Final Results and Data Published
The final Machine Learning results of our pilot study have been published on npj Parkinson’s Disease. Moreover, all relevant raw data and algorithms have been documented and published Open Access without the need for registration. This ensures easy Access and full reproducibility.
Analyses on Spiral Drawings Published
Looking for Parkinson’s Disease specific biomarkers in Tablet Drawings. Paper on two-sided tablet drawings by patients with Movement Disorders accepted in Scientific Reports: https://www.nature.com/articles/s41598-023-37388-3
Voice Analyses Results Published
Paper accepted for MIE 2023 conference in Göteburg: Brenner, A.; Van Alen, C. M.; Plagwitz, L.; Varghese, J. Classification of Parkinson’s Disease from Voice – Analysis of Data Selection Bias. Stud Health Technol Inform 2023, 302, 127–128. DOI: 10.3233/shti230079.
Submissions to MIE 2022 accepted
Both paper submissions to the MIE 2022 – the official European conference for Medical Informatics – have been accepted. Our team will present results on the machine learning impact of the PD-NMS questionnaire for Parkinson’s diagnosis and interpretability measures. See you in Nice, France! 28-05-2022, 09:45, Session 4: Supporting AI-Explainability by Analyzing Feature Subsets in…
Read more
Recruitment of Pilot Study Finished
Despite the corona pandemic we met our minimum recruitment target of 500 participants: We have measured 518 participants, including 282 PD participants, 24 patients with Essential Tremor, 97 with other differential diagnoses and 101 Healthy controls. The next months will focus on data cleaning and first analyses results of our AI-based system.
Winner of the Machine Learning Challenge 2021
Tim Segger and Cedric Vadder (middle of photo) implemented high precision models to distinguish Parkinson’s Disease from other movement disorders. Their model performed best and most robust in the project seminar at the Department of Computer Science, University of Münster (Accuracy >82%, SD <0.02). Congratulations!
Paper on Sensor Validation & Machine Learning published
Smartwatches are capable of measuring very subtle tremor phenomena almost at the level of seismometers! Our Open Access Paper on sensor validation and preliminary Machine Learning results is published by Sensors, the leading open access journal on the science and technology of sensors. The work resulted from a close collaboration with the Institute of Geophysics.…
Read more
More than 500 participants measured
Though the recruitment was persistently limited during the pandemic, we could still carry on to assess further participants with our new system (see figure). The App-based data capture und continuous data quality monitoring ensured 100% data completeness. We are going to soon submit further preliminary data analyses results to a special issue by Sensors.
Eight new students participating in intense Machine Learning Course for Winter term 2020/2021
The course is a 10-15 ECTS points project seminar for Computer Science students at the University of Münster. This time, we are focusing on promising Methods of Ensemble Learning with Stacking and Deep Neural Networks. We are looking forward for new interesting algorithms and of course the best-performing team to beat accuracies from last course.
Winner of the Machine Learning Challenge 2020
Our students Luisa Beerboom and Alexander Leifhelm implemented high precision models classifying Parkinson’s disease vs other movement disorders and healthy samples. Their model performed best and most robust in our intense project seminar at the Department of Computer Science, University of Münster (Accuracy >80%, SD <0.03). Congratulations!