Quantifying the effect of X-ray scattering for data generation in real-time defect detection.

BACKGROUND: X-ray imaging is widely used for the non-destructive detection of defects in industrial products on a conveyor belt. In-line detection requires highly accurate, robust, and fast algorithms. Deep Convolutional Neural Networks (DCNNs) satisfy these requirements when a large amount of labeled data is available. To overcome the challenge of collecting these data, different methods of X-ray image generation are considered. OBJECTIVE: Depending on the desired degree of similarity to real data, different physical effects should either be simulated or can be ignored. X-ray scattering is known to be computationally expensive to simulate, and this effect can greatly affect the accuracy of a generated X-ray image. We aim to quantitatively evaluate the effect of scattering on defect detection. METHODS: Monte-Carlo simulation is used to generate X-ray scattering distribution. DCNNs are trained on the data with and without scattering and applied to the same test datasets. Probability of Detection (POD) curves are computed to compare their performance, characterized by the size of the smallest detectable defect. RESULTS: We apply the methodology to a model problem of defect detection in cylinders. When trained on data without scattering, DCNNs reliably detect defects larger than 1.3 mm, and using data with scattering improves performance by less than 5%. If the analysis is performed on the cases with large scattering-to-primary ratio (1 < SPR < 5), the difference in performance could reach 15% (approx. 0.4 mm). CONCLUSION: Excluding the scattering signal from the training data has the largest effect on the smallest detectable defects, and the difference decreases for larger defects. The scattering-to-primary ratio has a significant effect on detection performance and the required accuracy of data generation.

CT-based data generation for foreign object detection on a single X-ray projection.

Although X-ray imaging is used routinely in industry for high-throughput product quality control, its capability to detect internal defects has strong limitations. The main challenge stems from the superposition of multiple object features within a single X-ray view. Deep Convolutional neural networks can be trained by annotated datasets of X-ray images to detect foreign objects in real-time. However, this approach depends heavily on the availability of a large amount of data, strongly hampering the viability of industrial use with high variability between batches of products. We present a computationally efficient, CT-based approach for creating artificial single-view X-ray data based on just a few physically CT-scanned objects. By algorithmically modifying the CT-volume, a large variety of training examples is obtained. Our results show that applying the generative model to a single CT-scanned object results in image analysis accuracy that would otherwise be achieved with scans of tens of real-world samples. Our methodology leads to a strong reduction in training data needed, improved coverage of the combinations of base and foreign objects, and extensive generalizability to additional features. Once trained on just a single CT-scanned object, the resulting deep neural network can detect foreign objects in real-time with high accuracy.

Quantitative Comparison of Deep Learning-Based Image Reconstruction Methods for Low-Dose and Sparse-Angle CT Applications.

The reconstruction of computed tomography (CT) images is an active area of research. Following the rise of deep learning methods, many data-driven models have been proposed in recent years. In this work, we present the results of a data challenge that we organized, bringing together algorithm experts from different institutes to jointly work on quantitative evaluation of several data-driven methods on two large, public datasets during a ten day sprint. We focus on two applications of CT, namely, low-dose CT and sparse-angle CT. This enables us to fairly compare different methods using standardized settings. As a general result, we observe that the deep learning-based methods are able to improve the reconstruction quality metrics in both CT applications while the top performing methods show only minor differences in terms of peak signal-to-noise ratio (PSNR) and structural similarity (SSIM). We further discuss a number of other important criteria that should be taken into account when selecting a method, such as the availability of training data, the knowledge of the physical measurement model and the reconstruction speed.

Registration-based multi-orientation tomography.

We propose a combination of an experimental approach and a reconstruction technique that leads to reduction of artefacts in X-ray computer tomography of strongly attenuating objects. Through fully automatic data alignment, data generated in multiple experiments with varying object orientations are combined. Simulations and experiments show that the solutions computed using algebraic methods based on multiple acquisitions can achieve a dramatic improvement in the reconstruction quality, even when each acquisition generates a reduced number of projections. The approach does not require any advanced setup components making it ideal for laboratory-based X-ray tomography.