Search

Outlier detection is a nontrivial but important task for many of the application areas. There exist several methods in literature to find outliers. However, there is no single method that outperforms in all cases. Thus, finding proper algorithm and the value of its relevant parameter is crucial. In addition, none of the methods are perfect and verification by the domain experts can confirm if the outliers detected are meaningful or not. Proper visual representation of the detected outliers may help experts to resolve anomalies. In this paper, we proposed a visual analytic system that finds proper algorithm and value of its relevant parameter for a specific dataset using training set. Later, the chosen method is applied to the test data to obtain the outlier ranking. After that, data points are visualized with parallel coordinate plot (PCP) where colors of the lines are obtained by using the outlier factor of the data points. PCP is one of the popular high-dimensional data visualization techniques where coordinates are parallel to each other and each data point is represented by a line. Using the visual, experts can provide feedback and update the result. Experiments with different datasets ensure the strength of our system.

The parallel processing of verbal elements like textual labels, annotations etc. with non-verbal i.e. visual elements, provides maximum integration of human mind. Textual labels and annotations aid in understanding medical data visualization. But it is a very challenging task to generate hand-made like illustration labels in an automated system. The paper presents a noble approach for dynamic external labeling of segmented 2D computer tomography (CT) slices. Various efficient algorithms are approached to solve labeling problems in modular form. The system emphasizes on the readability and clarity of labels while generating dynamic group labels in real time for 2D slices using external labeling layout.