Circular and oval-like objects are very common in cell and micro biology. These objects need to be analyzed, and to that end, digitized images from the microscope are used so as to come to an automated analysis pipeline. It is essential to detect all the objects in an image as well as to extract the exact contour of each individual object. In this manner it becomes possible to perform measurements on these objects, i.e. shape and texture features. Our measurement objective is achieved by probing contour detection through dynamic programming. In this paper we describe a method that uses Hough transform and two minimal path algorithms to detect contours of (ovoid-like) objects. These algorithms are based on an existing grey-weighted distance transform and a new algorithm to extract the circular shortest path in an image. The methods are tested on an artificial dataset of a 1000 images, with an F1-score of 0.972. In a case study with yeast cells, contours from our methods were compared with another solution using Pratt’s figure of merit. Results indicate that our methods were more precise based on a comparison with a ground-truth dataset. As far as yeast cells are concerned, the segmentation and measurement results enable, in future work, to retrieve information from different developmental stages of the cell using complex features.

Date Published: 16 April 2014
PDF: 5 pages
Proc. SPIE 9159, Sixth International Conference on Digital Image Processing (ICDIP 2014), 915903 (16 April 2014); doi: 10.1117/12.2063934

Published in SPIE Proceedings Vol. 9159:
Sixth International Conference on Digital Image Processing (ICDIP 2014)
Charles M. Falco; Chin-Chen Chang; Xudong Jiang, Editor(s)