The feasibility of head motion tracking in helical CT: a step toward motion correction

Estimate and compensate head motion in non-contrast head CT scans using partial angle reconstruction and deep learning

BACKGROUND

Patient head motion is a common source of image artifacts in computed tomography (CT) of the head, leading to degraded image quality and potentially incorrect diagnoses. The partial angle reconstruction (PAR) means dividing the CT projection into several consecutive angular segments and reconstructing each segment individually. Although motion estimation and compensation using PAR has been developed and investigated in cardiac CT scans, its potential for reducing motion artifacts in head CT scans remains unexplored.

PURPOSE

To develop a deep learning (DL) model capable of directly estimating head motion from PAR images of head CT scans and to integrate the estimated motion into an iterative reconstruction process to compensate for the motion.

METHODS

Head motion is considered as a rigid transformation described by six time-variant variables, including the three variables for translation and three variables for rotation. Each motion variable is modeled using a B-spline defined by five control points (CP) along time. We split the full projections from 360° into 25 consecutive PARs and subsequently input them into a convolutional neural network (CNN) that outputs the estimated CPs for each motion variable. The estimated CPs are used to calculate the object motion in each projection, which are incorporated into the forward and backprojection of an iterative reconstruction algorithm to reconstruct the motion-compensated image. The performance of our DL model is evaluated through both simulation and phantom studies.

RESULTS

The DL model achieved high accuracy in estimating head motion, as demonstrated in both the simulation study (mean absolute error (MAE) ranging from 0.28 to 0.45 mm or degree across different motion variables) and the phantom study (MAE ranging from 0.40 to 0.48 mm or degree). The resulting motion-corrected image,I D L , P A R ${I}_{DL,\ PAR}$ , exhibited a significant reduction in motion artifacts when compared to the traditional filtered back-projection reconstructions, which is evidenced both in the simulation study (image MAE drops from 178 ± $ \pm $ 33HU to 37 ± $ \pm $ 9HU, structural similarity index (SSIM) increases from 0.60 ± $ \pm $ 0.06 to 0.98 ± $ \pm $ 0.01) and the phantom study (image MAE drops from 117 ± $ \pm $ 17HU to 42 ± $ \pm $ 19HU, SSIM increases from 0.83 ± $ \pm $ 0.04 to 0.98 ± $ \pm $ 0.02).

CONCLUSIONS

We demonstrate that using PAR and our proposed deep learning model enables accurate estimation of patient head motion and effectively reduces motion artifacts in the resulting head CT images.

Common computed tomography artifact: source and avoidance

Background

Artifacts have significantly degraded the quality of computed tomography (CT) images, to the extent of making them unusable for diagnosis. The types of artifact that could be used are as follows: (a) streaking, which is commonly due to a discrepancy in a single measurement, (b) shading, which is due to a group of channels deviating gradually from the true measurement, (c) rings, which are due to errors in individual detector calibration and (d) distortion, which is due to helical reconstruction. It is occasionally possible to avoid scanning of a bony area, by means of changing the postion of the patient. Thus, this study aimed to evaluate the common artifacts that affect image quality and the method of correction to improve image quality.

Results

The data were collected by distributing a questionnaire to the CT technologist at different hospitals about the most common type of artifacts in the CT images, source of artifacts and methods of correction. A total of 95 CT technologists responded to the questionnaire, which included 67% males and 33% females. Most of the participants (70%) were experienced CT technologists, and 61% of the participants had not done any subspecialty CT scan courses. The most common artifact used in the CT departments was motion artifact in brain CT (73%), and the best method to reduce motion artifact was patient preparation (87%).

Conclusions

The most common shown artifact in this study was motion artifact, and the common cause was the patient-based artifact. It is important to understand why objects occur and how they could be prevented or suppressed to improve image quality.

Motion estimation and correction in SPECT, PET and CT

Patient motion impacts single photon emission computed tomography (SPECT), positron emission tomography (PET) and X-ray computed tomography (CT) by giving rise to projection data inconsistencies that can manifest as reconstruction artifacts, thereby degrading image quality and compromising accurate image interpretation and quantification. Methods to estimate and correct for patient motion in SPECT, PET and CT have attracted considerable research effort over several decades. The aims of this effort have been two-fold: to estimate relevant motion fields characterizing the various forms of voluntary and involuntary motion; and to apply these motion fields within a modified reconstruction framework to obtain motion-corrected images. The aims of this review are to outline the motion problem in medical imaging and to critically review published methods for estimating and correcting for the relevant motion fields in clinical and preclinical SPECT, PET and CT. Despite many similarities in how motion is handled between these modalities, utility and applications vary based on differences in temporal and spatial resolution. Technical feasibility has been demonstrated in each modality for both rigid and non-rigid motion, but clinical feasibility remains an important target. There is considerable scope for further developments in motion estimation and correction, and particularly in data-driven methods that will aid clinical utility. State-of-the-art machine learning methods may have a unique role to play in this context.

Rigid motion artifact reduction in CT using extended difference function

BACKGROUND

In computed tomography (CT), a patient motion would result in degraded spatial resolution and image artifacts.

OBJECTIVE

To eliminate motion artifacts, this study presents a method to estimate the motion parameters from sinograms based on extended difference function.

METHODS

Based on our previous work, we first divide the projection data into two parts according to view angles and take Radon transform. Then, we calculate the extended difference functions and search for the minimum points. The relative displacements can be determined by the minimum points, and the motion can be estimated by the relationships between the relative displacements and motion parameters. Finally, we introduce the estimated parameters into the reconstruction process to compensate for the motion effects.

RESULTS

The simulation results show that the running times can reduce by about 30% than our previous work. In phantom experiments, the relative mean rotation excursion (RMRE) and relative mean translation excursion (RMTE) of the new method are lower than the conventional Helgason-Ludwig consistency condition (HLCC) based method and comparable to our previous work. Compare with the HLCC method, the root mean square error (RMSE) of the new method also reduces, while the Pearson correlation coefficient (CC) and mean structural similarity index (MSSIM) increase.

CONCLUSIONS

The proposed new method yields the improved performance on accuracy of motion estimation with higher computational efficiency, and thus it can produce high-quality images.

Rigid motion artifact reduction in CT using frequency domain analysis

BACKGROUND

It is often unrealistic to assume that the subject remains stationary during a computed tomography (CT) imaging scan. A patient rigid motion can be decomposed into a translation and a rotation around an origin. How to minimize the motion impact on image quality is important.

OBJECTIVE

To eliminate artifacts caused by patient rigid motion during a CT scan, this study investigated a new method based on frequency domain analysis to estimate and compensate motion impact.

METHODS

Motion parameters was first determined by the magnitude correlation of projections in frequency domain. Then, the estimated parameters were applied to compensate for the motion effects in the reconstruction process. Finally, this method was extended to helical CT.

RESULTS

In fan-beam CT experiments, the simulation results showed that the proposed method was more accurate and faster on the performance of motion estimation than using Helgason-Ludwig consistency condition method (HLCC). Furthermore, the reconstructed images on both simulated and human head experiments indicated that the proposed method yielded superior results in artifact reduction.

CONCLUSIONS

The proposed method is a new tool for patient motion compensation, with a potential for practical application. It is not only applicable to motion correction in fan-beam CT imaging, but also to helical CT.

Correction for human head motion in helical x-ray CT

Correction for rigid object motion in helical CT can be achieved by reconstructing from a modified source-detector orbit, determined by the object motion during the scan. This ensures that all projections are consistent, but it does not guarantee that the projections are complete in the sense of being sufficient for exact reconstruction. We have previously shown with phantom measurements that motion-corrected helical CT scans can suffer from data-insufficiency, in particular for severe motions and at high pitch. To study whether such data-insufficiency artefacts could also affect the motion-corrected CT images of patients undergoing head CT scans, we used an optical motion tracking system to record the head movements of 10 healthy volunteers while they executed each of the 4 different types of motion ('no', slight, moderate and severe) for 60 s. From these data we simulated 354 motion-affected CT scans of a voxelized human head phantom and reconstructed them with and without motion correction. For each simulation, motion-corrected (MC) images were compared with the motion-free reference, by visual inspection and with quantitative similarity metrics. Motion correction improved similarity metrics in all simulations. Of the 270 simulations performed with moderate or less motion, only 2 resulted in visible residual artefacts in the MC images. The maximum range of motion in these simulations would encompass that encountered in the vast majority of clinical scans. With severe motion, residual artefacts were observed in about 60% of the simulations. We also evaluated a new method of mapping local data sufficiency based on the degree to which Tuy's condition is locally satisfied, and observed that areas with high Tuy values corresponded to the locations of residual artefacts in the MC images. We conclude that our method can provide accurate and artefact-free MC images with most types of head motion likely to be encountered in CT imaging, provided that the motion can be accurately determined.

Fiducial marker-based correction for involuntary motion in weight-bearing C-arm CT scanning of knees. Part I. Numerical model-based optimization

PURPOSE

Human subjects in standing positions are apt to show much more involuntary motion than in supine positions. The authors aimed to simulate a complicated realistic lower body movement using the four-dimensional (4D) digital extended cardiac-torso (XCAT) phantom. The authors also investigated fiducial marker-based motion compensation methods in two-dimensional (2D) and three-dimensional (3D) space. The level of involuntary movement-induced artifacts and image quality improvement were investigated after applying each method.

METHODS

An optical tracking system with eight cameras and seven retroreflective markers enabled us to track involuntary motion of the lower body of nine healthy subjects holding a squat position at 60° of flexion. The XCAT-based knee model was developed using the 4D XCAT phantom and the optical tracking data acquired at 120 Hz. The authors divided the lower body in the XCAT into six parts and applied unique affine transforms to each so that the motion (6 degrees of freedom) could be synchronized with the optical markers' location at each time frame. The control points of the XCAT were tessellated into triangles and 248 projection images were created based on intersections of each ray and monochromatic absorption. The tracking data sets with the largest motion (Subject 2) and the smallest motion (Subject 5) among the nine data sets were used to animate the XCAT knee model. The authors defined eight skin control points well distributed around the knees as pseudo-fiducial markers which functioned as a reference in motion correction. Motion compensation was done in the following ways: (1) simple projection shifting in 2D, (2) deformable projection warping in 2D, and (3) rigid body warping in 3D. Graphics hardware accelerated filtered backprojection was implemented and combined with the three correction methods in order to speed up the simulation process. Correction fidelity was evaluated as a function of number of markers used (4-12) and marker distribution in three scenarios.

RESULTS

Average optical-based translational motion for the nine subjects was 2.14 mm (± 0.69 mm) and 2.29 mm (± 0.63 mm) for the right and left knee, respectively. In the representative central slices of Subject 2, the authors observed 20.30%, 18.30%, and 22.02% improvements in the structural similarity (SSIM) index with 2D shifting, 2D warping, and 3D warping, respectively. The performance of 2D warping improved as the number of markers increased up to 12 while 2D shifting and 3D warping were insensitive to the number of markers used. The minimum required number of markers for 2D shifting, 2D warping, and 3D warping was 4-6, 12, and 8, respectively. An even distribution of markers over the entire field of view provided robust performance for all three correction methods.

CONCLUSIONS

The authors were able to simulate subject-specific realistic knee movement in weight-bearing positions. This study indicates that involuntary motion can seriously degrade the image quality. The proposed three methods were evaluated with the numerical knee model; 3D warping was shown to outperform the 2D methods. The methods are shown to significantly reduce motion artifacts if an appropriate marker setup is chosen.