A detection method for streak artifacts and radiological noise in a non-uniform region in a CT image

Abstract

By using the CT images obtained by subtracting two CT images acquired under the same conditions and slice locations, we have devised a method for detecting streak artifacts in non-uniform regions and only radiological noise components in CT images. A chest phantom was scanned using 16- and 64-multidetector row helical CT scanners with various mAs values at 120 kVp. The upper lung slice image was employed as a target image for evaluating the streak artifacts and radiological noise. One hundred parallel line segments with a length of 80 pixels were placed on the subtracted CT image, and the largest CT value in each CT value profile was employed as a feature variable of the streak artifacts; these feature variables were analyzed with the extreme value theory (Gumbel distribution). To detect only the radiological noise, all CT values contained in the 100 line profile were plotted on normal probability paper and the standard deviation was estimated from the inclination of its fitted line for the CT value plots. The two detection methods devised in this study were able to evaluate the streak artifacts and radiological noise in the CT images with high accuracy.

Introduction

The introduction of helical scanning algorithms and developments in multidetector row CT (MDCT) technology have provided reductions in scan time, and the entire chest and abdomen can be scanned in a single-breath hold. Consequently, the clinical applications of computed tomography (CT) have been expanded [1], [2]. Now, by superseding conventional radiography systems, CT plays a central role as a clinical screening tool (e.g., in lung cancer screening). In addition, in the development of computer-aided diagnosis (CAD) systems, CT images were frequently employed as the target images [3], [4], [5], [6], [7]. The CT image noise is well known to be one of the primary factors contributing to the degradation of CT images and to have detrimental effects on the detection of low contrast objects [8], [9]. Thus, the evaluation of CT image noise will be of clinical use and also yield profitable information for the improvement of CAD systems.

The image noise on CT images can be categorized into the following two major types: radiological noise and streak artifacts. Radiological noise is random fluctuations in CT values and its origins are due to quantum noise, structure noise, and electronic noise; in general, it has been assumed that quantum noise is the dominant source of radiological noise on CT images. On the other hand, it is widely known that streak artifacts on CT images (streak patterns in Fig. 1a) are caused by discrepancies between reconstructed Hounsfield units (HU) in CT images and their true attenuation coefficients. When an X-ray beam passes through highly attenuated areas such as shoulders, a noisy projection will be produced in the attenuation direction, and the reconstruction process will have the effect of greatly magnifying the noise, resulting in streak artifacts on the CT image; in brief, they are caused by X-ray photon starvation [10]. Therefore, the streak artifacts and radiological noise are very much related to the number of X-ray photons detected by X-ray detectors in the CT scanner [10], [11]. Presently, the standard deviation (SD) of CT values in a region of interest (ROI) placed in a homogeneous background without streak artifacts is used to evaluate the radiological noise on CT images. However, streak artifacts are always mixed with the radiological noise, so that this mixture makes it difficult to detect only the streak artifacts exactly.

In previous studies, we analyzed the streak artifacts on low-dose CT images using the extreme value theory [12], [13]. These results showed that the CT value variations caused by streak artifacts can be statistically modeled by a Gumbel distribution, which is one of the generalized extreme value distributions, and, based on these results, we devised a new method of detecting the streak artifacts on CT images, which we called the Gumbel evaluation method. However, in our devised Gumbel evaluation method, it is difficult to detect the streak artifacts in a ROI placed in a non-uniform image. Thus, the visual evaluation methods, such as a rating method, have been only available for the streak artifacts in a non-uniform image [14], [15], [16], [17], [18], [19]. The method of evaluating radiological noise when using the SD of CT values also has this same problem. Moreover, it is impossible to detect only radiological noise using this method, because the fluctuations in CT values are attributed not only to the radiological noise but also to the streak artifacts. This problem prevents us from understanding the inherent characteristics of radiological noises on CT images, especially when radiation doses are low and many streak artifacts are generated on CT images. From this viewpoint, it would be useful to devise a method for assessing only radiological noise in CT images.

In the present study, we have improved the Gumbel evaluation method to be able to detect the streak artifacts in non-uniform CT image regions, and have also proposed a method for detecting only the radiological noise. We call these methods for detecting the streak artifacts and radiological noise in CT images the improved Gumbel evaluation method and Gauss evaluation method, respectively.