Characterization of Time Evolving Graph Using State-Space Modelling and its Application in Alzheimer's Disease Detection

Early Detection of Alzheimer’s Disease Using Graph Signal Processing and Deep Learning

“…First, we describe in brief our previously proposed static connectivity based AD detection model [21,22]. Then we describe our novel dynamic connectivity based AD detection model, the preliminary version of which is presented in [28]. Finally, we explain our proposed integrated AD detection model wherein we combine the static and the dynamic graph connectivity based features extracted from the above models using the intermediate level integration approach.…”

“…Also, contemporary research in the field of time evolving GSP has discovered the correlation between the alterations in graph connectivity variation and the AD, thus giving rise to a new disease indicator [25][26][27]. To utilize the information in the brain connectivity variation for AD detection, in our previous work [28], we proposed a novel SSM based method to characterize the alterations in the graph connectivity matrices and used it to design a dynamic connectivity model for early diagnosis of AD. In this work, we design a modified dynamic connectivity based AD diagnosis model wherein the discriminating features extracted using the SSM based method are classified using a properly designed convolutional neural network classifier to improve the model performance.…”

“…Recently, the researchers working in the paradigm of time varying graph connectivity have discovered the alterations in the dynamics of brain graph connectivity [25][26][27] in case of AD, which makes it a plausible biomarker for early diagnosis of AD. So, to make use of the information present in the graph connectivity variation of a time varying graph for AD detection, in our previous work [28], we designed a dynamic connectivity based AD detection model wherein a novel state-space modelling (SSM) based method was proposed to characterize the time varying nature of the graph. In the present work, we first modify this basic dynamic connectivity based AD detection model wherein we design a more sophisticated deep neural network based classifier for classifying the SSM based features.…”