We aim to develop a non-invasive, low-cost pre-clinical biomarker for synucleinopathies such as Parkinson’s disease (PD) or Lewy body dementia (LBD), which will have real impact on both early treatment and drug development. Our hypothesis is that REM Sleep Behavior Disorder (RBD) patients who years later develop either PD or LBD present distinctive electroencephalography (EEG) alterations at the initial diagnostic exploration. EEG tracks brain wave activity. Our objective is to develop tools for the automatic classification of RBD patients according to the alterations in the EEG at group and individual levels.
We propose to retrospectively analyze spontaneous, waking EEG as recorded during the RBD diagnostic exploration of previous studies at the Center for Advanced Research in Sleep Medicine of Hospital Sacré-Coeur Montréal. Spectral features will be first extracted from EEG signals. A subset of the resulting data will be selected to identify classifiers for individual categorization, which will then be applied to the remaining data. We will explore how to generalize a biomarker, which works at a group level as we have shown in a pilot study, to individuals and how to formally evaluate performance. Currently no optimal approaches for performance evaluation exist, since de facto standards in machine learning are often over optimistic and do not apply statistical corrections.
Relevance to Diagnosis/Treatment of Parkinson’s Disease:
Prevention methods for Parkinson's and new treatments for PD patients critically depend on the identification of at-risk subjects, the characterization of pre-clinical brain alterations and the availability of evolution criteria. The proposed strategy addresses these issues via an EEG-based system that is potentially easy to use and low cost. Identification of EEG alterations will aid in characterizing pre-clinical brain changes in PD and allow earlier clinical trials of new PD therapies.
If successful, together with later multi-center and longitudinal controlled trials, our methodology will lead to the development of an integrated classification tool. The performance and limits of this tool will be defined within this project and a prototype will be delivered. Successful criteria in both cases will provide individual classification power similar to our pilot results with statistical validation.
We started with a group of 118 subjects (half control volunteers, half people with REM Behavior Disorder or RBD), where 80 percent of the RBD patients developed either Parkinson’s disease (PD) or Lewy body dementia (LBD) at a follow-up of eight years. The project attained classification at an individual level for different diagnosis and prognosis tools based on electroencephalography (EEG) data analysis. The EEG was collected in every patient at baseline, i.e. when they were still idiopathic RBD. We developed a classification system and a procedure for performance evaluation close to operational conditions.
Classification shows excellent performance levels in terms of Area Under the Curve 94-98% in all problems of operational relevance, i.e. RBD diagnosis, synucleinopathy prognosis and PD prognosis. The classification degrades when including a large group of RBD patients with a short follow-up (one year). The project confirms, therefore, recent research findings, which underscore the time to conversion as the key feature in idiopathic RBD as a biomarker of future synucleinopathies like Parkinson’s. With these results we take an important step towards the development of non-invasive, very low-cost pre-clinical markers for synucleinopathies (PD or LBD) and its utilization for conversion prognosis, which will have real impact on both early treatment and drug development options.