Computer Vision News - February 2021

One-click Segmentation 10 The challenges of medical image segmentation Image segmentation is a fundamental and essential process in image analysis and computer vision, which involves partitioning an image into multiple meaningful segments for future analysis and use. In the medical field, image analysis plays an exceptionally vital role in both the diagnosis process , as well as in interventional procedures . Medical image segmentation poses many challenges in several different areas: • Modality - a wide variety of imaging modalities are in use today, depending on availability as well as the task at hand. Such modalities include microscopy, X-ray, ultrasound (US), computed tomography (CT), magnetic resonance imaging (MRI), positron emission tomography (PET) and many more. For some tasks, several imaging modalities are used simultaneously, for example preoperative MRI with live US- guidance during prostate biopsies. One-click segmentation of multimodal medical images • ROI - in medical imaging, the region of interest (ROI) varies widely depending on the task at hand, be it identifying the location and extent of an arterial blockage, or accurately estimating the size and boundaries of a tumor. • Dimensionality - the resulting medical image may be a two- dimensional image of a slice made up of pixels, or a three-dimensional image of a volume made up of voxels. At times, 3D images of volumes are even reconstructed from a set of 2D slices. Accordingly, the ROI can be represented as a slice or a volume.