The invention concerns a computer-implemented method for segmenting a 3D image comprising round objects, the method comprising the steps of dividing the 3D image into 2D slice images, for each slice image, determining seed points within round areas of the slice image having brightness different than the background of the image and processing the determined seed points by selecting a starting seed point, for the starting seed point, determining a contour of a round area surrounding the starting seed point within its slice image plane, selecting neighboring seed points on other slice images that are in line with the starting seed point within the segmentation plane, determining contours of the round areas surrounding the neighboring seed points within their slice image planes based on the contours of the round area surrounding the starting seed point or another neighboring seed point, until all neighboring seed points are processed, combining the determined contours of the round areas associated with the first seed point and its neighboring seed points into a contour of a 3D round object.