The segmentation of lung lesions with an acceptable accuracy from computed tomography (CT) scans is more valuable for lung cancer research and can offer important information for clinical diagnosis and treatment by the doctors. It is a challengeable process to achieve a fully automatic lesion detection and segmentation with acceptable accuracy due to the various characteristics of lung lesions. Here, we propose an improved toboggan based growing automatic segmentation approach (ITBGA) .which are automatic initial seed point selection, The new approach does not require any human interaction or training dataset for lesion detection, yet it can provide a high lesion detection sensitivity (98.35%) .The results show a significant improvement of segmentation accuracy. The average time consumption for one lesion segmentation was under 9 seconds using proposed method. The improved toboggan based algorithm can achieve robust, efficient and accurate lung lesion segmentation in CT images automatically.