Segmentation of Heart Image Sequences Based on Human Way of Recognition - Brain Informatics - page 227

a) b)

c) d)

Fig. 2. Preprocessing phase: (a) - image after thresholding, (b) - potential contour of
both ventricles (scaled down image after thresholding with weak distance potential
filter), (c) - interventricular septum line (scaled down image after thresholding with
weak distance potential filter) (d) - interventricular septum (original image)

- C in , C ou - set of pixels in the image that lay inside and outside the current

contour, respectively;
- L ri ,

L le - set of pixels in the image that lay on the right and left side of the
current line separating ventricels, respectively;
- M ri,in ,

M ri,ou - set of pixels in the image that lay inside and outside the
right ventricle mask drawn by an expert, respectively;
- M le,in ,

M le,ou - set of pixels in the image that lay inside and outside the left
ventricle mask drawn by an expert, respectively;
- T i k T ou

k

- set of pixels in the image that lay inside the rectangle drawn
around the

,

k -th contour point and lay inside and outside the current contour,

respectively.
- n - function that value represents the number of pixels in a given set;

- D - function that value represents the intensity at the given point of a

processed image (the values are in range [0

, 1]) and which definition differs

between phases;

3.1 Preprocessing Phase

The first element of the proposed top-down detection process is localisation of
both heart ventricles. In this phase potential active contours are used. At this