Towards robust deconvolution of low-dose perfusion CT: Sparse perfusion deconvolution using online dictionary learning

Abstract

Computed tomography perfusion (CTP) is an important functional imaging modality in the evaluation of cerebrovascular diseases, particularly in acute stroke and vasospasm. However, the post-processed parametric maps of blood flow tend to be noisy, especially in low-dose CTP, due to the noisy contrast enhancement profile and the oscillatory nature of the results generated by the current computational methods. In this paper, we propose a robust sparse perfusion deconvolution method (SPD) to estimate cerebral blood flow in CTP performed at low radiation dose. We first build a dictionary from high-dose perfusion maps using online dictionary learning and then perform deconvolution-based hemodynamic parameters estimation on the low-dose CTP data. Our method is validated on clinical data of patients with normal and pathological CBF maps. The results show that we achieve superior performance than existing methods, and potentially improve the differentiation between normal and ischemic tissue in the brain.

Computational Health Informatics in the Big Data Age

“…Sparse representation has been effective in medical image denoising and fusion using group-wise sparsity ] and image reconstruction [Xu et al 2012;Gholipour et al 2010]. Recently, coupled with dictionary learning, Fang et al [2013] restored the hemodynamic maps in the low-dose computed tomography perfusion by learning a compact dictionary from the high-dose data, with improved accuracy and clinical value using tissue-specific dictionaries ] and applying to various types of medical images [Fang et al 2014]. The sparsity property in the transformed domain has also been important in restoring the medical information by combining with the physiological models ].…”

Measuring sparse temporal-variation for accurate registration of dynamic contrast-enhanced breast MR images

“…This important property has been widely used in the communities of medical imaging, computer vision, multimedia and signal processing. It has been successfully applied to practical applications of shape modelling [25], image segmentation [26][27][28], image reconstruction [29,30], motion analysis [31], bias correction [32,33], image registration [34], image retrieval [35,36] and deconvolution [37,38] in the fields of medial imaging and medical image analysis. In addition, it has also been used in a large variety of applications in the field of computer vision, including face recognition [39], image restoration [40], image denoising, deblurring, superresolution and object recognition [41][42][43][44][45][46][47][48].…”

Classification of multiple sclerosis lesions using adaptive dictionary learning

“…brain network model [11] and predicting cognitive data from medical images [12]. In addition, the dictionary learning framework has been used in deformable segmentation [13], image fusion [14], super-resolution analysis [15], denoising [16,17], deconvolution of low-dose computed tomography perfusion [18,19] and low-dose blood-brain barrier permeability quantification [20]. In each of these applications, the dictionaries are learned from the underlying data so that they are better suited for representation of the signal of interest.…”