Background: With the increasing incidence of lung cancer, it is prudent to do screenings for individuals at high-risk. Solitary pulmonary nodules (SPNs) are an indication of small tumors or early stages of disease. Therefore, accurate detection of SPNs is important to both clinicians
and radiologists. Since a large number of computed tomography (CT) scans are being acquired during a lung cancer screening, there is an urgent need for new automated techniques to detect SPNs. Methods: A novel algorithm for segmentation of SPNs in CT scans based on three-dimensional
connected voxels (3DCVs) can be used to screen out potential patients with SPNs. 120 cases of CT scans from a public database (100 positive cases with nodules and 20 negative cases without nodules) and 30 negative cases from the routine CT scans completed in a hospital were used to test the
algorithm. The algorithm is based on the fact that most pulmonary nodules are solitary at their early stages. First, find suitable CT values thresholds for CT values to convert pulmonary nodules, normal tissues and air spaces in each chest CT slice into black and white images. Then stack the
slices in their originally physical order. This will produce a three-dimensional (3D) matrix with pulmonary nodules and normal tissues constructing their own 3DCVs respectively. Results: Of the 100 positive cases, 93 cases showed positive detection of SPNs and 7 cases did not. Of the
50 negative cases, 48 cases returned a negative result and 2 cases showed as positive result. In this study, the sensitivity is 93% and the specificity is 96% with a 4% false positive rate (FPR). Conclusions: This algorithm can be used to screen out positive chest CT scans efficiently,
which will increase efficiency by two to three times than when compared with manual inspection and detection.
Standardized uptake value and radiological density attenuation as predictive and prognostic factors in patients with solitary pulmonary nodules: our experience on 1,592 patients