Abstract: It is hard to segment and extract the suspected nodular lesions from CT images for lung CAD system. To solve the key problem, an extraction algorithm is proposed for the suspected nodular lung lesions area based on the multi-scale morphology filtering which focuses on the feature of tissue geometry. A series of circle-like structure elements with different dimensions are built to do the morphologic filtering. After multi-scale processing, all nodules of different sizes are enhanced and stood out from the suspected lung lesion area. On the other hand, such linear tissues as trachea, bronchi tubes and blood-vessels are weakened and suppressed. And a particular advantage of the proposed method is found that the geometric parameters of morphology operator and binary threshold parameters needn't chose manually. Experiment results indicated that the method can extract suspected nodular lesions automatically and effectively.