Evaluating surface visualization methods in semi-transparent volume rendering in virtual reality

Abstract

ABSTRACT Perceptual visualisation of semi-transparent structures in volumetric datasets is challenging due to its inherent visual complexity. This is however of primary importance in medical visualisation where raymarching of volumetric data is common. While rendering volumetric data itself is a well-explored area, perception of volume-rendered images combined with semi-transparent mesh data remains an underexplored topic. As virtual reality (VR) is increasingly employed for medical data inspection, it becomes important to understand how different mesh visualisations affect performance in medical tasks in this context. When considering direct volume rendering in immersive VR, the effects of stereoscopic vision and motion parallax require additional consideration. We investigate how surface transparency modes affect task performance in VR. For two medical image analysis tasks, we conduct a user study (n = 23) to analyse the impact of mesh rendering methods on task outcome and subjective preference. Our evaluation indicates that user performance with wireframe rendering varies greatly between tasks while constant opacity and silhouette provide a stable benefit. Overall, transparent visualisations led to improved precision (lower error rate, lower inter-observer variability) and increased user confidence, while the efficiency of visualisation methods was task-dependent. This demonstrates how semi-transparent rendering will improve visual analysis in medical applications and other disciplines.