Search

CT metal artefact reduction (MAR) methods based on supervised deep learning are often troubled by domain gap between simulated training dataset and real-application dataset, i.e., methods trained on simulation cannot generalize well to practical data. Unsupervised MAR methods can be trained directly on practical data, but they learn MAR with indirect metrics and often perform unsatisfactorily. To tackle the domain gap problem, we propose a novel MAR method called UDAMAR based on unsupervised domain adaptation (UDA). Specifically, we introduce a UDA regularization loss into a typical image-domain supervised MAR method, which mitigates the domain discrepancy between simulated and practical artefacts by feature-space alignment. Our adversarial-based UDA focuses on a low-level feature space where the domain difference of metal artefacts mainly lies. UDAMAR can simultaneously learn MAR from simulated data with known labels and extract critical information from unlabeled practical data. Experiments on both clinical dental and torso datasets show the superiority of UDAMAR by outperforming its supervised backbone and two state-of-the-art unsupervised methods. We carefully analyze UDAMAR by both experiments on simulated metal artefacts and various ablation studies. On simulation, its close performance to the supervised methods and advantages over the unsupervised methods justify its efficacy. Ablation studies on the influence from the weight of UDA regularization loss, UDA feature layers, and the amount of practical data used for training further demonstrate the robustness of UDAMAR. UDAMAR provides a simple and clean design and is easy to implement. These advantages make it a very feasible solution for practical CT MAR.

In this article, a fault diagnosis method based on data augmentation technology is proposed to investigate the unequal length series by combining dynamic time warping (DTW) with deep residual network (Resnet) for excavator, where the hydraulic and power units are operated in a hybrid manner. A data augmentation method is designed to segment time series with unequal length by enhancing appropriate statistical features based on the working characteristics of excavators. Based on instantaneous calculation, an optimized DTW algorithm is proposed to reduce the calculation cost and ensure the effectiveness of the searched optimal warping path. In order to achieve fault diagnosis in industrial applications, a DTW-Resnet model combining optimized DTW algorithm with Resnet model is proposed, which overcomes the disadvantage of neural networks being unable to learn time series with unequal length easily. Results with respect to SANY excavator datasets and four publicly available datasets have well indicated that the proposed method is provided with preferable diagnostic performance compared with state-of-the-art neural networks and traditional models. Furthermore, the ablative experiment shows that the data augmentation and optimized DTW are of great significance to improve the classification accuracy of the model.

The image quality entitlement is evaluated for multi-energy bin photon counting (PC) spectral CT relative to that of energy integration an dual kVp (dkVp) imaging. Physics simulations of X-ray projection channel data and CT images are used to map contrast-to-noise metrics for simple numerical phantoms objects with soft tissue, calcium and iodine materials. The benefits are quantified under ideal detector conditions. Spectral optimization yields on the order of 2X benefit for iodine visualization measured by CNR^2/dose in two different imaging modes: optimal energy weighting, and optimal mono-energy imaging. In another case studied, strict dose equivalence is maintained by use of a composite spectrum for PC simulation that combines simultaneously the two kVp excitations used sequentially for dkVp. In this case, mono-energetic imaging of iodine contrast agent is shown to achieve 40% higher dose efficiency for photon counting compared to dual kVp although non-ideal characteristics of the photon counting response can eliminate much of this benefit.

A primary modulator made of erbium is evaluated in X-ray scatter correction using primary modulation. Our early studies have shown that erbium is the optimal modulator material for an X-ray cone-beam computed tomography (CBCT) system operated at 120 kVp, exhibiting minimum beam hardening which otherwise weakens the modulator's ability to separate scatter from primary. In this work, the accuracy of scatter correction is compared for two copper modulators (105 and 210 m of thickness) and one erbium modulator (25.4 m of thickness) with the same modulation frequencies. The variations in the effective transmission factors of these three modulators as functions of object filtrations are first measured to show the magnitudes of beam hardening caused by the modulators themselves. Their scatter correction performances are then tested using a Catphan�600 phantom on our tabletop CBCT system. With and without 300 m of copper in the beam, the measured variations for these three modulators are 4.3%, 7.8%, and 0.9%, respectively. Using the 105- and 210-m copper modulators, our scatter correction method reduces the average CT number error from 327.3 Hounsfield units (HU) to 19.4 and 20.9 HU in the selected regions of interest, and enhances the contrast-to-noise ratio (CNR) from 10.7 to 16.5 and 15.9, respectively. With the 25.4-m erbium modulator, the CT number error is markedly reduced to 2.8 HU and the CNR is further increased to 17.4.