Segmentation and 3D visualization of volumetric image for detection of tumor in cancerous brain

Development of an accurate three dimensional (3D) model of human skull showing the shape and relative position of a tumor inside the cancerous brain is crucial for taking pre-surgical or pre-therapy measures. This paper presents a new method for automatic segmentation of volumetric image which is a collection of series of spatially distributed slice images of a cancerous brain. The proposed method provides a 3D visualization in order to identify the position, shape, and size of a tumor inside the brain of a skull. Each of the images of the series is segmented using a level-based segmentation method using the histogram of image intensities. Such a segmentation of single image in the series results in spurious regions which may cause misleading 3D shapes of tumors. In order to obtain an accurate 3D model of the tumor inside the brain, both the intra- and inter-slice area-based thresholds are exploited for removing the spurious regions from the initially segmented images. The proposed segmentation scheme not only removes the false regions regarding the segmentation of tumor, but also effectively maintains the size and shape of the original tumor. Extensive experimentations on a number of volumetric images reveal that the proposed method can provide minimum volumetric noise in the extraction process of a tumor as compared to the recently introduced vector flow method.