Segmentation of lung fields in digital chest radiographs by artificial neural networks

Lung field segmentation is a basic step for virtually any quantitative procedure. In this view, due to the imaging process and the complexity of the imaged district, an efficient use of prior anatomical knowledge is crucial. In this report we describe a new approach to lung field segmentation which is based on fuzzy boundary modeling and a neural network architecture including supervised multilayer networks and topology preserving maps. In this report we describe a new approach to lung field segmentation which is based on fuzzy boundary modeling and a neural network architecture including supervised multilayer networks and topology preserving maps.