Efficient Exploration of 3D Medical Images Using Projected Isocontours
Marius Gavrilescu · 2019
The exploration and visualization of multidimensional medical images is an intricate and complicated problem, considering the amount and complexity of the structures and features within such data. A lot of such structures are identifiable as specific isosurfaces. Consequently, we propose a method for the intuitive selection and representation of meaningful isosurfaces from CT medical volume data. We first employ an isosurface detection algorithm which is easy to incorporate within a raycasting-based traversal of the volume. We then render 2D projections of multiple detected isosurfaces and subject them to various image processing and information visualization techniques meant to assist with the selection of meaningful structures and features from the whole body of data. Subsequently, we use a shape similarity metric to build a map which allows for the efficient selection of isosurfaces based on the relative differences of their projected contours. We then cluster the isosurfaces considering these relative differences, allowing for the fast selection of isosurface clusters which form the geometry of key anatomical structures. Our approaches are interactive and allow a potential user to quickly and effectively traverse the isosurface space, focusing on the selection and representation of meaningful portions of the volume data.