Hence, researchers have adopted objective and subjective image quality assessments to evaluate the feasibility of using neural networks in CT imaging. However, neural networks have not yet been widely used in practice owing to their confidence scores. Such networks have been used in applications such as spectral distortion correction for photon-counting X-ray CT, dual-domain learning for two-dimensional and three-dimensional low-dose CT reconstruction, sparse-view and limited-angle CT reconstruction, noise suppression in the sinogram domain and image domain, dual-energy imaging with energy-integrating detectors and photon-counting detectors, and CT artifact reduction. Several studies on the applications of such networks have been published, and their potential in solving several problems in the field of CT imaging has been extensively evaluated. In recent years, neural networks have been applied in computed tomography (CT) imaging. LeakyReLU outperforms Swish in terms of noise reduction. U-net is more effective in improving the image quality of smooth regions compared with that of the edge. The results indicate that the network cannot be applied to projections with high noise levels and possesses limitations in terms of efficiency for processing low-noise projections. In addition, we use covariance estimation to provide three measurements for the qualitative and quantitative performance evaluation of U-net. The results reveal that the covariance estimation obtained from the proposed analytical method agrees well with that obtained from the image samples for different phantoms, noise levels, and activation functions, demonstrating that propagating noise from only a single projection is feasible for CNN methods in CT reconstruction. A commonly used U-net structure is adopted for validation. Nesting of the linear and nonlinear layer noise propagation constitutes the covariance estimation of the CNN. We propagate noise through nonlinear layers in a CNN using the Taylor expansion. In this study, we propose an analytical method for noise propagation from a single projection to efficiently evaluate convolutional neural networks (CNNs) in the CT imaging field. Model observers that mimic subjective assessment require the mean and covariance of images, which are calculated from numerous image samples however, this has not yet been applied to the evaluation of neural networks.
In contrast, subjective assessments are trustworthy, although they are time- and energy-consuming for radiologists. Assessing the image quality of neural networks using traditional objective measurements is not appropriate because neural networks are nonstationary and nonlinear. Thus, quality assessments of the results obtained from a neural network are necessary to evaluate the neural network. Interestingly, among various experiments gelu seems to outperform swish in quite a lot of experiements.Neural network methods have recently emerged as a hot topic in computed tomography (CT) imaging owing to their powerful fitting ability however, their potential applications still need to be carefully studied because their results are often difficult to interpret and are ambiguous in generalizability. This work does include gelu in comparison experiments. mish(2019)Īccording to the paper mish can handle more deeper layered networks than swish, and in other aspects mish is normally slightly better than swish.īut overall, mish and swish performances are nearly identical. This is advantageous to relu since relu suffers from ‘dying RELU’ problems where significant amount of neuron in the network become zero and don’t practically do anything. It is differentiable in all ranges, and allows to have gradients(although small) in negative range. while relu is suddenly zero in negative input ranges, gelu is much smoother in this region. Like relu, gelu as no upper bound and bounded below. Despite introduced earlier than relu, in DL literature its popularity came after relu due to its characteristics that compensate for the drawbacks of relu.
0 kommentar(er)
