Dissertation: AI analysis of brainwave tests helps detect Parkinson’s and Alzheimer’s disease
Neurodegenerative diseases such as Parkinson’s disease and Alzheimer’s disease are becoming more common as populations age. Today, diagnosis often relies on specialised imaging and clinical assessments that are expensive, slow and usually performed only after symptoms are clearly visible. There is a need for simple tools that could support earlier detection and follow-up.
In his doctoral dissertation Mingliang Zhang shows that short, non-invasive brainwave recording (resting-state EEG) combined with artificial intelligence can help identify people with Parkinson’s or Alzheimer’s disease from healthy older adults.
EEG is inexpensive and widely available, but its signals are often noisy and vary from person to person, which has limited its use in everyday clinical work.
“My research explores how raw EEG signal data can be transformed into compact images that show how brain activity changes over time and across scalp regions. It also focuses on training lightweight neural networks to recognise patterns typical for Parkinson’s disease, Alzheimer’s disease and healthy brain function”, Zhang says.
All methods are evaluated at the subject level, meaning that participants are completely separate from the training data, giving results that better reflect real clinical conditions.
Using this framework, Zhang has developed a rapid screening method for Parkinson’s disease based on short resting EEG recordings. He showed that the method can reliably detect patients with Parkinson's disease or Alzheimer's disease.
Zhang also created region-aware representations that highlight which parts of the scalp carry the most useful information for detecting diseases. For Parkinson’s disease, the central sensorimotor area, which is also essential for movement control, proved especially important. The same approach was adapted to frontal EEG recordings for Alzheimer’s disease to capture disease-related changes in brain rhythms.
In another part of his research, Zhang explored ways for hospitals to collaborate without moving sensitive patient data.
Zhang used a technique called federated learning, in which AI models are trained in parallel on multiple datasets. Each hospital trains its own model with its own EEG data, and when the model is updated, only the model’s learning results are updated centrally among the collaborating hospitals. This allows hospitals to benefit from larger combined datasets while still respecting patient privacy and strict data protection rules.
Overall, the results suggest a realistic pathway from routine EEG recordings to AI-assisted tools for the detection and monitoring of neurodegenerative diseases.
“In the future, such tools could complement existing clinical assessments, support earlier identification of patients and enable more personalised follow-up, ultimately supporting better treatment outcomes and quality of life”, Zhang concludes.
Mingliang Zhang defends his doctoral dissertation “Time–Frequency Deep Learning Framework for EEG-Based Neurodegenerative Disease Detection” on 16 December 2025 at the Faculty of Information Technology, University of Jyväskylä. The opponent is Professor Karen Egiazarian (Tampere University), and the custos is Professor Tommi Kärkkäinen (University of Jyväskylä).
The language of the event is English. The dissertation event can be attended in Agora Auditorium 2 or online.
