**7. Discussion**

We developed the Rehapiano device for the quick evaluation of action tremor. Tremor is measured during a steady state, when fingers are exerting the desired force. From a clinical viewpoint, action tremor, specifically kinetic (see Figure 1), appears during the targeted motion. Such movement is represented many times during everyday activities. Therefore, we decided to develop a device that measures this tremor, and based on the results of a particular activity, a physician can observe the development of PT.

Concerning Hypothesis 1, we rejected the null hypothesis based on the low frequency hypothesis test. Based on our observations, we still have confidence that the Rehapiano is capable of reliably measuring tremors of lower frequencies and that the problem of verifying it lies in the DC motor and the rubber band used for its verification. The difference between the measurements from the encoder and the Rehapiano may be due to the low torque of the DC motor at low voltage. The rubber band slows down the motor too much when being stretched and then, when being contracted, causes a peak in its angular velocity. This may distort the measurement. The authors will build a new verification device and repeat the procedure in the near future. For medium- and high-frequency tremors, we have confirmed Hypothesis 1.

Concerning Hypothesis 2, two factors influence the quality of the measurements and the resulting accuracy of the classifiers. The first factor is the subject's stress. According to the work in [16,45], the amplitude and frequency of tremor under psychological stress conditions increase significantly compared to a calm state. The second factor is the medication the subject receives. Based on the subjective opinions expressed by the subjects, they did not feel stressed during the measurement. All PD-affected subjects (except for the one measured in relation to Hypothesis 3) were taking medication to relieve PD symptoms. All the subjects completed the measurement protocol successfully. Two subjects had to repeat the measurement routine for one of their fingers due to wrong placement of the fingertip on the strain gauge. Based on the experimental results, we consider Hypothesis 2 to be confirmed.

As for Hypothesis 3, we subsequently measured both hands of a PD patient seven times. PT most often presents unilaterally and later progresses to include both sides of the body [46]. Both upper limbs of our subject were affected. Although the measurements of the left hand were valid, we had to ignore the measurements of the right hand. The patient was unable to keep the fingertips of the right hand in steady contact with the sensor, tapping the strain gauge more or less. From our perspective, this involuntary motion invalidated most of the right hand measurements. We discovered that, on multiple occasions, the signal contained several dominant frequencies. Therefore, we are currently working on several design and technological improvements to our Rehapiano device, primarily to improve the contact between the finger and strain gauge using neodymium magnets and gloves. Based on the experimental results of the PD subject's left hand measurements, we consider Hypothesis 3 to be confirmed.

#### **8. Conclusions**

This paper presents the Rehapiano system for measuring force applied by the fingertips, used here to detect and quantify hand tremors. We focused on measuring the action hand tremor of healthy subjects and PD patients in this study. We can confirm that the Rehapiano is capable of measuring tremors with frequencies higher than 3 Hz, that the Rehapiano measurements contain detectable and useful tremor information, and that the Rehapiano produces repeatable results. We have implemented a measurement protocol aided by a virtual nurse. Our device is noninvasive and non-wearable. It differs from the existing solutions due to its low measurement time, its comfort level, and the rapid processing and evaluation of the results. The average time of the measurement procedure was three minutes, and the subjects were not fatigued. We will be replacing the Hx711 converters to increase the sampling frequency, thus obtaining finer and higher quality data, which will also enable measurement of orthostatic tremor (between 14 Hz and 16 Hz).

We plan to expand the sample of the test subjects and to create measurement protocols for patients of specific groups. Future work also includes using the Rehapiano as a hand fine motor rehabilitation device. We are developing a rehabilitation computer game with the Rehapiano as the game controller. The game should motivate patients with motor disorders to exercise and improve their fine motor skills. Based on the field work and cooperation with the medical team, we learned that no PD patient barring one had problems in using the Rehapiano. The strain gauges' pods are equipped with neodymium magnets, although we have not yet used this feature. We plan to use fitting gloves with magnetic inserts to help keep the fingertips in place during the interaction. We have also learned that quantifying the tremor of PD patients at regular intervals could aid in determining the proper medication for them and in adjusting the dosage.

After carrying out all the planned improvements to the Rehapiano, we plan to use it to detect other types of tremor in clinical settings. Further, we want to compare Rehapiano measurements with reference measurements from IMUs and EMG. These experiments should provide conclusive evidence about the clinical feasibility of the Rehapiano.

**Author Contributions:** Conceptualization, N.F., M.J., M.B., and F.C.; Rehapiano hardware, N.F.; methodology, M.B., M.J., and F.C.; N.F. and M.J. proposed and described the algorithm and the methodology; software, M.J. and N.F.; protocol, N.F. and M.J.; investigation, M.B. and N.F.; N.F. designed and carried out the experiments with the patients; data processing, M.J.; writing—original draft preparation, N.F., M.J., M.B., and F.C.; editing, M.B. and F.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This publication is the result of the APVV grant Multimodal Human–Robot Interaction Using Cloud Resources (APVV-15-0731) (50%), H2020-MSCA-RISE-2018,824047—LIFEBOTS Exchange (25%), and VEGA, 2017-2020, 1/0663/17 Intelligent Cyber-Physical Systems in Heterogeneous Environments with the Support of IoE and Cloud Services (25%).

**Acknowledgments:** The authors gratefully acknowledge all the volunteers who participated in the trial. The study was performed in cooperation with the University Hospital of L. Pasteur in Košice, Svet Zdravia Hospital in Trebišov, and Retirement Home Trebišov, Slovakia. We thank Theodoz Molcanyi, Miriam Dziakova, Peter Mucha, Lukas Zbojovsky, Frantisek Mihalcik, and Nikola Harmadiova for their help and advice.

**Conflicts of Interest:** The authors declare no conflicts of interest.

#### **References**


c 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
