The main principle for the system-based study of lung cancers in CT images is cancer cell recognition and segmentation. Anyhow, in low-contrast pictures, it is a complex job as the low-level images are too small to detect. We are proposing a new technique in this project for the automated detection of lung cancers. Alternatively, by probability density function estimation, we enhance the intensity contrast of CT images. We use the expectation maximization / maximization of the posterior marginal to find cancerous areas. Finally, to decrease noise and classify focal cancers, we use shape limitation. The resolution of more than 95 percent of this fuzzy-based segmentation method is achieved and 9 percent accuracy is also given.