The At-Home Testing Device (AHTD) database was described in Goetz et al. (2009): it is a novel telemonitoring device built by Intel Corporation for collecting data from PWP. Originally, 52 subjects with idiopathic PD with diagnosis within the previous five years at trial onset were recruited. A PD diagnosis was given if the subject had at least two of the following: rest tremor, bradykinesia (slow movement) or rigidity, without evidence of other forms of Parkinsonism. The study was supervised by six US medical centers: Georgia Institute of Technology (7 subjects), National Institutes of Health (10 subjects), Oregon Health and Science University (14 subjects), Rush University Medical Center (11 subjects), Southern Illinois University (6 subjects) and University of California Los Angeles (4 subjects). All patients gave written informed consent, and remained un-medicated for the six- month duration of the study. We disregarded data from 10 recruits – two that dropped out the study early, and a further eight that provided insufficient test data. The 42 PWP used in this study had at least 20 valid study sessions during the trial period.
The 28 male participants had an age range (mean ± standard deviation) 64.8 ± 8.1, min. 49, max. 78, median 65 years, with 63.0 ± 61.9, min. 1, max. 260, median 48 weeks since
diagnosis. UPDRS was assessed at baseline (onset of trial), after three, and after six months: (20.3 ± 8.5, 21.9 ± 8.7, 22.0 ± 9.2), min. (6, 6, 5), max. (36, 38, 41), median (21, 22, 20) points for motor UPDRS, and (27.5 ± 11.6, 30.4 ± 11.8, 31.0 ± 12.4), min. (8, 7, 7), max. (54, 55, 54), median (27, 28.5, 26.5) points for total UPDRS.
The 14 female participants had an age range 63.6 ± 11.6, min. 36, max. 85, median 64 years with 89.7 ± 81.2, min. 4, max. 252, median 60 weeks since diagnosis. Their UPDRS at baseline, after three and after six months was: (17.6 ± 7.4, 21.2 ± 10.5, 20.1 ± 9.4), min. (6, 6, 8), max. (32, 38, 38), median (18, 18.5, 19.5) points for motor UPDRS, and (24.2 ± 9.1, 27.4 ± 12.1, 26.8 ± 10.8), min. (10, 7, 10), max. (42, 46, 49), median (25, 28, 24.5) points for total UPDRS. At baseline, the combined (male and female) scores were 19.42 ± 8.12, min. 6, max. 36, median 18 points for motor UPDRS, and 26.39 ± 10.80, min. 8, max. 54, median 25.5 points for total UPDRS. After three months: 21.69 ± 9.18, min. 6, max. 38, median 21 points for motor UPDRS, and 29.36 ± 11.82, min. 7, max. 55, median 28 points for total UPDRS, and after six months: 29.57 ± 9.17, min. 5, max. 41, median 20 points for motor UPDRS, and 29.57 ± 11.92, min. 7, max. 54, median 26 points for total UPDRS.
Fig. 6.1 displays graphically the data acquisition and UPDRS estimation procedure. The data is collected at the subject‘s home, transmitted over a dedicated, purpose-built secure server, and processed in the clinic to predict the UPDRS score. The AHTD is designed to facilitate remote, Internet-enabled measurement of a variety of PD-related motor impairment symptoms: it contains a docking station for measuring tremor, paddles and pegboards for assessing upper body dexterity, a high-quality microphone headset for recording patient voice signals and a USB data stick to store test data. A Liquid Crystal Display (LCD) guides the subject in taking the tests.
Fig. 6.1: Schematic diagram depicting the data acquisition procedure and the methodology to estimate the average Parkinson‘s disease symptom severity expressed using the Unified Parkinson's Disease Rating Scale (UPDRS). The device that collects the data is known as the At-Home-Testing-Device (AHTD). The red box (steps 6-8) is the focus of this study.
Audible prompts instruct the subject to undertake tasks to measure tremor, bradykinesia, complex co-ordinated motor function, speech and voice. As part of a trial to test the effectiveness of the AHTD system in practice, PWP were recruited and trained to use the device. Subsequently, an AHTD was installed in their home and they performed tests on a
Parkinson’s patient speaks into microphone Home telemonitoring device records speech signal Speech transferred to USB stick Internet
Patient‘s home Medical Centre
Data into patient’s personal computer Signal processing algorithms applied to speech Statistical mapping of algorithms to UPDRS Predicted UPDRS report to clinical staff Data into dedicated server in the clinic 0 0.5 1 0 10 20 PPE UP DRS J=|F0,i-F0,i+1|
weekly basis. Each patient specified a day and time of the week during which they had to complete the test protocol, prompted with an automatic alarm reminder on the device. The collected data was encrypted and transmitted to a dedicated server automatically when the USB stick was inserted into a computer with internet connection. Further details of the AHTD apparatus and trial protocol can be found in Goetz et al. (2009). Henceforth, we refer to this trial as the AHTD trial.
The audio recordings are of two types: sustained phonations and running speech tests. In the sustained vowel phonations, the subjects were instructed to say ―ahh…‖ and keep the pitch as steady as possible, for as long as possible. In the running speech tests the subject was instructed to describe photographs displayed on the AHTD‘s screen, specifically chosen to elicit emotional responses. The voice samples were recorded using a head-mounted microphone placed approximately 5 cm from the patient‘s lips. The AHTD uses a spoken instruction followed by a ―beep‖ prompting the subject to begin phonation; an audio amplitude threshold detector triggered the capture of audio, and subsequently the capture was stopped one second after the detected signal amplitude dropped below that threshold, or 30 seconds of audio had been captured (whichever occurred sooner). The voice signals were recorded directly to the AHTD USB stick sampled at 24 KHz with 16 bit resolution.
Following initial screening to remove faulty recordings (for example failure to record a phonation, subject coughing, or initialization and very early termination of the phonation), 5875 sustained vowel /a/ phonations were digitally processed using algorithms implemented in the Matlab software package. Six phonations were recorded each day on which the test was performed: four at comfortable pitch and loudness and two at twice the initial loudness (but without shouting).