Segmentation of lung lesion Nodules using DICOM with structuring elements and noise-a comparative study

Lung cancer is one of the major considerations that the field of science and medicine has to overcome. Various medical imaging modalities like X-ray, CT, chest radiography, SPECT, NM, MRI, CT, US, PET and optical modalities like Endoscopy, Microscopy or Photography exists to identify the presence of disease. Automated Computer Aided Diagnosing (CAD) system is more useful tool for advanced decision making in radiology. CAD system performs diagnosis and detection of malignancy from suspect regions in medical image. Aggregation of cells (Nodules) is unusual appearances that are numerous, clustered, irregularly shaped and sized and branching in orientation. Detection sensitivity of cancer depends on identification of malignant nodules. Major challenge lies in the lesion ROI segmentation during the clinical evaluations. For clinical work flow medical images are stored in PACS. In this paper the nodules are extracted from the DICOM lung image in the noise environment such as Gaussian, salt and pepper, Poisson and speckle using different edge detection operators such as Gaussian, Average, Laplacian and Sobel. To increase the reliability, contour detection is followed by morphological analysis with various sizes of the Disk and Diamond structuring element and watershed algorithm. These results enable to analyze the accuracy of nodule extraction from DICOM images and the impact of the noise during Diagnosis.