Interactive Visualization of Ultrasound Image Volumes for Oncologic Diagnostics

摘要

In today's computer assisted diagnostics, high-value 3-D imaging modalities are commonly used, since they permit differentiation between normal anatomical and pathological structures and also show their spatial relationships. While 2-D imaging leaves spatial relationships unclear, 3-D volume visualization can give clear insights into spatial relationships, but needs sophisticated visualization techniques and experienced interactive and tedious setting of parameters. Iso-surface rendering, direct volume rendering, and surface reconstruction are established techniques and are applied to CT and MR I image volumes for identification of vessels - used as anatomical landmarks - and tumors. Their visualizations are evaluated and compared to those of semi-automatically segmented 3-D ultrasound image volumes of the same anatomical structures. Even in CT and MRI, which have higher contrast than noisy ultrasound image volumes, vessels and tumorous lymph nodes can only be delineated with time consuming interaction. In comparison, semi-automatically segmented 3-D ultrasound image volumes present similar or even superior visualizations, but necessitate specifically adapted delineation algorithms. Their development is only worthwile, if inter-patient reproducibility and - as far as possible - artifact-free imaging conditions can be assured.