whole dataset totaled then for almost 2.2 MB.

DISCUSSION
Electrical stimulation has been proposed as a way of restoring somatosensations (15,(63)(64)(65)(66)(67) in cases where they have been lost due to injury or disease in both the CNS or the PNS. In fact, sensory feedback is crucial to improve the motor control of robotic limbs or prostheses, enabling the patient to be more efficient in manipulating objects (4,28,55,68). The sensations evoked thus had to be characterized in detail in patients receiving stimulation in order to restore the sensory information. The psychometric questionnaires were able to register all the aspects of the sensations being restored in a reliable and efficient way, considering more descriptors than in previous studies (34) and using a user-friendly platform.
Currently, there are many important sensation properties which need to be collected in order to obtain an intuitive and rich sensory feedback. In particular, the sensation location, type, quality, and intensity are valid and extendable for all the approaches in different neurological conditions. Considering the previously presented interface to collect stimulation-evoked somatosensory percepts, Geng et al. (35) showed a platform used to evaluate electrical stimulation to relieve PLP. Their platform was interfaceable with one type of neural stimulator and contained three questions to characterize the evoked sensation considering 12 sensation descriptors. The psychometric platform presented here reports somatosensory percepts based on five questionnaires containing 20 standard sensory descriptors ( Table 1). The platform exploits a customizable, fast, and easyto-use GUI which can be efficiently connected to several neural stimulators (31,69,70).
Since several groups are currently using electrical stimulation to restore sensory feedback, a standard somatosensory platform could facilitate their comparison, assessment, and optimization. Our findings support the conclusion that this psychometric platform could help and accelerate the development of sensorimotor neuroprostheses.
Given the simple software architecture, this platform is flexible in terms of modifications and upgrades. It is possible to add new questionnaires regarding other aspects of sensory feedback restoration. For example, two important features to be considered for the development of the next generation of somatosensory neuroprostheses are embodiment (42,71,72) and psychological/affective aspects (73).
The psychometric platform is simple to interface with other devices and also with existing software, thanks to its open and platform-agnostic interfaces: in ODF mode the HTTP interface accepts commands regardless of the device and the programming language of the sender application (all major languages can implement HTTP communication effortlessly). Answers to the questionnaires are exported in a CSV format, making it easy for any other software program to import and analyze them.
Considering the results of the usability assessments (Figure 5), users highlighted various positive and negative aspects which will then help us to improve the platform. The most positive aspect in terms of the "overall reaction to the software" was that the software is easy to use, which is crucial both for patients and experimenters.
The QUIS answers revealed that this system is consistent and very clear; however, we still need to improve error and warning messages. These aspects mainly regard the experimenters' side. The SUS again indicated that the system is easy to use and intuitive, but additional material and instructions should be included as support. Moreover, the NAU showed a high user satisfaction along with a request for more error messages. Finally, the ASQ revealed "the ease of completing this task, " thus highlighting the need for more support, information, and documentation. We thus intend to improve the platform using these usability results.
Usage metrics confirmed the ease of the fill-in process, with a mean time needed for a complete session of 74 s. Thanks to that stimulation spots were probed with a suitable rate (128 ± 39.6 measurements per day), for the intended objectives. The small dimension of the measurements made easy the subsequent steps of data revision, analysis, and storage, anyway authors are planning further reduction in the memory needed.

STUDY LIMITATIONS
There are several limitations connected to the patient attention at the time of testing. To solve this issue, it is important to repeat the test multiple times over multiple days in order to increase its reliability. The test is also highly subjective, and the mapping results could strongly depend on the sensation of the patient and his/her personal experience (74). The individual subjective differences remain a big challenge for interpreting the somatosensory results and also the semantic differences. The usability and the utility of the platform need to be demonstrated with a larger population of patients with sensory deficits and with other technologies for the restoration of sensory feedback (e.g., non-invasive stimulation, brain stimulation).
Sham (placebo) and blind stimulations could also be delivered to test individual response bias and identify possible unreliable self-reports.
The software design, particularly the GUI, was inspired by the principles of the ISO 9241 standard. In fact, the users' opinions of the platform were taken into consideration during the design phase and the assessment.
The software will be actively used during experiments, and the user experience will be monitored to improve new versions, ensuring an iterative development driven by user feedback, as also stated in ISO 9241. In particular, in addition to the online availability of the software, the platform needs to be used in future clinical trials for both upper-and lower-limb amputees provided with a fully implantable sensory feedback system in long-term studies.

CONCLUSIONS
This study has presented a psychometric platform used to record a complete somatosensory percept description, which can be evoked by several different methods of electrical stimulation in humans. The subjective somatosensory sensation type, location, quality, and intensity are collected and used to develop a somatosensory questionnaire, which can be used for neuroprosthesis calibration and optimization. The psychometric toolbox is implemented in a user-friendly software program. The platform was validated in patients with electrodes implanted in the PNS.
We believe that this new somatosensory psychometric system will help to establish a standard and uniform methodology of subjective sensory reports, which is a pivotal step to uniformly develop, adapt, and improve somatosensory neuroprostheses.

DATA AVAILABILITY STATEMENT
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request. The Easy Quest Android application is available from the Google Play Store platform: https://play. google.com/store/apps/details?id=me.francescoiberite.research. easy_quest.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Institutional Ethics Committees of Policlinic A. Gemelli at the Catholic University and Italian Ministry of Health. The patients/participants provided their written informed consent to participate in this study. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.

AUTHOR CONTRIBUTIONS
GV designed the study, developed the software, analyzed the data, and wrote the paper. FI developed the software, analyzed the data, and reviewed the manuscript. IS designed the study, developed the software, and reviewed the manuscript. ED'A developed the software and reviewed the manuscript. GG, RD, and PR tested the platform with patients. TS developed the TIME electrodes. TS and PR discussed the results and reviewed the manuscript. SR designed the platform. FP designed the platform and discussed the results. SM designed the study, discussed the results, and reviewed the manuscript. All authors read, commented, and approved the manuscript.

ACKNOWLEDGMENTS
This manuscript has been released as a preprint at Biorxiv (75).

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmedt.

2021.619280/full#supplementary-material
Supplementary Video 1 | Example of collection of somatosensory sensation data. The video shows the ODF procedure, where the Easy Quest app and the software controlling the stimulator interact. After the login procedure, the user sets the android application in ODF, opening the communication port. As soon as the controller software, after a verification procedure, detects the running app, there is a visual confirmation in the dashboard on the right-hand side of the screen. When stimulation parameters are correctly set, the software can deliver the stimulation and automatically send a sensation questionnaire to the subject's device. Instead of being stored in the device, the answers are sent back to the control software and are shown in the lower part of the interface.