The Clinical Value of the MAR⁺ Metal Artifact Reduction Algorithm for Postoperative Assessment of Lumbar Internal Fixation

Abstract

Background:

With the widespread use of lumbar pedicle screws for internal fixation, the morphology of the screws and the surrounding tissues should be evaluated. The metal artifact reduction (MAR) technique can reduce the artifacts caused by pedicle screws, improve the quality of computed tomography (CT) images after pedicle fixation, and provide more imaging information to the clinic.

Purpose:

To explore whether the MAR⁺ method, a projection-based algorithm for correcting metal artifacts through multiple iterative operations, can reduce metal artifacts and have an impact on the structure of the surrounding metal.

Materials and Methods:

A total of 57 patients who underwent lumbar spine CT examination after lumbar internal fixation from January to December 2023 in our hospital were retrospectively enrolled. The CT images were reconstructed using MAR⁺ and non-MAR⁺ techniques and were subdivided into MAR⁺ and non-MAR⁺ groups. The CT number (in Hounsfield units) and the SD noise values of the spinal canal, vertebral body, psoas major muscle, and adjacent fat were measured in the 2 groups of CT images, and the signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were calculated. The subjective score was evaluated by two diagnostic radiologists using a double-blind method for image quality evaluation of the MAR⁺ group and the non-MAR⁺ group, and the image quality was classified on a 5-point scale. The rank-sum test was utilized to compare the subjective and objective scores of the 2 groups.

Results:

The SD values of the spinal canal (Z=−4.12, P<0.01), vertebral body (Z=−3.81, P<0.01), and psoas major muscle (Z=−3.87, P<0.01) in the MAR⁺ group were significantly lower than those in the non-MAR⁺ group (P<0.05). However, the SD values of the adjacent fat (Z=−2.03, P=0.42) in the MAR⁺ group, although smaller than those in the non-MAR⁺ group, were not statistically significant. The CNR values of vertebral canal (Z=−2.67, P=0.008) and fat (Z=−2.60, P=0.009) were higher in the MAR⁺ group than in the non-MAR⁺ group, whereas the CNR values of the vertebral body (Z=−6.74, P<0.01) in the MAR⁺ group were smaller than those in the non-MAR⁺ group, and the difference of all of them was statistically significant (P<0.05). Furthermore, for both CT and SNR values, the MAR group’s values were all less than those of the non-MAR group and were statistically significant (P<0.05). The subjective scores of the measurement points were all higher in the MAR⁺ group than in the non-MAR⁺ group.

Conclusions:

The MAR⁺ technique has a noise reduction effect on different tissues and artifacts are significantly reduced. Although the artifacts caused by metal screws were not completely eliminated, the MAR⁺ technique was able to reduce the interference of artifacts in the diagnosis of CT images, thus improving their diagnostic quality.

The spine plays a very important role in supporting and protecting the body as well as in hematopoiesis. In recent years, the incidence of diseases such as traumatic or osteoporosis vertebral fractures, spondylolisthesis, spinal stenosis, or spinal deformity has been increasing. Percutaneous pedicle screw fixation is widely used in clinical practice not only to achieve rigid fixation of the vertebral body but also to treat many vertebral diseases. Compared with traditional open surgery, the development of percutaneous pedicle screw fixation provides a less invasive alternative for the treatment of lumbar fractures, which can not only reduce the operation time and hospital stay but also reduce postoperative bleeding and pain, which helps patients recover faster.¹ For the examination and review of patients with these metal screws implanted, the choice of an effective and convenient examination is necessary. Digital radiography shows limited anatomic details around the lumbar spine and internal fixation due to overlapping images. Magnetic resonance imaging is not routinely used in patients with metal implants, although it has superior soft tissue resolution due to its long scanning time and possible limitations due to the material of the metal implant. Computed tomography (CT) is the imaging modality of choice to evaluate postoperative outcomes and potential complications of lumbar internal fixation and adjacent anatomy.² CT plays a key role in accurately identifying the position and morphology of metal implants after surgery and in detecting associated complications such as loosening, fracture, and hematomas or abscesses.³

Pedicle screws are usually made of high-density metals such as titanium and cobalt alloys.⁴ These artifacts manifest as streaks and shadows around the implant due to photon starvation, photon scattering, and beam hardening.^5–8 Metal implants can cause various artifacts during CT scanning, and these metal artifacts are a diagnostic obstacle because they compromise the image quality of the tissue adjacent to the metal material. These streaked shadows can significantly blur the region of interest and are disruptive to the assessment of the implant itself and the interface between the implant and the vertebral body, as well as surrounding structures. The degree of artifacts is linked to the composition, thickness, size, orientation, and geometry of the metal implant.⁹ The presence of metal artifacts makes CT images difficult to interpret, thereby increasing the likelihood that surgical complications may be missed.¹⁰ In recent years, a variety of methods in CT scanning and image postprocessing, such as hardware filtering and changing CT scanning parameters, have been used to minimize the impact of metal artifacts on image quality. For example, higher tube voltage (kV) and tube current (mAs), virtual monochromatic imaging (VMI) using dual-energy CT (DECT), and postprocessing metal artifact reduction (MAR) reconstruction algorithms.^11–14 Conventional filtered back-projection (FBP) reconstruction algorithms, which are commonly used in CT image reconstruction, are unable to reconstruct images clearly due to inaccurate projection data. Many MAR algorithms have been suggested to solve the serious metal artifact problem. MAR is an automated projection-based postprocessing technique that corrects for metal artifacts by segmentation and reconstruction based on CT number thresholding, revealing structural details obscured by metal artifacts and reconstructing a more accurate image.^15–17 The most commonly utilized MAR algorithm is the Interpolated Sinusoidal Map Repair MAR algorithm, which aims to interpolate missing projection data by replacing corrupted data from nearby slice information and then reconstructing the image utilizing FBP using adjacent data.¹⁸

This study intends to compare images before and after the use of the MAR⁺ technique through observational analyses of clinical cases to investigate whether the MAR⁺ method reduces metal artifacts while at the same time having an impact on the tissue structure around the metal.

MATERIAL AND METHODS

Statement

G.W. were from the industry (CT Business Unit, Neusoft Medical System), and the data were controlled by J.F. The data are free of falsification and bias.

Ethics Approval

The local ethics committee approved the study. The study was approved by the Medical Ethics Committee of the hospital (decision number: BY2024026).

Patient Cohort

A total of 57 patients who underwent lumbar spine CT after lumbar internal fixation were enrolled in this study between January and December 2023. The study population consisted of 28 males and 29 females, with an age range of 17 to 78 years and an average age of 55.56 years. The 2 sets of general data were found to be comparable (P>0.05). Inclusion criteria: (1) all patients underwent general lumbar CT scanning, and the imaging data were complete; (2) all patients were postimplanted after lumbar spine metal internal fixation; (3) the lumbar spine image includes the entire internal fixation material and at least one vertebral body beyond the upper and lower edges of the internal fixation material. Patients were excluded if they were pregnant or breastfeeding or if they exhibited poor cooperation.

CT Imaging Acquisition and Reconstruction

A 126-slice spiral CT scanner (NeuViz Extra CT, Neusoft Medical Systems Co. Ltd., Shenyang, China) was utilized to scan the lumbar internal fixation area in patients with lumbar metal internal fixation. The patient is placed in the supine position with arms raised, and the scanning area includes the metal endoprosthesis and more than 1 vertebra above and below, and the scanning direction is from cephalad to caudal. The following scan parameters: collimation=32×0.625 mm; rotation=0.5 seconds; pitch=0.5; tube voltage 120 kV, automatic tube current, reference value 300 mAs; layer thickness 1 mm; layer spacing 1 mm; scan field 160×160 mm²; matrix 512×512. Image reconstruction was performed at routine lumbar spine scans using the ClearView 50% reconstruction algorithm. According to whether the MAR⁺ function is applied, it is divided into the MAR⁺ group and the non-MAR⁺ group. The scanned images were saved in the Neusoft postprocessing workstation.

Image Evaluations

(1) Image objective evaluation: 4 circular region of interests (ROIs) were manually selected at the level where the pedicle screws produced the largest artifacts in combination with soft tissue and bone windows (spinal canal, psoas muscle, adjacent fat, vertebral body).¹⁹ The area was 50±1 mm² and the CT and SD values were measured and repeated 2 times, and the average value was considered as the final result. The AVW (Neusoft Medical Systems Co. Ltd.) workstation was used only for measurement, and the level location and size of the CT images of the MAR⁺ group and the non-MAR⁺ group were consistent (Figs. 1, 2). Signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) are calculated in accordance with the specified formulae.²⁰

FIGURE 1.

A man, 61 years old, underwent transforaminal posterior lumbar decompression and fusion for corresponding lumbar spinal stenosis due to a bulging disc at L4-5. (A) Axial bone window image at the level of the pedicle in the non-MAR⁺ group. (B) Axial soft tissue window image at the same level in the non-MAR⁺ group. (C) Axial bone window image at the level of the pedicle in the MAR⁺ group. (D) Axial soft tissue window image at the same level in the MAR⁺ group.

FIGURE 2.

A 53-year-old male who underwent percutaneous pedicle screw fixation for a compression fracture of the L1 lumbar vertebra. (A) Axial bone window image at the level of the pedicle in the non-MAR⁺ group. (B) Axial soft tissue window image at the same level in the non-MAR⁺ group. (C) Axial bone window image at the level of the pedicle in the MAR⁺ group. (D) Axial soft tissue window image at the same level in the MAR⁺ group.

SNR=CT_ROI / SD_ROI, the CT_ROI and SD_ROI are the CT and SD values within the 4 ROIs;

CNR=(CT_ROI−CT _{psoas muscle})/SD _{psoas muscle}, CT_ROI is the CT value within the 4 ROIs, and the CT _{psoas muscle} and SD _{psoas muscle} are the CT value and SD value of the psoas muscle.

(2) Image subjective evaluation: 2 senior radiologists evaluated the image quality of the MAR⁺ group and the non-MAR⁺ group by a double-blind method, selecting the soft tissue window and bone window images with the largest layer of pedicle screw artifact. Both physicians employed a 5-point Likert scale to assess image quality. A score of 5 indicated the presence of few or no artifacts, allowing for a complete diagnosis. A score of 4 indicated the presence of mild artifacts, which did not affect the diagnostic process. A score of 3 indicated the presence of moderate artifacts, which still permitted diagnostic accuracy. A score of 2 indicated the presence of severe artifacts, which made it challenging to make a diagnosis. A score of 1 indicated the presence of severe artifacts, rendering the image undiagnostic. A score of at least 3 is deemed sufficient for clinical diagnosis.

Statistical Analysis

The SPSS 27.0 software was employed for statistical analysis, with the resulting data expressed as mean±SD for continuous variables and as frequency for categorical variables. The objective evaluation parameter values tested did not conform to a normal distribution. The CT and SD values and subjective scores of the spinal canal, vertebral body, psoas muscle, and adjacent fat between the MAR⁺ group and the non-MAR⁺ group were analyzed using the Wilcoxon signed-rank test. The consistency of subjective scoring between 2 radiologists for groups of MAR⁺ and non-MAR⁺ images was tested using κ analysis. A κ value of ≥0.75 indicates good consistency, 0.40 to 0.74 indicates moderate consistency, and <0.40 indicates poor consistency. In this study, a P-value of <0.05 was considered to be statistically significant, and the test level was set at α=0.05. The test was statistically significant and conducted on both sides.

RESULTS

Objective Evaluation Results

Comparison of CT Values

ROI-based analysis revealed that CT values for all 4 regions (spinal canal, vertebral body, psoas muscle, and adjacent fat) were reduced in the MAR⁺ group compared with the non-MAR⁺ group, and the difference was statistically significant (P<0.05) (Table 1).

TABLE 1.

Comparison of Objective Values in Groups MAR⁺ and Non-MAR⁺

	MAR+	Non-MAR+	Z	P
Spinal canal
CT (HU)	49.1±21.5	94.9±49.0	−5.47	＜0.01
SD	55.9±10.9	66.5±17.0	−4.12	＜0.01
SNR	0.89±0.40	1.48±0.79	−4.20	＜0.01
CNR	0.32±0.67	−0.16±1.34	−2.67	0.008
Vertebral body
CT (HU)	93.8±39.5	260.1±81.7	−8.62	＜0.01
SD	71.4±22.7	94.7±48.5	−3.81	＜0.01
SNR	1.36±0.54	3.04±1.12	−7.54	＜0.01
CNR	1.42±1.01	3.47±1.82	−6.74	＜0.01
Psoas muscle
CT (HU)	35.7±14.3	103.8±37.0	−8.74	＜0.01
SD	42.4±8.9	46.8±7.5	−3.87	＜0.01
SNR	0.88±0.43	2.22±0.74	−8.33	＜0.01
Adjacent fat
CT (HU)	−116.3±19.3	−85.3±18.5	−6.90	＜0.01
SD	39.9±6.9	41.7±8.3	−2.03	0.42
SNR	−2.89±0.74	−2.25±0.67	−4.52	＜0.01
CNR	−3.72±0.87	−4.10±0.86	−2.60	0.009

Values are given as mean HU and SD.

P<0.05 was considered to indicate a statistically significant difference.

CNR indicates contrast-to-noise ratio; CT, computed tomography; MAR, metal artifact reduction; SNR, signal-to-noise ratio.

Comparison of SD Values

The SD values of all ROIs in the MAR⁺ group were lower than those in the non-MAR⁺ group. The spinal canal demonstrated a significant negative correlation with Z=−4.12 (P<0.01), the vertebral body with Z=−3.81 (P<0.01), and the psoas muscle with Z=−3.87 (P<0.01). The SD values for adjacent fat (Z=−2.03, P=0.42) were not statistically significant, although they were similarly lower in the MAR⁺ group than in the non-MAR⁺ group (Table 1).

Comparison of SNR Values

The SNR values of all ROIs in the MAR⁺ group were lower than those in the non-MAR⁺ group. The spinal canal (Z=−4.20, P＜0.01), vertebral body (Z=−7.54, P＜0.01), psoas muscle (Z=−8.33, P＜0.01), adjacent fat (Z=−4.52, P＜0.01) exhibited statistically significant differences (P<0.05) (Table 1).

Comparison of CNR Values

The CNR values of the vertebral canal (Z=−2.67, P=0.008) and adjacent fat (Z=−2.60, P=0.009) were higher in the MAR⁺ group than in the non-MAR⁺ group. Conversely, the CNR values of the vertebral body (Z=−6.74, P<0.01) were observed to be lower in the MAR⁺ group than in the non-MAR⁺ group (Table 1).

Subjective Evaluation Results

In terms of the display of the spinal canal, vertebral body, psoas muscle, and adjacent fat interface around the screw, the image score of the MAR⁺ group was significantly higher than that of the non-MAR⁺ group, with a statistically significant difference (P<0.05). In the consistency test for the MAR⁺ group between the 2 radiologists, the κ value=0.814, and for the non-MAR⁺ group, the κ value=0.802. The P<0.05 indicates that the consistency test results of the 2 test results were statistically significant and the consistency was good (Tables 2 and 3).

TABLE 2.

Consistency Test of Subjective Scores of 2 Radiologists in Groups MAR⁺ and Non-MAR⁺

			Radiologist B
Radiologist	Group	Score	1	2	3	4	5	κ	P
Radiologist A		1	0	0	0	0	0	0.814	＜0.01
		2	0	0	0	0	0
	MAR+	3	0	0	5	1	0
		4	0	0	1	42	1
		5	0	0	0	1	6
	Non-MAR+	1	1	0	0	0	0	0.802	＜0.01
		2	0	10	2	0	0
		3	0	3	38	0	0
		4	0	0	0	3	0
		5	0	0	0	0	0

MAR indicates metal artifact reduction.

TABLE 3.

Comparison of Subjective Values in Groups MAR⁺ and Non-MAR⁺

Subjective evaluation	MAR+	Non-MAR+	Z	P
Radiologist A	4.02±0.48	2.80±0.54	−8.670	＜0.01
Radiologist B	4.05±0.51	2.79±0.55	−8.645	＜0.01

MAR indicates metal artifact reduction.

DISCUSSION

For patients with metallic orthopedic implants, being able to visualize the bone structure and the relationship between the implant and the bone and surrounding soft tissues is key to ruling out complications such as implant loosening or fractures. The effect of using MAR⁺ technology to remove metal artifacts was confirmed in this study, and it was found that for lumbar implants, MAR⁺ images can significantly improve image quality. In the study, the image quality and value of using MAR⁺ to reduce metal artifacts in CT scans of lumbar implants were evaluated. When there is a metal foreign body, such as dentures or metal fixations, within the CT scan field or reconstruction field of the human body being scanned, the extremely large density difference between the metal foreign body and the surrounding critical tissues leads to radially high-density artifacts in the image. The artifacts obscure the visualization of the surrounding tissues, thus limiting their diagnostic quality. MAR⁺, by decreasing a human body model, replaces the algorithm in the abnormally high-density areas, thereby eliminating high-density artifacts and restoring the actual tissue structure. The subjective image quality was much improved, from a situation where severe artifacts significantly affected the accuracy of diagnosis to one with only minor artifacts that do not affect the accuracy of diagnosis. This underscores the clinical importance of the technology in reducing image artifacts. In this study, by assessing the subjective and objective scores of the MAR⁺ group and the non-MAR⁺ group after lumbar fixation surgery, taking into full consideration the different surrounding tissue structures, the results showed that the SD values of different tissues in the MAR⁺ group were all less than those in the non-MAR⁺ group. This indicates that the MAR⁺ technique has a noise-reducing effect on different tissues, and the artifacts were significantly decreased. In this study, in the region of the vertebral body and psoas muscle, where metal artifacts were more severe, the CT values after treatment with the MAR⁺ technique were 93.8±39.5 and 35.7±14.3 HU, respectively, which were significantly lower than the pretreatment values of 260.1±81.7 and 103.8±37.0 HU, and were closer to the CT values of normal tissues. At the same time, the SD values of most regions were also significantly reduced, with an average decrease of 35.44% in the SD values after the MAR⁺ technique compared with the SD values before the treatment. Among them, the SD value of the psoas muscle, which is more affected by the high-density artifacts, was reduced by 65.64% after the MAR⁺ technique, and the difference was statistically significant (P<0.05). In addition, in terms of subjective scores, the 2 radiologists showed good subjective agreement and an average improvement in diagnostic scores of 30.7%, and again, the difference in subjective scores was statistically significant. However, it should be pointed out that the SNR of any anatomic region did not increase, so our objective results may only indicate a changed distribution of artifacts, which is consistent with the findings of Feldhaus et al.²¹ Although the use of MAR⁺ technology did not completely eliminate the metal artifacts caused by metal screws, it was able to reduce their impact on CT images, thereby improving image quality.

Many previous studies have evaluated iterative MAR algorithms to reduce artifacts in vertebral bodies and other implanted areas. However, there are differences in the effectiveness of MARs depending on the different metal implants used in orthopedic surgery and the different CT vendors, with numerous studies yielding different results, depending on the algorithm used.²² Huber et al²³ compared different CT MAR strategies in an animal sheep study using titanium pedicle screws. It was demonstrated that, in addition to the primary factor of hardware material, there are also significant differences in the efficiency of different MAR reconstruction methods. Increasing the availability of model-based iterative reconstruction in the clinical setting may help to reduce radiation dosage further. Compared with the standard weighted filtered back-projection reconstruction method, MAR reconstruction has been proven to significantly improve the visualization of blurred soft tissue structures and diagnostic quality.²⁴ Currently, many scholars are using methods such as DECT, dual-source CT, and spectral CT to eliminate metal artifacts.^{2,3,9,15,25–27} However, these methods involve repeated scanning or simultaneous scanning at two different energy levels. Although the removal of metal artifacts is significant, it also results in a noticeable increase in the radiation dose perceived by the patient. In this paper, the MAR⁺ technique is a post-reconstruction processing technology that eliminates metal artifacts through multiple iterative calculations using corrections to the projection data without increasing the radiation dose of the scan. It is suitable for orthopedic large metal implants, such as the screws used in vertebral fixation surgery in this study. On the other hand, the projection-based MAR⁺ algorithm utilizes a Hounsfield unit threshold to detect and segment the metal part from the original image, allowing for compensation of the image for the photon starvation phenomenon caused by the metal using a designated algorithm. Therefore, while reducing bright and dark band artifacts caused by metal hardware, it also preserves the contrast enhancement of iodine.²⁸ Moreover, this method can be applied retrospectively, permitting physicians to decide whether to apply the algorithm after reviewing the scanned images.²⁹ Although the application of the MAR⁺ method has clear benefits in reducing metal artifacts, it must also be acknowledged that MAR⁺ may introduce its own unique new artifacts.^21,22,30,31 The MAR⁺ technique effectively restores the original image information in the process of removing metal artifacts, but the repair effect is poor in structures with severe metal artifacts that are very close to the metal area. These new artifacts introduced by MAR⁺ are not unique to a specific MAR algorithm or CT scanning equipment,²² and the reason may be that MAR⁺ failed to correctly identify the pedicle screws, mistakenly identifying part of the screws as damaged projection data, and removing them. The best way to avoid this pitfall is still to keep the original images without MAR⁺ for review. Therefore, in clinical practice, it may be necessary to create multiplanar reconstructed images and combine the 2 sets of images before and after artifact removal in conjunction with the original images, viewing them from multiple perspectives to avoid misinterpretation.

In recent years, high-kiloelectron volts synthetic VMI based on DECT-derived imaging has played an important role in minimizing metal artifacts.^32–37 DECT can obtain 2 multicolor spectra, whereas conventional CT obtains only one. This additional information can be used to generate virtual single-energy images for more clinically useful information. However, it has been suggested that the VMI implementation may not be as effective as the MAR algorithm in removing metal artifacts when dealing with severe metal artifacts caused by bilateral hip implants.³² On the other hand, Nils Große et al³⁵ reported that both the VMI and MAR algorithms reduced the spinal metal artifacts, and there was no clear advantage between the 2 methods. Furthermore, Patzer et al³³ showed that the combined application of VMI ≥110 keV and iMAR leads to the least metal artifacts. At the same time, research by Layer et al³⁴ have also shown that VMI 100 keV has achieved the best results in combination with MAR. Therefore, the combination of the 2 methods may be promising, and we can compare and analyze the 2 methods in further work.

In addition, this study has the following limitations. Firstly, the relatively small number of patients included in this study may have led to reduced reliability; follow-up work could expand the patient population for further study. Secondly, the study did not differentiate based on various metal compositions, although the majority of the implants were titanium alloys. Different types of titanium alloys have unique shapes and compositions, and the effectiveness of postprocessing techniques to reduce metal artifacts also varies. In the future, we may distinguish according to the different materials. Lastly, this study only focused on lumbar spinal fixation and did not evaluate other metal fixation devices in different parts of the body, which could be the subject of further research.

CONCLUSION

In conclusion, the MAR⁺ technology can be an effective tool to reduce metal artifacts and can significantly improve image quality and evaluate metal implants. The application of MAR⁺ technology can reduce the metal artifacts caused by metal screws after lumbar pedicle fixation, which is beneficial for accurately observing the position of the screws and the surrounding tissue structure. In summary, MAR⁺ technology can be an effective tool to reduce metal artifacts and can significantly improve image quality and evaluate metal implants.