Evaluation of the effects of head movement-related artifacts on the quality of cone-beam computed tomography images

INTRODUCTION

Cone-beam computed tomography (CBCT) is a valuable 3D imaging tool, extensively utilized in dentistry for multiple applications [1]. It plays a crucial role in various tasks, such as implant planning, assessment of impacted teeth, diagnosis of cysts and tumors, evaluation of ma­xillofacial trauma, periodontal treatments, orthodontic planning, and endodontic procedures [2]. CBCT imaging offers numerous benefits, including reduced radiation exposure, 3D reconstruction at diverse cross-sections, and enhanced image processing, when compared with traditional CT scans. With sub-millimeter voxel resolution in orthogonal planes, CBCT units can achieve high-resolution imaging. Additionally, the inclusion of cursor-driven measurement algorithms enables real-time dimensional assessment, annotation, and measurements, providing practitioners with interactive capabilities [3, 4]. However, this modality has some deficiencies, such as restricted imaging field and artifacts caused by metal objects as well as patient motion during scanning [3]. In the case of patient movement during CBCT scanning, it introduces geometric errors into the reconstruction process, ultimately compromising spatial resolution. If the object being imaged shifts during scanning, the geometry of images used for three-dimensional reconstruction is altered. Therefore, artifacts caused by patient movement can degrade image quality and diminish diagnostic accuracy of CBCT scans [4].
Artifacts are structures formed in the reconstruction of CBCT images, which do not actually exist [5]. The main sources of artifacts are: 1) electrical and photon counting noise; 2) photons from scattered X-rays; 3) extinction and beam hardening effects, e.g., due to metal implants; 4) approximations in reconstruction due to finite beam width and detector pixel size; 5) aliasing due to finite pixel size and cone-beam divergence; 6) ring artifacts due to defect or mis-calibrated detector elements; and 7) patient movement. Motion artifacts arise since the reconstruction assumes that the scanned patient is stationary. However, periodic respiratory or cardiac motions, such as breathing and heart beat in the chest and lung region as well as non-periodic, including abrupt patient motion, gas bubbles in the abdomen and digestive system, all lead to acquiring projections from different states of motion, resulting in evident and undesirable, typically streak-shaped, image artifacts after reconstruction [6]. Artifacts are the main cause of image quality degradation, and even in some cases, they make images unreadable and non-interpretable [7]. The occurrence of patient movement during CBCT imaging can be attributed to a CBCT reconstruction algo­rithm. When object movement takes place, the algo­rithm back-projects pixels depicting the same region of interest into various positions, resulting in generation of hazy images characterized by blurred lines, double-contours, and diminished sharpness [10]. Another factor influencing the occurrence of artifacts is the acquisition time, which may vary from 5 to 40+ seconds depending on the device [8]. The motion also strengthens metal artifacts [9]. Various algorithms and methods for reducing motion artifacts have been studied to be effective [10]. Currently, the technical community is actively engaged in the advancement of artifact reduction techniques. Certain equipment necessitates a head tracking system for effective motion artifact correction, which mostly involve post-processing algorithms applied to 3D volu­me data. While such approaches may significantly diminish certain visible artifact structures, it is essential to acknowledge, from a physical perspective, that the error has already been incorporated into the volume [6, 11].

OBJECTIVES

The aim of this study was to determine the effects of artifacts caused by different head movements on the quality of CBCT images.

MATERIAL AND METHODS

An experimental study was carried out with ethics code number: IR.UMSHA.REC.1399.716 issued by the Hamadan University of Medical Sciences. A dry human skull with teeth without any restoration or filling was used in the study. The skull was secured onto a robotic system [12] to mimic various movements observed in patients (Figure 1A).
Using the fabricated robot, patient’s movements were simulated at different axes at adjustable speed and angles. In order to achieve this, the robot was capable of simulating four head movements, implemented based on clinical evidences [13]. Four head movements were posterior (4 mm), anterior (3, 7, 10 mm), lateral rotational (5, 10, 15°), and tremor (5°) motions. Recent in vivo studies indicated that movements of ≥ 0.5 mm occur in approximately 80% of CBCT examinations, with patient movements ≥ 3 mm significantly affecting CBCT image quality and interpretability [11]. Additionally, a master pilot study was conducted to assess the impact of patient movement on CBCT image quality based on radiation geometry of CBCT devices under examination. The rotational movement was simulated by a stepper motor (Greensky Power Co., Hangzhou, Zhejiang, China) and anterior, posterior, and right-side transitio­nal movements were simulated by four DC motors with a gearbox (Greensky Power Co., Hangzhou, Zhejiang, China). To control movements and motors, an Arduino board (Fut-electronic Tech Co., Shenzhen, Guangdong, China) was employed, and to control the robot remotely, a Wi-Fi connection was used. Rotation angle was ± 15° (totally 30°), with 0.9° accuracy. In addition, movement range was adjustable up to 10 mm, and movement speed was 0.9 mm/20 ms. The chosen movement ranges were determined considering motions that can occur during an extended acquisition time of CBCT images, particularly in pediatric cases and patients with conditions characterized by involuntary movements, such as Parkinson’s disease [14]. Figure 1B shows different movements of the dry human skull. For CBCT imaging, two different CBCT systems with specific characteristics were applied (Table 1).
CBCT images with and without head movements were taken and stored in OnDemand3D software (Cyber­med, Seoul, Korea). Axial and cross-sectional images were evaluated at two levels, including maxillary and mandibula alveolar crest. Cross-section thickness of 1 mm with 1 mm intervals was used to examine the quality of images, and images with artifacts were categorized as streak-like groups, double-contour, ring-like, and blurred images. To quantify image quality and resolution, a four-point scale was applied, and images were scored from zero (lowest quality images) to three (highest quality images). Based on literature review, scores defined as zero indicated streak-like and double-contour artifacts, which prevented diagnostic analysis of the images; one were streak-like and contour artifacts, with feasible diagnostic analysis of the images; two were streak-like artifacts, with feasible diagnostic analysis of the images; and three indicated no artifact, with maxi­mum diagnostic and analysis quality. This four-point scale is commonly utilized in radiologic research, with confirmed high validity and reliability [14-16]. Additionally, a comprehensive description of each scale was provided in a questionnaire, enabling radiologists to assess images based on specific criteria associated with each scale, and select the most suitable rating.
To examine the measuring accuracy of CBCT images, 12 markers were used; markers with acrylic balls (6 mm in diameter) were placed 4 mm from buccal and lingual aspects of the alveolar crest on the posterior and anterior areas of the maxilla and mandible.
The distance between the two acrylic balls on the skull was experimentally measured using a caliper (Mitutoyo Corp., Kawasaki, Japan), and the distance on the inner surface of the two balls adjacent to the ridge was measured on the images in OnDemand3D software, and the results were compared with those of the skull (Figure 2).
Two oral and maxillofacial radiologists with more than 10 years of experience in CBCT image analysis, assessed the images. Some of the scans were given to two observers as training images, and they were blinded to type, level and pattern of movements. These two observers were able to adjust the light, contrast, and zoom level of the images. In addition, to examine movement artifacts and general quality of the images, they were allowed to assess all parts of the images at axial, reconstructed panoramic, and cross-sectional levels. Distance measurements were performed by a software on one of the cross-sections, where the acry­lic ball was completely focused and sharp (Figures 3-6). The measurements and examinations were conducted twice with two weeks interval, and the results were recorded on a checklist [17]. The collected data analysis was done with SPSS version 16.0 (IBM Corp., Armonk, NY, USA) using descriptive statistics (number, frequency, mean, and standard deviation) and Wilcoxon signed-rank test. Inter-class correlation coefficient (ICC) and kappa statistic were employed for evaluating inter- and intra-rater reliability, and agreement for quantitative and non-quantitative measurements, respectively.

RESULTS

Streak artifacts were observed in all rotational, ante­rior, posterior, and tremor movements with Cranex 3D and Carestream devices (six images). In other words, all images taken by the two CBCT devices contained artifacts with rotational, anterior, posterior, and tremor movements. There was no case of ring-like artifacts in the images obtained by the two devices. The highest frequency of double-contour was observed with posterior movement (3 mm and 10 mm) provided by Cranex 3D unit (n = 6), and with rotational movement (5°) by Carestream device (n = 5). In addition, the highest frequency of blurred image artifacts was noted with rotational (15°) and posterior (3, 7, and 10 mm) movements showed Carestream device (n = 6), and posterior (10 mm) and anterior (4 mm) movements by Cranex 3D (n = 6) (Table 2).
In general, the highest and the lowest number of artifacts were streak and double-contour artifacts, respectively. With 7 mm and 10 mm posterior and 5° tremor movements in Carestream device, the majority of images had a low quality. All images with rotational (5°) movement in Carestream device, and anterior (4 mm) movement and tremor (5°) motion in Cranex 3D device, showed a moderate quality. In addition, images with 3 mm posterior movement in Carestream and 7 mm posterior movement in Cranex 3D device were mostly of moderate quality. As to rotational movement (5°) in Cranex 3D unit, all images were of good quality. There was no image with excellent quality (Table 3).
Regarding quantitative measurement of inter-class correlation coefficient (ICC) assessed by the two obser­vers, it was equal to 0.825 (95% CI: 0.556-0.938%). In addition, intra-class correlation coefficient (ICC) of the observer No. 1 for two quantitative measurements was equal to 0.903 (95% CI: 0.715- 0.968%), while for the observer No. 2, it was equal to 0.873. For non-quantitative measurements, kappa for inter-observer agreement (0.761) and intra-observer reliability (0.812-0.831) showed substantial values. Statistical comparison results as to the linear distances between the acrylic markers at different maxilla and mandible anterior and posterior areas with rotational (5°, 10°, and 15°), posterior (3, 7, and 10 mm), anterior (4 mm), and tremor movements (5°), are listed in Table 4.
In Carestream device, the lowest effects were observed in measuring the distance between the markers with posterior movements (3 and 10 mm), and there was no significant difference between the skull and software measurements. Whereas in Cranex 3D unit, there was a significant difference between the skull and software measurements in most areas. In the distance measurement accuracy, rotational movement of 10 and 15° decreased the accuracy of measurements in both Carestream and Cranex 3D devices more than in other types of movements. The difference of distance measurement in the majority of jaw regions between the skull and CBCT images was significant. In the posterior mandible area, tremor movement created a significant difference in terms of accuracy of measurement, while the difference in measurements was not significant compared with other movements.

DISCUSSION

In this research, a dry human skull was secured onto a robot to replicate various patient movements. The skull was manipulated using a long rigid pole as a fulcrum, allowing for the generation of small movements by positioning the fulcrum near the skull phantom. This method facilitated the simulation of movements similar to those observed in vivo, as the skull phantom was placed directly on the same support utilized for real patients. However, certain complex and combined movements, which more closely resemble actual patient movements, could not be replicated in the skull phantom [18]. Artifacts in images reduce the visibility of details and affect diagnostic accuracy negatively in all medical imaging procedures. Re-imaging is required when image quality is insufficient for reporting. Re-scanning with CBCT causes patient to be exposed to an extra dose of radiation, which does not comply with as low as reasonably achievable (ALARA) principle [3].
The causes for the problem are established as discrepancies in the acquired images when the object of investigation (i.e., patient) moves during the examination, leading to mismatched voxel intensities used by the image-reconstruction algorithm [6], resulting in stripe-like artifacts, overall un-sharpness, and double-contours [19].
With different frequencies, all the simulated head movements of patients created artifacts in CBCT images, except for ring-like artifacts, which were not seen in the images. Streak artifacts were observed in all simulated movements in the two CBCT devices with identical frequencies. Streak artifacts were associated with the majority of movement patterns. There was no case of ring artifacts in the captured images using the two CBCT devices nor in rotational, posterior, anterior, and tremor movements. In addition, double-contour artifacts were observed in different tremor and posterior movements. The highest frequency of double-contour artifacts was observed with posterior (3 mm and 10 mm) movements by Cranex 3D (n = 6) and tremor movements (5°) by Carestream device (n = 5). Blurred image artifacts were mostly noted with posterior, anterior, and tremor movements with both CBCT machines. The highest frequency of blurred image artifacts was observed with rotational (15°) and posterior (3, 7, and 10 mm) movements in Carestream, and with posterior (10 mm) and anterior (4 mm) movements in Cranex 3D. In general, the most intense artifacts were streak artifacts, and the least intense were double-contour artifacts. Various studies have explored the impact of motion artifacts on images, and the influence of different patient positioning on motion artifacts. Nardi et al. [18] investigated the effect of various head movements on image quality, and observed that different motion types do not equally impact image quality, with short duration and gradual movements affecting images differently. In contrast, Keris et al. [3] conducted a retrospective study comparing motion effects on images with different patient positioning, and found no significant impact of motion positioning on image quality. In a study by Spin-Neto et al. [17], the frequency of stripe and lack of sharpness artifacts in images taken with heads’ movement was high. They reported that after using artifacts correcting system, stripe and lack of sharpness artifacts were limited to tremor movements, which is an indicative of the importance and role of tremor movements in artifacts creation. In 2016, Nardi et al. [18] investigated the effects of motion artifacts on CBCT images, and showed that all head movements that caused motion artifacts had no identical impact on the quality of images; the rotational movements had a worse effect on the diagnostic quality of CBCT images. This finding is relatively consistent with the present study. On the other hand, in 2013, Spin-Neto et al. [4] reported motion artifacts effects on the quality of CBCT images after simulating head movements using a skull robot. They also reported that stripe-like artifacts were more common in all types of movements in images produced by all CBCT devices. According to the results, head movements, regardless of the type, cause artifacts in CBCT images, and their effects on image quality depend on the position and extent of skull movement [4]. This is inconsistent with our findings. The current study differed from Spin-Neto et al.’s [4] in terms of the size of field of view as well as the type and magnitude of the movements examined.
In 2016, Nardi et al. [18] showed that not all movements caused motion artifacts, while rotational movements created artifacts in 100% of cases. In addition, the rate of artifacts in shaking and tilting movements was 92% and 67%, respectively. Various results in this regard depend on different movements of anatomic structures. Additionally, Nardi et al. [18] utilized a NewTom 3G machine in patients in a supine position, whereas the CBCT machines employed in our study required patients to be in a standing position, increasing the likelihood of patient movement in this scenario.
The position in which an artifact appears, depends on the angle of imaging and divergence between X-ray beam and detector center. This phenomenon is rooted in the projection geometry of 2D images, so that anatomic structures with more distance from the detector have a fewer contribution in reconstructing volume data of the structure under study. As a result, the obtained geometry is widely different from the ideal geometry (back-projection) [18].
Intensity of artifacts also depends on the movement time, position of X-ray source, and its base displacement. In the case of CBCT images, because of longer gantry rotation time, images are taken in shorter time intervals. In addition, patients rarely experience movements lasting for 2-6 sec (1-23%) [18].
The quality of CBCT images has a direct relationship with artifacts. As the obtained results demonstrate, the majority of CBCT images of the simulated head movements had a moderate quality, but good quality images were only obtained with Cranex 3D device with rotational movements (5, 10, and 15°). There was no case of excellent quality image.
Most of the significant differences between the actual and software measurements of the distance between acrylic markers on the skull were noted in the images taken by Cranex 3D device. This indicates that Carestream is relatively more accurate; for 3 mm and 10 mm posterior movements, there was no significant difference between the actual and software measurements of the distance between acrylic markers on the skull. However, as to Cranex 3D device, all movements created significant differences on various areas between the actual and software measurements of the distance between acrylic markers on the skull.
In general, the type of CBCT devices did not have a notable effect on the quality of images, while the type of artifacts had the main effect on the quality of images. Tremor, rotational, posterior, and anterior movements, all created images with streak artifacts at C-arm rotation range in the respective CBCT devices, which use different methods to prepare data from the field of views (FOVs). In addition, the exposure time was different in these two machines, which can be a reason for different frequency of various artifacts. In 2007, Katsumata et al. [20] examined effects of image artifacts on CBCT images, and reported that devices equipped with a flat panel detector had fewer artifacts when compared with a CBCT system with an image intensifier.
Rotation range of device and its software can also affect the frequency and intensity of artifacts [21]. In this study, two FOVs were used by the two CBCT devices with small differences, in which the size and number of FOVs were not the same. The selected FOVs need to cover the information necessary for diagnosing and treatment plans. However, since patients are not the same in terms of body size, FOV parameter changes depending on the patient’s size. On the other hand, the selected FOV size is the most important scan parameter to limit the rational dose and image quality [22]. According to studies by Parsa et al. [22] and Hassan et al. [23], FOV has a notable effect on the level of gray shadows in CBCTs. Bigger FOVs have a lower resolution and contrast compared with smaller FOVs, so that they directly affect the observable anatomic structures produced by CBCT devices [24].
Distance measurement accuracy in CBCT images was evaluated by assessing linear distance between acrylic markers and comparing the results with the readings by Cranex 3D on anterior mandible and posterior maxilla, showing mostly notable and significant differences. In the case of Carestream device, the anterior mandible and posterior maxilla had more significant differences in terms of image quality based on the actual and software measurements. The frequency of significant differences in Carestream device was less than that of Cranex 3D. On the other hand, there were fewer significant diffe­rences between the actual and software measurements in the posterior mandible areas for all rotational, anterior, posterior, and tremor movements. This can be explained based on a lower level of anatomic complicacies in this region. In addition, it is notable that the processing of projecting images on display decreases the resolution and quality of images. This can be solved by using high definition displayers.
Quality of CBCT images for diagnosing complicated cases, such as the connection between mandible canal and third molar teeth roots, or anatomic structures in bone injuries, including mental foramen, can be affected by artifacts [25-28]. It is also possible that artifacts have less impact on diagnosing diseases and probable pathologies in less complicated anatomic structures. Furthermore, the quality of cross-sectional images, in all cases, are affected by the type of patients’ head movements.
Many efforts have been made to prevent patient motion during CBCT acquisition. Patients are generally immobilized using a head strap, chin rest, and fixed bite block. However, relying on head fixation methods may not prevent all potential motions [27], since as small as 3 mm motion displacement can significantly affect the quality of image [17]. With short scanning time, the patient would have less chance to make a movement, which of course, is not feasible by the given limited resolution of flat panel detectors in CBCT devices [29]. Using movement tracking sensors and markers on patient’s head can be considered a way to limit patient’s movement and artifacts [30]. A survey used a gyroscope to measure patients’ head movements, but it still appears that intrinsic lack of accuracy in these systems creates errors in measuring angular and spatial positions within obtained images [30, 31].
Different levels of motion artifacts have been shown in previous studies. Yildizer Keris et al. [3] reported motion artifacts in 6.7% of patients. Spin-Neto et al. [29] revealed the rate of patient’s movement in artifact detection studies as 41%. One of the reasons for patients’ movements during CBCT imaging is anxiety, which leads to motion and physical symptoms, such as restlessness, vertigo, breath shortness, and shaking during scanning [32]. To control anxiety, it is important to provide adequate education to patients. Patient anxiety prior to CBCT examination can be alleviated through effective communication, informing patients about the CBCT procedure as well as by implementing anxiety-reduction techniques [14].
The findings in this study highlight the importance of paying more attention to patient’s movement during CBCT scanning, and, hence, reducing the artifacts. To control motion artifacts and reduce the necessity of repeating CBCT scans, further studies are needed on the relationship between patient’s position (supine, sitting, or standing) and their movements. Methods to restrain movement of the mouth, chin, and jaws, are also needed to be examined. Ensuring adequate fixation of the patient’s head during scanning can restrict movement options for the patient. Furthermore, advancements in detector hardware are expected to facilitate quicker detector read-out, leading to decreased scanning durations and minimizing the likelihood of patient movements [6, 33]. It is also possible to use special mathematical software to correct images when the patient moves during CBCT scanning.
To mitigate the occurrence of motion artifacts, future research can explore and assess different techniques for skull immobilization. Additionally, the development of software incorporating mathematical algorithms to detect and eliminate images captured during patient movements may enhance the quality of reconstructed data sets, potentially reducing the necessity for retakes, and minimizing additional radiation exposure to patients [18].

CONCLUSIONS

The simulated head movements created artifacts in CBCT images in most cases, and the frequency of artifacts was relatively similar in using two different CBCT devices (Carestream and Cranex 3D).
In general, the highest and the lowest number of artifacts were streak and double-contour, respectively. Ring-like artifact was not observed in the images. The majority of CBCT images with head movements showed a moderate level of quality. The posterior movement (10 mm) decreased the quality of images more than other types of movements. The rotational movement (5°) had the lowest negative effects on the quality of images. In terms of the accuracy of distance measurement, in both Carestream and Cranex 3D devices, the rotational movement decreased the accuracy more than other types of movements. The accuracy of distance measurement with Carestream device was relatively higher than Cranex 3D in the mandible anterior and maxilla posterior areas. The accuracy of distance measurement in the posterior mandible area was less affected by different movements compared with other areas of the jaws.

Disclosures

1. Institutional review board statement: This study was approved by the Ethics Committee of the Hamadan University of Medical Sciences (approval number: IR.UMSHA.REC.1399.716).
2. Assistance with the article: This work is a part of dentistry thesis, supported by Dental Research Center, Vice Chancellor of Research, Hamadan University of Medical Sciences (grant number: 9911288444).
3. Financial support and sponsorship: None.
4. Conflicts of interest: The authors declare no potential conflicts of interest concerning the research, authorship, and/or publication of this article.

References