Cell Nuclei Segmentation Combining Multiresolution Analysis, Clustering Methods and Colour Spaces

In this paper a new method for medical images analysis has been proposed. It is based in a multiresolution schema in combination with a k-means clustering algorithm. The edge detection and classification schema is based on the analysis of the data obtained by a multiresolution image analysis (MRA) using Mallat and Zhong's wavelet. The edge detection and classification algorithm developed has been tested defining five contour types: step, ramp, stair, pulse and 'noise'. The cell nuclei presented in medical images can be perfectly isolated with the help of the 'cellular nucleus' contour, a special noise reduction achieved by means of the previous classification schema and a segmentation process provided by a k-means algorithm. We have proposed an algorithm to estimate the number of cells appearing in tissue samples, as well as the estimate of positivity levels in tumour tissues. This is part of a software tool for tumour detection and diagnosis of diseases.