COVID-19 and Parkinson’s disease are emergency problems on a large scale that require fast care. Therefore, developing methodologies to diagnose these diseases is imperative. Wearables, such as smartwatches, and solutions based on mobile health (mHealth) and electronic health (eHealth) concepts could be justified answers for these challenges. They could also support ‘enabling aging in place’.
In her doctoral dissertation Justyna Skibińska introduces solutions which can potentially effectively detect the COVID-19 cases at an early stage, and thereby limit the number of infected people. Moreover, she presents the methodologies of Parkinson’s disease recognition on the combined utilisation of video and audio analysis, considering hypomimia and hypokinetic dysarthria (HD) symptoms, respectively.
“It was possible to distinguish people infected with COVID-19 from representatives of healthy control group in the early stage of illness in 78 % cases based on wearable records making use of the k-nearest neighbours (k-NN) algorithm,” said Skibińska.
She utilises a publicly available dataset in her work. Previous work on the dataset (by Mishra et. all) involved just an anomaly detection-based solution, while not considering specificity. In addition to the above, Skibińska and her collaborators introduce the classification of COVID-19 cases and healthy control group.
“The target of the designed methodology was to identify COVID-19 cases two days before the visible onset of the disease before the highest contagiousness period. The positive consequence of this approach will be the limitation of the number of infected people.”
In her doctoral dissertation she also introduces a multimodal approach to detecting Parkinson’s disease. Using audio and video modalities and machine learning approaches allows the development of a support system methodology. The prediction of Parkinson’s disease achieved a balanced accuracy of 83 % for the general model. The multimodal approach outperforms solutions based on a single modality thanks to the XGBoost classifier.
43 various Czech speech exercises were utilised in her thesis to identify Parkinson’s disease. The most accurate occurred to be one of tongue twisters. The original sentence in Czech is: “Celý večer se učí sčítat.” The meaning of the sentence is: “He’s been learning to count all night”, but rather than its meaning it is important that the sentence is hard to pronounce. The prediction model based on this speech task and multimodality achieved 74 % balanced accuracy.
“This work explored the clinical value, and various speech exercises power in the prediction of Parkinson’s disease with the potential to be applied as a mHealth solution. This proposed solution could facilitate the lives of Parkinson’s disease patients, their families, and doctors likewise limiting the burden of the healthcare system,” Skibińska remarked.
Justyna Skibińska conducted her doctoral research during 2019–2022 in a joint/double doctoral degree programme at Brno University of Technology, Czech Republic and Tampere University, Finland. It was funded by the European Union’s Horizon 2020 Research and Innovation programme under the Marie Skłodowska-Curie grant agreement No. 813278, A-WEAR.
Public defense on 4 December 2023
The doctoral dissertation of M.Sc. Eng. Justyna Skibińska in the field of Electronics and Information Technologies titled Machine Learning-Aided Monitoring and Prediction of Respiratory and Neurodegenerative Diseases Using Wearables will be publicly examined at the Faculty of Electrical Engineering and Communication, Brno University of Technology (address: Technická 12, 616 00, Brno, Czech Republic) at 10.30 a.m. (CET) on Monday 4 December 2023.
The Opponent will be Prof. dr. Peter Peer, University of Ljubljana, Slovenia. The Custos will be Assoc. Prof. Jiri Mekyska, Brno University of Technology, Czech Republic.
Photograph: Iwona Liszcz
Brno University of Technology, Czech Republic