An Automatic Estimation of Arterial Input Function Based on Multi-Stream 3D CNN

Abstract

Arterial input function (AIF) is estimated from perfusion images as a basic curve for the following deconvolution process to calculate hemodynamic variables to evaluate vascular status of tissues. However, estimation of AIF is currently based on manual annotations with prior knowledge. We propose an automatic estimation of AIF in perfusion images based on a multi-stream 3D CNN, which combined spatial and temporal features together to estimate the AIF ROI. The model is trained by manual annotations. The proposed method was trained and tested with 100 cases of perfusion-weighted imaging. The result was evaluated by dice similarity coefficient, which reached 0.79. The trained model had a better performance than the traditional method. After segmentation of the AIF ROI, the AIF was calculated by the average of all voxels in the ROI. We compared the AIF result with the manual and traditional methods, and the parameters of further processing of AIF, such as time to the maximum of the tissue residue function (Tmax), relative cerebral blood flow, and mismatch volume, which are calculated in the Section Results. The result had a better performance, the average mismatch volume reached 93.32% of the manual method, while the other methods reached 85.04 and 83.04%. We have applied the method on the cloud platform, Estroke, and the local version of its software, NeuBrainCare, which can evaluate the volume of the ischemic penumbra, the volume of the infarct core, and the ratio of mismatch between perfusion and diffusion images to help make treatment decisions, when the mismatch ratio is abnormal.

Optimal Scaling Approaches for Perfusion MRI with Distorted Arterial Input Function (AIF) in Patients with Ischemic Stroke

“…The changed AIF amplitude used in the cSVD algorithm shifts the maximum of R(t) to a higher time points ‘t’ which accounts for higher Tmax values. The change in Tmax is consistent with a previous study where different AIFs with changed amplitudes and similar shape selected by different algorithms resulted in change of Tmax values [ 1 ]. The increased Tmax maps generated by the rescaled AIF may allow clinicals to visualize the critically hypoperfused regions which are likely to be salvageable.…”

“…Selecting more voxels or a large region for AIF estimation can lead to significant PVE. However, PVE-corrected, multi-voxel AIF is necessary as AIF obtained from a single voxel or a small region is not reliable enough due to noise in spatial measurements and motion in temporal measurements [ 1 ]. In this study, we intend to make use of scaling as a way out to calibrate a multiple voxel AIF which would further lead to reduce the effect of the PVE on the quantification of absolute CBF and Tmax values.…”

“…Dynamic susceptibility contrast (DSC) MR perfusion imaging refers to perfusion scanning after an intravenous injection of a paramagnetic contrast agent containing gadolinium [ 1 , 2 , 3 , 4 ]. By utilizing the signal loss during the passage of a contrast agent through the brain tissues, it is possible to estimate perfusion parameters that are essential to identify ischemic core and penumbra, such as the cerebral blood flow (CBF), cerebral blood volume (CBV), mean transit time (MTT) and perfusion–diffusion mismatch [ 2 , 3 , 5 , 6 ].…”

“…In perfusion quantification, one of the most important functions, which is required as input from the perfusion data, is the arterial input function (AIF) [ 10 ]. AIF is the function that describes contrast agent concentration over time as it enters the brain through the brain feeding artery [ 1 , 2 ]. When the voxel or region that has been selected for quantification of AIF has only some fraction of the arterial blood, the partial volume effect (PVE) arises.…”

Computerized Tomography Image Feature under Convolutional Neural Network Algorithm Evaluated for Therapeutic Effect of Clarithromycin Combined with Salmeterol/Fluticasone on Chronic Obstructive Pulmonary Disease

“…The activation function used was the Sigmoid function [ 11 ], and the pooling operation was an average pooling operation. Rand random function [ 12 ] was used to randomize the feature vector, weight, and bias in the last layer of the fully connected layer.…”