Deep Learning-based Classification of Dementia using Image Representation of Subcortical Signals
Authors:
Shivani Ranjan,
Ayush Tripathi,
Harshal Shende,
Robin Badal,
Amit Kumar,
Pramod Yadav,
Deepak Joshi,
Lalan Kumar
Abstract:
Dementia is a neurological syndrome marked by cognitive decline. Alzheimer's disease (AD) and Frontotemporal dementia (FTD) are the common forms of dementia, each with distinct progression patterns. EEG, a non-invasive tool for recording brain activity, has shown potential in distinguishing AD from FTD and mild cognitive impairment (MCI). Previous studies have utilized various EEG features, such a…
▽ More
Dementia is a neurological syndrome marked by cognitive decline. Alzheimer's disease (AD) and Frontotemporal dementia (FTD) are the common forms of dementia, each with distinct progression patterns. EEG, a non-invasive tool for recording brain activity, has shown potential in distinguishing AD from FTD and mild cognitive impairment (MCI). Previous studies have utilized various EEG features, such as subband power and connectivity patterns to differentiate these conditions. However, artifacts in EEG signals can obscure crucial information, necessitating advanced signal processing techniques. This study aims to develop a deep learning-based classification system for dementia by analyzing scout time-series signals from deep brain regions, specifically the hippocampus, amygdala, and thalamus. The study utilizes scout time series extracted via the standardized low-resolution brain electromagnetic tomography (sLORETA) technique. The time series is converted to image representations using continuous wavelet transform (CWT) and fed as input to deep learning models. Two high-density EEG datasets are utilized to check for the efficacy of the proposed method: the online BrainLat dataset (comprising AD, FTD, and healthy controls (HC)) and the in-house IITD-AIIA dataset (including subjects with AD, MCI, and HC). Different classification strategies and classifier combinations have been utilized for the accurate mapping of classes on both datasets. The best results were achieved by using a product of probabilities from classifiers for left and right subcortical regions in conjunction with the DenseNet model architecture. It yields accuracies of 94.17$\%$ and 77.72$\%$ on the BrainLat and IITD-AIIA datasets, respectively. This highlights the potential of this approach for early and accurate differentiation of neurodegenerative disorders.
△ Less
Submitted 20 August, 2024;
originally announced August 2024.
