Dataset Information

Detecting Parkinson Disease Using a Web-Based Speech Task: Observational Study.

ABSTRACT:

Background

Access to neurological care for Parkinson disease (PD) is a rare privilege for millions of people worldwide, especially in resource-limited countries. In 2013, there were just 1200 neurologists in India for a population of 1.3 billion people; in Africa, the average population per neurologist exceeds 3.3 million people. In contrast, 60,000 people receive a diagnosis of PD every year in the United States alone, and similar patterns of rising PD cases-fueled mostly by environmental pollution and an aging population-can be seen worldwide. The current projection of more than 12 million patients with PD worldwide by 2040 is only part of the picture given that more than 20% of patients with PD remain undiagnosed. Timely diagnosis and frequent assessment are key to ensure timely and appropriate medical intervention, thus improving the quality of life of patients with PD.

Objective

In this paper, we propose a web-based framework that can help anyone anywhere around the world record a short speech task and analyze the recorded data to screen for PD.

Methods

We collected data from 726 unique participants (PD: 262/726, 36.1% were women; non-PD: 464/726, 63.9% were women; average age 61 years) from all over the United States and beyond. A small portion of the data (approximately 54/726, 7.4%) was collected in a laboratory setting to compare the performance of the models trained with noisy home environment data against high-quality laboratory-environment data. The participants were instructed to utter a popular pangram containing all the letters in the English alphabet, "the quick brown fox jumps over the lazy dog." We extracted both standard acoustic features (mel-frequency cepstral coefficients and jitter and shimmer variants) and deep learning-based embedding features from the speech data. Using these features, we trained several machine learning algorithms. We also applied model interpretation techniques such as Shapley additive explanations to ascertain the importance of each feature in determining the model's output.

Results

We achieved an area under the curve of 0.753 for determining the presence of self-reported PD by modeling the standard acoustic features through the XGBoost-a gradient-boosted decision tree model. Further analysis revealed that the widely used mel-frequency cepstral coefficient features and a subset of previously validated dysphonia features designed for detecting PD from a verbal phonation task (pronouncing "ahh") influence the model's decision the most.

Conclusions

Our model performed equally well on data collected in a controlled laboratory environment and in the wild across different gender and age groups. Using this tool, we can collect data from almost anyone anywhere with an audio-enabled device and help the participants screen for PD remotely, contributing to equity and access in neurological care.

SUBMITTER: Rahman W

PROVIDER: S-EPMC8564663 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:BACKGROUND:Regular testing for sexually transmitted infections (STIs) is important to maintain sexual health. Self-sampling kits ordered online and delivered in the post may increase access, convenience, and cost-effectiveness. Sexual health economies may target limited resources more effectively by signposting users toward Web-based or face-to-face services according to clinical need. OBJECTIVE:The aim of this paper was to investigate the impact of two interventions on testing activity across a whole sexual health economy: (1) the introduction of open access Web-based STI testing services and (2) a clinic policy of triage and signpost online where users without symptoms who attended clinics for STI testing were supported to access the Web-based service instead. METHODS:Data on attendances at all specialist public sexual health providers in an inner-London area were collated into a single database. Each record included information on user demographics, service type accessed, and clinical activity provided, including test results. Clinical activity was categorized as a simple STI test (could be done in a clinic or online), a complex visit (requiring face-to-face consultation), or other. RESULTS:Introduction of Web-based services increased total testing activity across the whole sexual health economy by 18.47% (from 36,373 to 43,091 in the same 6-month period-2014-2015 and 2015-2016), suggesting unmet need for testing in the area. Triage and signposting shifted activity out of the clinic onto the Web-based service, with simple STI testing in the clinic decreasing from 16.90% (920/5443) to 12.25% (511/4172) of total activity, P<.001, and complex activity in the clinic increasing from 69.15% (3764/5443) to 74.86% (3123/4172) of total activity, P<.001. This intervention created a new population of online users with different demographic and clinical profiles from those who use Web-based services spontaneously. Some triage and signposted users (29.62%, 375/1266) did not complete the Web-based testing process, suggesting the potential for missed diagnoses. CONCLUSIONS:This evaluation shows that users can effectively be transitioned from face-to-face to Web-based services and that this introduces a new population to Web-based service use and changes the focus of clinic-based activity. Further development is underway to optimize the triage and signposting process to support test completion.

Project description:ObjectiveSubthalamic nucleus deep brain stimulation (STN-DBS) is an effective treatment for advanced Parkinson disease (PD). Following STN-DBS, speech intelligibility can deteriorate, limiting its beneficial effect. Here we prospectively examined the short- and long-term speech response to STN-DBS in a consecutive series of patients to identify clinical and surgical factors associated with speech change.MethodsThirty-two consecutive patients were assessed before surgery, then 1 month, 6 months, and 1 year after STN-DBS in 4 conditions on- and off-medication with on- and off-stimulation using established and validated speech and movement scales. Fifteen of these patients were followed up for 3 years. A control group of 12 patients with PD were followed up for 1 year.ResultsWithin the surgical group, speech intelligibility significantly deteriorated by an average of 14.2%±20.15% off-medication and 16.9%±21.8% on-medication 1 year after STN-DBS. The medical group deteriorated by 3.6%±5.5% and 4.5%±8.8%, respectively. Seven patients showed speech amelioration after surgery. Loudness increased significantly in all tasks with stimulation. A less severe preoperative on-medication motor score was associated with a more favorable speech response to STN-DBS after 1 year. Medially located electrodes on the left STN were associated with a significantly higher risk of speech deterioration than electrodes within the nucleus. There was a strong relationship between high voltage in the left electrode and poor speech outcome at 1 year.ConclusionThe effect of STN-DBS on speech is variable and multifactorial, with most patients exhibiting decline of speech intelligibility. Both medical and surgical issues contribute to deterioration of speech in STN-DBS patients.Classification of evidenceThis study provides Class III evidence that STN-DBS for PD results in deterioration in speech intelligibility in all combinations of medication and stimulation states at 1 month, 6 months, and 1 year compared to baseline and to control subjects treated with best medical therapy.

Project description:BACKGROUND:Novel alcohol prevention strategies are needed for heavy-drinking young adults. Sleep problems are common among young adults who drink heavily and are a risk factor for developing an alcohol use disorder (AUD). Young adults, interested in the connection between sleep and alcohol, are open to getting help with their sleep. Therefore, sleep interventions may offer an innovative solution. This study evaluates social media advertising for reaching young adults and recruiting them for a new alcohol prevention program focused on sleep. OBJECTIVE:This study aims to evaluate the effectiveness and cost of using Facebook, Instagram, and Snapchat advertising to reach young adults who drink heavily for a sleep intervention; characterize responders' sleep, alcohol use, and related concerns and interests; and identify the most appealing advertising content. METHODS:In study 1, advertisements targeting young adults with sleep concerns, heavy alcohol use, or interest in participating in a sleep program ran over 3 months. Advertisements directed volunteers to a brief web-based survey to determine initial sleep program eligibility and characterize the concerns or interests that attracted them to click the advertisement. In study 2, three advertisements ran simultaneously for 2 days to enable us to compare the effectiveness of specific advertising themes. RESULTS:In study 1, advertisements generated 13,638 clicks, 909 surveys, and 27 enrolled volunteers in 3 months across the social media platforms. Fees averaged US $0.27 per click, US $3.99 per completed survey, US $11.43 per volunteer meeting initial screening eligibility, and US $106.59 per study enrollee. On average, those who completed the web-based survey were 21.1 (SD 2.3) years of age, and 69.4% (631/909) were female. Most reported sleep concerns (725/909, 79.8%) and an interest in the connection between sleep and alcohol use (547/909, 60.2%), but few had drinking concerns (49/909, 5.4%). About one-third (317/909, 34.9%) were identified as being at risk for developing an AUD based on a validated alcohol screener. Among this subsample, 8.5% (27/317) met the final criteria and were enrolled in the trial. Some volunteers also referred additional volunteers by word of mouth. In study 2, advertisements targeting sleep yielded a higher response rate than advertisements targeting alcohol use (0.91% vs 0.56% click rate, respectively; P<.001). CONCLUSIONS:Social media advertisements designed to target young adults with sleep concerns reached those who also drank alcohol heavily, despite few being concerned about their drinking. Moreover, advertisements focused on sleep were more effective than those focused on drinking. Compared with previous studies, cost-effectiveness was moderate for engagement (impressions to clicks), excellent for conversion (clicks to survey completion), and reasonable for enrollment. These data demonstrate the utility of social media advertising focused on sleep to reach young adults who drink heavily and recruit them for intervention.