Generalizable electroencephalographic classification of Parkinson's disease using deep learning
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Growing interest surrounds the use of electroencephalography (EEG) and deep learning for diagnosing neurological conditions like Parkinson's Disease (PD). Despite the existing proof-of-concept literature demonstrating the potential of deep learning in classifying PD from EEG data, neurologists have been slow to adopt these tools due to insufficient evidence of their real-world diagnostic generalizable performance. Our study aimed to evaluate the potential of deep learning for inter-subject PD classification using a conservative training approach and testing on an external independent dataset. Specifically, we utilized publicly available resting-state EEG data from PD patients at two separate centers, the University of New Mexico (n = 54) and the University of Iowa (n = 28), as our training and testing sets, respectively. Each of these recordings had a minimum of 2 min of data. We implemented a channel-wise convolutional neural network, tuning it with a leave-one-subject-out cross-validation approach. Our approach achieved a patient-level accuracy of 80.4% (epoch-level accuracy = 72.7%), which remained consistent when tested on the external dataset (patient-level accuracy = 82.8%, epoch-level accuracy = 75.7%). Our model performs equal-or-better than other standard classification models and our approach compares favourably to similar works. Our publicly available code serves as a foundation for future research exploring different deep learning architectures, investigating other pathologies, and involving larger datasets with the hope of accelerating the adoption of objective computational approaches for the diagnosis and monitoring of neurological disorders.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.002 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it