Fast Contextual View Generation and Region of Interest Selection in 3D Medical Images Via Superellipsoid Manipulation, Blending and Constrained Region Growing
thesisposted on 2022-11-03, 16:57 authored by Ken Lagos
This thesis presents a 3D widget user-interface (UI), super-ellipsoid shape primitives and a customized volume rendering algorithm that together create a system effective for exploring 3D medical images and for selecting a 3D region within these images. Using a “painting” metaphor, the widget UI supports the fast and precise positioning of a super-ellipsoid shaped paint “blob”. The paint blob can be “deposited” and automatically blended with previously deposited blobs to form arbitrarily-shaped regions enclosing target image features. The rendering of these “focus” regions can be controlled separately from the surrounding contextual region, allowing medical experts to examine and measure image features relative to the context. The system’s core algorithms are designed to execute on Graphics Processing Units (GPUs), resulting in real-time interaction and high-quality visualizations. The focus plus context visualization system presented in this thesis is validated via a user study and a series of experiments.