Motor Impairment Estimates via Touchscreen Typing Dynamics Toward Parkinson's Disease Detection From Data Harvested In-the-Wild

Iakovakis, Dimitrios; Hadjidimitriou, Stelios; Charisis, Vasileios; Bostantjopoulou, Sevasti; Katsarou, Zoe; Klingelhoefer, Lisa; Reichmann, Heinz; Dias, Sofia B.; Diniz, José Alves; Trivedi, Dhaval; Chaudhuri, K. Ray; Hadjileontiadis, Leontios J.

doi:10.3389/fict.2018.00028

Cited by 30 publications

(32 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A correlation analysis between the subjects’ upper extremity UPDRS Part III single items 22/23/31 and the hybrid model predicted parameters, i.e., dRSi, dAFSi, DBSi, exhibited correlation of 0.66/0.73/0.58, respectively. These results show that the proposed hybrid approach outperforms the previous method developed 19 , which achieved corresponding correlations of 0.64/0.58/0.55, respectively. In Fig.…”

Section: Resultsmentioning

confidence: 72%

“…3 . In particular, the performance of the predicted dRSi, dAFSi, DBSi values per subject, when compared to the previous method 19 , the UPDRS Single Item score 22/23/31 and the sum of UPDRS Part III items, is depicted. Specifically, in T1 scenario, the dRSi was the best performing typing-based index, achieving 0.89 [with 0.83–0.97 95% confidence interval (CI)] area under the ROC curve (AUC), and with 0.90/0.83 sensitivity/specificity, whereas the UPDRS Part III items 22/23/31 achieved AUC of 0.84 (with 0.77–0.92 95% CI)/0.77 (with 0.69–0.86 95% CI)/0.93 (with 0.87–0.98 95% CI) with {0.94/0.73 ,0.94/0.60, 0.88/0.95} sensitivity/specificity per item, respectively.…”

Section: Resultsmentioning

confidence: 99%

“…This approach, provided a granular quantification of the severity footprint of the FMI. Therefore, in a subsequent analysis 19 , the best performing features were fed into a random forest regression algorithm to estimate specific UPDRS Part III single-item scores, aiming at extracting a specific relationship between the keystroke features and the FMI symptoms, achieving up to 0.83 correlation with the UPDRS Part III Single Item 22 (rigidity-related) in-the-clinic setting. In the case of remotely and unobtrusively acquired typing data in-the-wild, the produced touchscreen-based symptoms estimates showed good discrimination performance (0.84 AUC) on a self-reported cohort.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Screening of Parkinsonian subtle fine-motor impairment from touchscreen typing via deep learning

Iakovakis

Klingelhoefer

Bostantjopoulou

et al. 2020

Sci Rep

Self Cite

View full text Add to dashboard Cite

Fine-motor impairment (FMI) is progressively expressed in early Parkinson’s Disease (PD) patients and is now known to be evident in the immediate prodromal stage of the condition. The clinical techniques for detecting FMI may not be robust enough and here, we show that the subtle FMI of early PD patients can be effectively estimated from the analysis of natural smartphone touchscreen typing via deep learning networks, trained in stages of initialization and fine-tuning. In a validation dataset of 36,000 typing sessions from 39 subjects (17 healthy/22 PD patients with medically validated UPDRS Part III single-item scores), the proposed approach achieved values of area under the receiver operating characteristic curve (AUC) of 0.89 (95% confidence interval: 0.80–0.96) with sensitivity/specificity: 0.90/0.83. The derived estimations result in statistically significant ( ) correlation of 0.66/0.73/0.58 with the clinical standard UPDRS Part III items 22/23/31, respectively. Further validation analysis on 9 de novo PD patients vs. 17 healthy controls classification resulted in AUC of 0.97 (0.93–1.00) with 0.93/0.90. For 253 remote study participants, with self-reported health status providing 252.000 typing sessions via a touchscreen typing data acquisition mobile app (iPrognosis), the proposed approach predicted 0.79 AUC (0.66–0.91) with 0.76/0.71. Remote and unobtrusive screening of subtle FMI via natural smartphone usage, may assist in consolidating early and accurate diagnosis of PD.

show abstract

Section: Resultsmentioning

confidence: 72%

Section: Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Screening of Parkinsonian subtle fine-motor impairment from touchscreen typing via deep learning

Iakovakis

Klingelhoefer

Bostantjopoulou

et al. 2020

Sci Rep

Self Cite

View full text Add to dashboard Cite

show abstract

“…Similarly, Iakovakis et al (Iakovakis et al 2018a) explored in-the-clinic assessment to classify early PD patients and controls. In this study, estimation of individual fine-motor impairment severity scores is employed to interpret the footprint of specific underlying symptoms, such as brady-/hypokinesia (B/H-K) and rigidity (R), to keystroke dynamics that cause group-wise variations.…”

Section: Related Workmentioning

confidence: 99%

Early diagnosis of Parkinson’s disease using EEG, machine learning and partial directed coherence

et al. 2020

View full text Add to dashboard Cite

Background Parkinson's disease (PD) is a neurodegenerative disease, which has an upward progression. In advanced stages, motor symptoms cause functional impairment to patients due to the degeneration of the substantia nigra. In early stages of PD, there is a sensory impairment, and patients report visual processing dysfunction. There is still no cure for PD, and early diagnosis is essential to slow disease progression. New method Given the good anatomical representation and organization of the visual system in the cerebral cortex, in this study, we propose a biomarker of PD using EEG signals, photic stimulation, partial directed coherence (PDC) to perform feature extraction, and machine learning (ML) techniques. Our goal is to classify participants into three distinct groups: PD patients who are medicated; patients with PD and drug deprivation; and healthy subjects. Results We were able to achieve outstanding results, above 99% of accuracy and kappa statistic up to 0.98 using random forests and feature selection techniques. Comparison with existing methods: Our approach was evaluated using several ML methods. As features, we initially used the electrodes, without explicitly extracting feature vectors over signal samples. Conclusions The good results we obtained by using random forests made possible clinical applications for the early detection of PD and, consequently, better prognosis and patient's quality of life.

show abstract

“…In recent years, many researchers have thought of techniques for diagnosing PD which do not entail trained specialists or a complex medical environment. In some studies, the authors use touch screen data from several users of smart phones in order to detect PD [6,7]. Meanwhile in others, the authors use drawing movements as a biomarker to determine the motor impairments related to early PD symptoms [8].…”

Section: Introductionmentioning

confidence: 99%

Detection of early Parkinson’s disease with wavelet features using finger typing movements on a keyboard

2020

View full text Add to dashboard Cite

This study presents a new Parkinson's disease diagnosis technique based on wavelets extracted features and machine learning paradigms. The present-day diagnosis techniques suffer from low diagnosis accuracy and also require the patient to go to a medical facility, where the diagnosis is done by a specialist. In this work, we propose an automatic diagnosis method where by, all the patient has to do is to type some keys on their keyboard, and the algorithm will calculate the latency time, flight time and hold time of each key pressed, to make a diagnosis of Parkinson's disease. We use several wavelets to extract some features that are classified into Parkinson's disease or non-Parkinson's disease. The results are very encouraging and we obtain a classification accuracy of up to 100% in some of the cases, using a tenfold crossvalidation technique. Wavelets are a tool that can be used to complement and improve the detection of Parkinson's disease. These results will permit the amelioration of some state-of-the-art methods which use a similar technique to detect Parkinson's disease.

show abstract

Motor Impairment Estimates via Touchscreen Typing Dynamics Toward Parkinson's Disease Detection From Data Harvested In-the-Wild

Cited by 30 publications

References 28 publications

Screening of Parkinsonian subtle fine-motor impairment from touchscreen typing via deep learning

Screening of Parkinsonian subtle fine-motor impairment from touchscreen typing via deep learning

Early diagnosis of Parkinson’s disease using EEG, machine learning and partial directed coherence

Detection of early Parkinson’s disease with wavelet features using finger typing movements on a keyboard

Contact Info

Product

Resources

About