PUBLICATIONS

Abstract: Artificial intelligence has emerged as an essential tool for investigating neurological diseases, driven by the rapid expansion of neuroimaging, clinical, physiological, and wearable data. This systematic review, based on a structured survey of peer-reviewed literature identified through major biomedical and engineering databases, summarizes advancements in machine learning and deep learning associated with Alzheimer’s disease, stroke, Parkinson’s disease, brain tumors, and traumatic brain injury, with emphasis on data modalities, publicly accessible datasets, model architectures, and performance trends reported in recent studies. Our analysis reveals that convolutional and encoder--decoder architectures dominate imaging-related tasks across these diseases, while hybrid and multimodal frameworks increasingly integrate neuroimaging with clinical, electrophysiological, speech, gait, and sensor data. Emerging paradigms, including federated learning, self-supervised representation learning, and foundation models, demonstrate potential for addressing data scarcity, privacy constraints, and inter-institutional variability; however, substantial challenges in generalizability, interpretability, and clinical translation persist. This review synthesizes evidence across neurological domains to clarify ML and DL’s current strengths and limitations in brain care and outlines key research directions toward robust, interpretable, and clinically actionable decision-support systems.

Abstract: This work investigates the second-order partitioned time-stepping method for the sophisticated multiphysics parabolic model problem. In this paper, we consider a coupled system of heat equations through two adjacent materials which are coupled across their shared and rigid interface with two interface conditions. We perform the variational formulation of the heat-heat coupled fluid flow model and report the well-posedness of the model. On the other hand, efficient second-order Crank-Nicolson and second-order implicit-explicit finite element discretized algorithm is proposed to solve the parabolic two domain problems numerically. We conduct several numerical tests to achieve optimal convergence order. We also designed several conceptual model problems to demonstrate the validity, accuracy, and efficiency of the heat-heat multiphysics model problem. The applicability and complicated flow characteristics are shown by illustrating the heat flux, conduction of the heat and contour plots in the conjugate computational domain.