Objective To propose an algorithm that can extract certain types of lung nodule from thoracic CT images in order to aid the screening and treatment evaluation of lung cancer. Methods In the algorithm，after the radiologist determines the seed point and the volume of interest，the nodule is identified as non-juxta-pleural or juxta-pleural based on the raw segmentation result. Then a hybrid method of multi-level thresholding combined with distance transform is used to extract non-juxta-pleural nodule，ray casting and line fitting are applied for juxta-pleural nodule. Finally，the datasets including 85 patients (232 nodules) are utilized to evaluate the proposed algorithm，and the accuracy is evaluated by one experienced radiologist. Results This algorithm，with strong adaptive capacity，can recognize non-juxta-pleural and juxta-pleural nodules accurately，and extract the isolated solid，juxta-vascular，ground glass opacities and juxta-pleural nodules properly. The average segmentation accuracy exceeds 90% with high segmentation speed. Conclusions This algorithm can extract certain kinds of lung nodules from thoracic CT images，whose robustness，accuracy and efficiency can satisfy clinic requirements，and is helpful for lung cancer screening，diagnosis and treatment evaluation.