Analysis of adductors angle measurement in Hammersmith infant neurological examinations using mean shift segmentation and feature point based object tracking

Abstract

This paper presents image and video analysis based schemes to automate the process of adductors angle measurement which is carried out on infants as a part of Hammersmith Infant Neurological Examination (HINE). Image segmentation, thinning and feature point based object tracking are used for automating the analysis. Segmentation outputs are processed with a novel region merging algorithm. It is found that the refined segmentation outputs can successfully be used to extract features in the context of the application under consideration. Next, a heuristic based filtering algorithm is applied on the thinned structures for locating necessary points to measure adductors angle. A semi-automatic scheme based on the object tracking of a video has been proposed to minimize errors of the image based analysis. It is observed that the video-based analysis outperforms the image-based method. A fully automatic method has also been proposed and compared with the semi-automatic algorithm. The proposed methods have been tested with several videos recorded from hospitals and the results have been found to be satisfactory in the present context.