Reliability of a computer-aided detection system in detecting lung metastases compared to manual palpation during surgery

Abstract

For resection of lung metastases computed tomography (CT) is needed to determine the operative strategy. A computer-aided detection (CAD) system, a software tool for automated detection of lung nodules, analyses the CT scans in addition to the radiologists and clearly marks lesions. The aim of this feasibility study was to evaluate the reliability of CAD in detecting lung metastases. Preoperative CT scans of 18 patients, who underwent surgery for suspected lung metastases, were analysed with CAD (September-December 2009). During surgery all suspected lesions were traced and resected. Histological examination was performed and results compared to radiologically suspicious nodes. Radiological analysis assisted by CAD detected 64 nodules (mean 3.6, range 1-7). During surgery 91 nodules (mean 5.0, range 1-11) were resected, resulting in 27 additionally palpated nodules. Histologically all these additional nodules were benign. In contrast, all 30 nodules shown to be metastases by histological studies were correctly described by CAD. The CAD system is a sensible and useful tool for finding pulmonary lesions. It detects more and smaller lesions than conventional radiological analysis. In this feasibility study we were able to show a greater reliability of the CAD analysis. A further and prospective study to confirm these data is ongoing. © 2011 Published by European Association for Cardio-Thoracic Surgery.