Segmentation of Radiological Images

Abstract

Today’s typical hospital environment is well-equipped with medical scanners that routinely provide valuable information to aid with the diagnosis or treatment planning for a particular patient. Computerised tomography (CT), magnetic resonance imaging (MRI), ultrasound, and positron emission tomography (PET) are examples of imaging modalities that are frequently used. An experienced radiologist can gain much insight by viewing the individual images that a scanner produces. However, segmentation of the radiological images to extract or classify a specific region or volume of interest is often required to partition the image data into its constituent components. Image segmentation provides quantitative information about relevant anatomy, for example to determine the size or volume. It also enables an accurate three-dimensional (3D) visualisation of a particular structure using surface triangulation, isosurfacing or volume rendering. There is no single approach that can generally solve the problem of segmentation, and different methods will be more effective depending on the image modality being processed.