Muhammad Laiq Ur Rahman Shahid

• Obstructive sleep apnea (OSA) is a public health problems affecting at least 2-4% of the middle-aged population and characterized by repeated collapsing of the soft tissue structures of the throat. However, the pathogenesis of this problem is poorly understood. Volumetric analysis of the upper airway and para-pharyngeal fat pads can help us to study the OSA syndrome. A reliable automatic segmentation technique plays a vital role in investigating the OSA. The proposed context-based automatic algorithm uses prior knowledge about the topology and geometry of the pharynx. The comprehensive segmentation pipeline consists of three major stages for the delineation of pharyngeal structures from the rest of the soft tissues of the head MRI. The first stage consists of a coarse segmentation of soft tissues of the head MRI using intensity clustering. The second stage involves the feature extraction of the coarse objects and designs a supervised classifier using the interactive visual analysis of the multidimensional feature space. The third stage comprises of the refinement on the local region using multiOtsu thresholding and watershed transform. The approach is quantitatively evaluated against the ground truths of ten datasets resulting in an average of approximately 90% Dice coefficient. An interactive multidimensional visual analytics tool is proposed and implemented to visualize and analyze quantitative and categorical data simultaneously. Visual analysis of quantitative and categorical variables shows that the age variable carries the highest weight in separating healthy subjects group from apneic patients in 2D projection space. Other worth mentioning variables are BMI, gender, blood pressure, minimum pharyngeal laterolateral axial length and parapharyngeal fat pads volume etc. This multidimensional visual analytics tool helps us to study the impact of numerous features and variables to investigate the OSA syndrome.