Presentation

Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Lightning Talks

Classification between PPA, MCI and Healthy Controls using EEG and Artificial Intelligence

Poster A53 in Poster Session A, Tuesday, October 24, 10:15 am - 12:00 pm CEST, Espace Vieux-Port

Panteleimon Chriskos¹, Christos Frantzidis^1,2, Alexandros Afthinos³, Jessica Gallegos³, Brenda Rapp⁴, Panagiotis Bamidis¹, Kyrana Tsapkini^3,4; ¹Laboratory of Medical Physics and Digital Innovation, School of Medicine, Faculty of Health Sciences, Aristotle University of Thessaloniki, Thessaloniki, Greece, ²School of Computer Science, University of Lincoln, Lincoln, United Kingdom, ³Department of Neurology, Johns Hopkins School of Medicine, Baltimore, MD, USA, ⁴Department of Cognitive Science, Johns Hopkins University, Baltimore, MD, USA, ⁵Department of Physical Medicine and Rehabilitation, Johns Hopkins School of Medicine, Baltimore, MD, USA

Primary progressive aphasia (PPA) is a rare neurodegenerative disorder in which language impairments are the primary symptom. Diagnosis is very challenging and requires an expert physician and a combination of methods often involving expensive imaging modalities such as MRI or PET scans. Previous research has shown the importance of connected speech in PPA subtyping[1] and we have recently employed Natural Language Processing and Machine Learning methodologies for subvariant diagnosis[2]. However, early differential diagnosis between PPA and other neurodegenerative disorders[3], e.g., mild cognitive impairment (MCI), is even more challenging due to similar clinical profiles initially. Therefore, it would be beneficial if diagnosis could be assisted with alternative methods such as low-density EEG and suitable functional connectivity metrics, which would sufficiently differentiate between healthy individuals and individuals with MCI and PPA. To achieve accurate healthy, MCI and PPA patient classification, we used 3-minute recordings of 8-channel (F7, F8, T7, T8, CP3, CP4, P5 and P6) EEG signals recorded during an eyes-closed resting state session. The data were collected from 8 healthy elderly control participants (HC), 8 MCI and 14 PPA patients (9 lvPPA, 5 nfvPPA). The recorded data were re-referenced using the common average method, and subsequently preprocessed to remove noise using a set of second order Butterworth filters. (Non-)Linear were removed using least-squares fit method. The data were separated into 4096 sample epochs resulting in a total of 974 epochs (361 healthy, 319 MCI, 294 PPA). The Relative Wavelet Entropy (RWE) method was used to calculate the functional connectivity matrix, based on the Orthogonal Wavelet Transform using Morlet wavelets and the Shannon entropy. Classification was conducted using the k-Nearest Neighbor (kNN) and Support Vector Machine (SVM) classifiers. The training set contained 705 epochs (270 healthy, 225 MCI, 210 PPA) and the test set 269 epochs (91 HC, 94 MCI, 84 PPA). Classification was evaluated on a participant level (epoch voting). Three classification experiments were conducted using various classifier parameters. The highest accuracy rates on the test set reported are: (a) for HC-MCI 100.00% using the kNN classifier with Cityblock distance and, k=1, (b) for HC-PPA 100.00% using an SVM with an RBF kernel and σ=0.7, and, (c) for HC-MCI-PPA 77.78% using an SVM with an RBF kernel and σ=1.38. Three minutes of low-density EEG using functional connectivity metrics can provide adequate accuracy classification between the healthy, MCI and PPA. We are in the process of automating this analysis now. These results indicate that this low-cost, easy method can be used by clinicians to assist the differential diagnosis of PPA vs. other neurodegenerative disorders. [1]Wilson, Stephen M., et al. "Connected speech production in three variants of primary progressive aphasia." Brain 133.7 (2010): 2069-2088. [2]Themistocleous, Charalambos, et al. "Automatic subtyping of individuals with Primary Progressive Aphasia." Journal of Alzheimer's Disease 79.3 (2021): 1185-1194. [3]Frantzidis, Christos A., et al. "Cognitive and physical training for the elderly: Evaluating outcome efficacy by means of neurophysiological synchronization." International Journal of Psychophysiology 93.1 (2014): 1-11.

Topic Areas: Disorders: Acquired, Computational Approaches