Stories, Papers, WIKIs

Title Body
A framework for volume segmentation and visualization using Augmented Reality (ACM)

Abstract:
We propose a two-handed direct manipulation system to achieve complex volume segmentation of CT/MRI data in Augmented Reality with a remote controller attached to a motion tracking cube. At the same time segmented data is displayed by direct volume rendering using a programmable GPU. Our system achieves visualization of real time modification of volume data with complex shading including transparency control by changing transfer functions, displaying any cross section, and rendering multi materials using a local illumination model. Our goal is to build a system that facilitates direct manipulation of volumetric CT/MRI data for segmentation in Augmented Reality. Volume segmentation is a challenging problem and segmented data has an important role for visualization and analysis.

Paper available at ACM.
An Accelerative Method for Multimodality Medical Image Registration Based on CUDA (IEEE)

Abstract:
Image registration algorithm based on normal mutual information (NMI) has been adopted in multimodality medical image registration research for a few years. Basing on the whole information of the image pixels, this method has no incertitude of the problems about the similar parts, different range of values, segmentation, boundary, and so on. The results have always been more satisfied by doctors than other registration methods. However, medical images usually contain a huge amount of data and the calculation of mutual information takes a lot of statistical analyses. Therefore the operation time is too long to apply in clinical. According to this problem, researchers have intended to improve the algorithm on two aspects, namely software and hardware. Compute Unified Device Architecture (CUDA) is a new technology of parallel computing with broad-ranging application on image processing, reconstruction, analysis, and much more. Above all, we proposed a new improvement CUDA based method to accelerate NMI registration algorithm. Experiments show it cannot only ensure to provide a correct registration images but also an efficiency improvement as well. And the total speedup ratio nearly reaches 10.

Paper available at IEEE.
Denosing 3D Ultrasound Images by Non-local Means Accelerated by GPU (IEEE)

Abstract:
In Ultrasound imaging, speckle noise is the most serious problem which affects the performance of images. Non-local mean filter is a nice method to remove the speckle noise, but the algorithm' computational complexity makes it's a highly time-consuming method. In many applications like image-guided surgical intervention, real-time de-noising is required. This paper implements a NLM method accelerated by GPU for real-time denoising of 3D ultrasound images. The experimental results show the proposed accelerated de-noising method is efficient in terms of denoising quality and real-time.

Paper available at IEEE.
Fast and Accurate 3D Compton Cone Projections on GPU using CUDA (IEEE)

Abstract

We present a fast and accurate method for reconstructing single photons detected by a Compton camera using 3D cone projection operations formulated to run on a graphics processing unit (GPU) and the compute unified device architecture (CUDA) framework. With these projection operations, image quality and accuracy of modalities such as positron emission tomography (PET) can be improved by incorporating Compton scatter events. We also use Monte Carlo simulation to produce a model of the blurring effects caused by limited energy and spatial resolution of the detectors to improve the quality and accuracy of the reconstructed images. The blur model is then incorporated into the cone projections in a cone-by-cone basis. Our method overcomes challenges such as compute thread divergence, and exploits GPU capabilities such as shared memory and texture memory. Unique challenges for projecting cones compared with projecting lines are also addressed for the GPU. The projection operations are integrated into a list-mode ordered subsets expectation maximization (OSEM) framework to reconstruct images from a Compton camera. The algorithm with blurring model achieves an average of 17.3% improvement on CNR compared with the image reconstructed without the blurring model. The whole reconstruction algorithm takes 2.2 seconds per iteration to process 50,000 cones in a 96×96×32 image on a NVIDIA GeForce GTX 480 GPU, including forward projection, backprojection, and multiplicative update. On a single core state-of-the-art central processing unit (CPU), it takes 3.1 hours for the same task with the same level of accuracy in blur modeling. Images generated using the CPU and GPU implementing the same blurring model are virtually identical, with root mean squared (RMS) deviation of 0.01%.

Paper available at IEEE.
Implementing Geant4 on GPU for Medical Applications (IEEE)

Abstract

Monte Carlo simulation (MCS) plays a key role in medical applications, especially for emission tomography (ET) and radiotherapy (RT). Unfortunately MCS is also associated with long calculation times that prevent for using it in routine clinical practice. Actually, a solution based on the use of computer clusters to solve the intensive computational issues is not realistic within routine clinical environment. Recently graphics processing units (GPU) became in many domains a cheap solution for the acquisition of a high power computation. The objective of this work was to develop an efficient framework for the implementation of MCS on GPU architectures. Geant4 was used as the MCS engine for targeting medical imaging and radiotherapy applications. We propose the definition of a global strategy and associated structures for such a GPU based simulation. The different steps needed for a Geant4 simulation were implemented on GPU. The first validations have shown equivalence in the underlying photon physics processes between the Geant4 and the GPU codes. Based on these simplistic simulations, we are expecting a speedup factor of over 200 for a complete simulation in emission tomography or in radiotherapy dosimetry.

Paper available at IEEE.
Measurement-Based Spatially-Varying Point Spread Function for List-Mode PET Reconstruction on GPU (IEEE)

Abstract

We present a novel method to accurately model the spatially-varying point spread function (PSF) of a PET system reformulated for list-mode reconstruction on the graphics processing unit (GPU). The spatially-varying PSF for each LOR is modeled as an asymmetric Gaussian function whose variance changes asymmetrically according to the orientation of the line of response (LOR) and the voxel geometry. To fit the PSF parameters, a point source is imaged at twelve locations in a Philips Gemini TF PET system. To avoid tedious mechanical calibrations, the accurate point source location is estimated directly from the list-mode data. We introduce canonical sinogram to enable reading out the sampled PSF directly from a stack of sinograms by exploring the rotational symmetry of the system matrix. The critical parameters for the PSF model are obtained by solving a convex optimization problem based on the measured point source data. The spatially-varying PSF is efficiently incorporated into the image reconstruction process on the GPU using the CUDA texture memory. The reconstruction algorithm incorporating the measurement-based shift-varying PSF takes 103 milliseconds per iteration to process a million LORs in a 75×75×26 image on a GeForce GTX 480 GPU, which is 190 times faster than a non-PSF implementation on a state-of-the-art central processing unit (CPU), and only 6.8% slower than a spatially-invariant fixed-width Gaussian kernel on the same GPU. Compared with no PSF modeling, this shift-varying PSF shows an average improvement of spatial resolution and contrast to noise ratio for point sources at the periphery of 2.95 ± 0.44% and 159.62 ± 31.54%, respectively. Improvements of the same parameters compared to the spatially-invariant PSF are 1.00 ± 0.26% and 41.11 ± 9.45%, respectively. These results indicate that the fast and accurate spatially-varying PSF reconstruction promises better resolution and contrast recovery with very s- all additional computational cost.

Paper available at IEEE.
GPU-Based Fast Projection-Backprojection Algorithm for 3-D PET Image Reconstruction (IEEE)

Abstract

Iterative image reconstruction algorithms based on a stochastic model of emission tomography have been widely studied, because they can provide better image quality than analytic reconstruction algorithms. However, their long reconstruction time has been a major bottle neck for further developments of high resolution PET scanners and their applications. In recent years, there have been several attempts to reduce the PET image reconstruction time by using a graphic processing unit (GPU). To obtain high computational performance on a massive parallel GPU, however, global memory coalescing and branching diversity are to be carefully considered, which are not considered in most existing GPU-based algorithms. To increase global memory coalescing, we propose the image-rotation-based (IR-based) projection and frame-rotation-based (FR-based) backprojection schemes. We then successfully incorporate the geometrical symmetry property into the proposed schemes to reduce the branching diversity. Thereby, we effectively reduce the total image reconstruction time from many hours to a few seconds. Experimental results show that the proposed algorithm reduces the computation time by a factor of about 539 compared with a CPU-based straightforward implementation.

Paper available at IEEE.
GPU Based Calculation of a SPECT Projection Operator for Content Adaptive Mesh Model (IEEE)

Abstract

In this paper we explore the use of a graphical processing unit (GPU) in a fast calculation of a projection operator using a ray casting algorithm for a content-adaptive mesh model (CAMM). Previously we had introduced 2D and 3D tomographic image reconstruction using a CAMM for emission tomography (EM). The proposed GPU based projection operator calculation is fast and allows incorporation of a non-uniform attenuation and distance-dependent spatial resolution of the imaging system. The GPU allows for fast computation, therefore establishing a necessary step for the future practical development of a CAMM tomographic reconstruction.

Paper available at IEEE.
Performance Evaluation of Scatter Modeling of the GPU-Based “Tera-Tomo” 3D PET Reconstruction (IEEE)

Abstract

In positron emission tomography (PET), photon scattering inside the body causes significant blurring and quantification error in the reconstructed images. To solve this problem we have developed Monte Carlo (MC) based 3D PET reconstruction algorithms implemented on the Graphics Processing Unit (GPU). Our implementation takes multiple Compton scattering into account without any significant additional cost. The performance of the scatter correction is evaluated using GATE simulation as well as by comparing reconstruction results of Tera-Tomo to the reference reconstruction implementation of the Philips Gemini TOF PET which applies attenuation correction and single scatter simulation (SSS) for scatter correction. The comparative reconstruction results are based on the NEMA NU2-2007 image quality phantom.

Paper available at IEEE.
Automatic Monte-Carlo Based Scatter Correction For X-Ray Cone-Beam CT using General Purpose Graphic Processing Units (GP-GPU): A Feasibility Study (IEEE)

Abstract

Scattered photons highly degrade the quality of X-ray images and their effect has become more important due to the increasing interest in cone-beam geometry for the acquisition of CT (CBCT) and micro-CT data. The random nature of scatter events and the great influence of the sample suggest that the most accurate methods for their estimation are Monte Carlo (MC) techniques, but their use is usually hampered by the large computation time required to obtain an acceptable estimation of the scattered radiation.

Paper available at IEEE.