You are here

Brain Perfusion Imaging: Performance and Noise Reduction

Fan Zhu

2nd September

Brain perfusion weighted images acquired using dynamic contrast studies may potentially have an important clinical role in acute stroke diagnosis and decisions to treatment. However, these images are quite noisy especially the computed tomography (CT) perfusion images and the analysis of perfusion imaging is time consuming. Methods: The majority of current approaches for denoising CT images are optimized for 3D (spatial) information, examples include spatial decimation (spatially weighted mean filters) and techniques based on wavelet and curvelet transforms. However, perfusion imaging is 4D as it also contains temporal information. Our approach using Gaussian process regression (GPR) take advantage of the temporal information to reduce the noise. Furthermore, local AIFs are one of techniques used to improve the accuracy. However, it leads to fairly slow performance. To solve this problem, a GPGPU based parallel implementation is developed to satisfy the demand to speed up the perfusion analysis. Results and Conclusions: Our GPR implementation gains 99% contrast to noise ratio (CNR) improvement and spatial decimation based GPR gains up to 102% CNR improvement. Our parallel implementation gains speedups of 2.6 and 4.8 depending on data size.