Contour expansion algorithm preprocessed by hough transform and circular shortest path for ovoid objects

Circular and ovoidal objects such as cells and supramolecular complexes are very common in bio-imaging. In order to do measurements on such objects, the accurate detection of these objects from microscope images is essential. Therefore our objective is finding accurate and reliable methods to extract the contour delineating the object. In earlier work, we have shown to be successful in achieving this objective through dynamic programming, where we used Hough transform and a minimal path algorithm to estimate the contour location. However, due to the inherently fuzzy nature of edges and as microscope imaging can be very delicate, a refinement of the initial estimate is sometimes required. In these cases the minimal path is no longer the ultimate valid representation of the contour. This paper describes an expansion algorithm in the polar image that is created for each object; with this new algorithm we achieve the necessary refinement of contours. From previous results we can assess the improvement to existing segmentation methods, including our own approach. The exact contour contributes in the accomplishment of precise measurements of the objects so that machine learning techniques can better recognize the subtle patterns within the data. As the method is geared to ovoid like objects, it can be easily generalized to other image of other modalities. Here we apply the method in the domain of yeast cell biology.