GPU‐based interactive visualization framework for ultrasound datasets

Abstract

Ultrasound imaging is widely used in medical areas. By transmitting ultrasound signals into the human body, their echoed signals can be rendered to represent the shape of internal organs. Although its image quality is inferior to that of CT or MR, ultrasound is widely used for its speed and reasonable cost. Volume rendering techniques provide methods for rendering the 3D volume dataset intuitively. We present a visualization framework for ultrasound datasets that uses programmable graphics hardware. For this, we convert ultrasound coordinates into Cartesian form. In ultrasound datasets, however, since physical storage and representation space is different, we apply different sampling intervals adaptively for each ray. In addition, we exploit multiple filtered datasets in order to reduce noise. By our method, we can determine the adequate filter size without considering the filter size. As a result, our approach enables interactive volume rendering for ultrasound datasets, using a consumerlevel PC.

Feature of Interest‐Based Direct Volume Rendering Using Contextual Saliency‐Driven Ray Profile Analysis

“…Our method does not require any modality‐specific parameter settings and therefore can be applied to other modalities, e.g. ultrasound [LKS09]. Its applicability to time‐varying data is another interesting research topic.…”

A fast 3D adaptive bilateral filter for ultrasound volume visualization

“…Ultrasound data filtering can be performed during pre-processing or post-processing. In many cases, 2D filters are used during the pre-processing stage [1,2], but 2D filtering techniques cannot produce good outcomes because these techniques consider only the relationships between sample points in a single 2D slice and ignore the relationships between adjacent slices. Although the outcomes can be improved using 3D filters, 3D filtering techniques have been applied using only simple filters because of the size of the ultrasound volume data and the amount of computations required [3,4].…”

“…(1) A nonlinear bilateral filtering technique that could not be applied previously due to calculation costs was used to quickly filter the ultrasound volume data. (2) Adaptive kernels that consider the sampling characteristics of the ultrasound volumes were applied such that filtering could be performed more equally than with existing methods. Although our method proposes to extend 2D filtering to 3D ultrasound coordinates, it also can be used for filtering of 4D ultrasound data because it is able to handle more than 10 volume data sets per second.…”

Optimized Graphical Processing Unit Processing Framework for Surface Rendering in 4D Ultrasound Imaging

“…In spite of the success of such methods for their respective applications, such approaches were not intended for real-time processing in addition to not being optimized for implementation on the presently available GPU technology. 11 Several groups proposed GPU-based framework for the ultrasound data volume render [11][12][13][14] with main focus on processing time. An approach that would combine the optimization methods used in offline 3D visualization and real-time processing on GPUs would offer an excellent platform for 4D imaging applications.…”