Performance evaluation of edge detectors - morphology based ROI segmentation and nodule detection from DICOM lung images in the noisy environment

Several lung diseases are diagnosed detecting patterns of lung tissue in various medical imaging obtained from MRI, CT, US and DICOM. In recent years many image processing procedures are widely used on medical images to detect lung patterns at an early and treatment stages. Several approaches to lung segmentation combine geometric and intensity models to enhance local anatomical structure. When the lung images are added with noise, two difficulties are primarily associated with the detection of nodules; the detection of nodules that are adjacent to vessels or the chest wall corrupted and having very similar intensity; and the detection of nodules that are non-spherical in shape due to noise. In such cases, intensity thresholding or model based methods might fail to identify those nodules. Edges characterize boundaries and are hence of fundamental importance in image processing. Image edge detection significantly reduces the amount of data by filtering and preserving the important structural attributes. So understanding of edge detecting algorithms is necessary. In this paper Morphology based Region of interest segmentation combined with watershed transform of DICOM lung image is performed and comparative analysis in noisy environment such as Gaussian, Salt & Pepper, Poisson and speckle is performed. The ROI lung area blood vessels and nodules from the major lung portion are extracted using different edge detection filters such as Average, Gaussian, Laplacian, Sobel, Prewitt, Unsharp and LoG in presence of noise. The results are helpful to study and analyse the influence of noise on the DICOM images while extracting region of interest and to know how effectively the operators are able to detect, overcoming the impact of different noise. The evaluation process is based on parameters from which decision for the choice can be made.