Abstract:Abstract Background . Computed tomography perfusion (CTP) imaging plays a critical role in the acute stroke syndrome assessment due to its widespread availability, speed of image acquisition, and relatively low cost. However, due to its repeated scanning protocol, CTP imaging involves a substantial radiation dose, which might increase potential cancer risks. Methods . In this work, we present a novel deep learning model called non-local perfusion texture learning network (NPTN) for high-quality CTP imaging at low-dose cases. Specifically, considering abundant similarities in the CTP images, i.e. latent self-similarities within the non-local region in the CTP images, we firstly search the most similar pixels from the adjacent frames within a fixed search window to obtain the non-local similarities and to construct non-local textures vector. Then, both the low-dose frame and these non-local textures from adjacent frames are fed into a convolution neural network to predict high-quality CTP images, which can help better characterize the structure details and contrast variants in the targeted CTP image rather than simply utilizing the targeted frame itself. The residual learning strategy and batch normalization are utilized to boost the performance of the convolution neural network. In the experiment, the CTP images of 31 patients with suspected stroke disease are collected to demonstrate the performance of the presented NPTN method. Results . The results show the presented NPTN method obtains superior performance compared with the competing methods. From numerical value, at all dose levels, the presented NPTN method has achieved around 3.0 dB improvement of average PSNR, an increase of around 1.4% of average SSIM, and a decrease of around 4.8% of average RMSE in the low-dose CTP reconstruction task, and also has achieved an increase of around 3.4% of average SSIM and a decrease of around 61.1% of average RMSE in the cerebral blood flow (CBF) estimation task. Conclusions . The presented NPTN method can obtain high-quality CTP images and estimate high-accuracy CBF map by characterizing more structure details and contrast variants in the CTP image and outperform the competing methods at low-dose cases.

Temporally Downsampled Cerebral CT Perfusion Image Restoration Using Deep Residual Learning.

Deep Spatial-Temporal Convolutional Neural Networks for Medical Image Restoration

STIR-Net: Deep Spatial-Temporal Image Restoration Net for Radiation Reduction in CT Perfusion

Contrast-Medium Anisotropy-Aware Tensor Total Variation Model for Robust Cerebral Perfusion CT Reconstruction with Low-Dose Scans

Learning non-local perfusion textures for high-quality computed tomography perfusion imaging

Self-Supervised Dynamic CT Perfusion Image Denoising With Deep Neural Networks

An Efficient Iterative Cerebral Perfusion CT Reconstruction Via Low-Rank Tensor Decomposition with Spatial-Temporal Total Variation Regularization.

Clustering Based Low Dose Cerebral Computed Tomography Perfusion Spatio-temporal Restoration

Deep learning-based correction for time truncation in cerebral computed tomography perfusion

CPCT-LRTDTV: Cerebral Perfusion CT Image Restoration Via a Low Rank Tensor Decomposition with Total Variation Regularization

Reduction of scan duration and radiation dose in cerebral CT perfusion imaging of acute stroke using a recurrent neural network

Value of 4D CT Angiography Combined with Whole Brain CT Perfusion Imaging Feature Analysis under Deep Learning in Imaging Examination of Acute Ischemic Stroke

CT Perfusion Map Synthesis from CTP Dynamic Images Using a Learned LSTM Generative Adversarial Network for Acute Ischemic Stroke Assessment

SDCNet: Smoothed Dense-Convolution Network for Restoring Low-Dose Cerebral CT Perfusion

Robust Restoration of Low-Dose Cerebral Perfusion CT Images Using NCS-Unet

U-net Models Based on Computed Tomography Perfusion Predict Tissue Outcome in Patients with Different Reperfusion Patterns

Low-dose Cerebral Perfusion Computed Tomography Image Restoration Via Low-Rank and Total Variation Regularizations

Neural network-derived perfusion maps: A model-free approach to computed tomography perfusion in patients with acute ischemic stroke

Spatio-temporal Constrained Adaptive Sinogram Restoration for Low-dose Dynamic Cerebral Perfusion CT Imaging

Reconstructing Quantitative Cerebral Perfusion Images Directly From Measured Sinogram Data Acquired Using C-arm Cone-Beam CT

Deep learning-based identification of acute ischemic core and deficit from non-contrast CT and CTA