Gravity Segmentation of Human Lungs from X-ray Images for Sickness Classification

Automated classification of patients based on the analysis of their medical images will aid in directing the attention of medical professionals to cases that require the most urgent attention. Our current work has been focused on the segmentation of lung areas from an X-ray image of human lungs. We use the Watershed segmentation with a new distance measure, called “gravity distance”, which is derived from a field based on a relief function. The relief field has similar properties to those of gravity, including potential energy. When compared to the Watersnakes distance, the gravity measure has a stronger physical meaning. With the gravity distance, the Watershed segmentation obtained by region growing is equivalent to that with energy minimization. We also introduce a method to control the smoothness of the Watershed line. Preliminary results obtained from the segmentation of human lungs’ X-ray images are very encouraging.