Evaluation of algorithm for extraction of lung regions in CT exams.

Recently has grown the development of Computer-aided Detection Systems - CAD to improve the diagnosis of diseases identifying it at the initials stages using medical images. In this work is made a analysis of the performance of an algorithm for lung region extraction in computed tomography exams - CT. The implementation was made in MATLAB and applied to 9 CT scans of patients with lung disease (corresponding to a set of 479 images). The detection of image lung boundaries were classified by an expert radiologist as "Good" and "Poor" according the presence of errors. The results showed deficiencies which damage the algorithm performance, only 52% of the images were rated as "Good" . The problems identified were listed, and when resolved will give the needed quality to the process to be used by radiologist doctors in the patient care.