A Neural Network System for Detecting Lung Nodules in Chest Radiograms

A system for detecting small lung nodules is described. According to computer vision paradigms, the system is modular and based on a multi-scale approach. An attention focusing mechanism (operating at low spatial resolution) locates image regions with possible lesions. A second module analyzes (at full spatial resolution) the regions around the attention foci. Finally, a classification module estimates the probability of lesion presence.