Improvement of image quality and diagnostic confidence using Smart MAR – a projection-based CT protocol in patients with orthopedic metallic implants in hip,spine,and shoulder

Abstract

Background

In computed tomography (CT) scans, artifacts caused by metallic orthopedic implants still hamper the visualization of important, periprosthetic tissues. Smart MAR metal artifact reduction tool is a promising three-stage, projection-based, post-processing algorithm.

Purpose

To determine whether the Smart MAR tool improves subjective and objective image quality and diagnostic confidence in patients with orthopedic implants of the hip, spine, and shoulder.

Material and Methods

Seventy-two patients with orthopedic screws, hip/shoulder replacement, or spine spondylodesis were included. CT scans were performed on a single-source multislice CT scanner, raw data were post-processed using Smart MAR. Image quality was evaluated both quantitatively (ROI-based) and qualitatively (rater-based) and compared to iterative reconstructions (ASIR V). As comparative standard for artificial prosthetic breaks or loosening, follow-up examinations were used.

Results

Smart MAR reconstructions of the hip (n = 23), spine (n = 26), and shoulder (n = 23) showed a significantly reduced attenuation and noise of regions adjacent to metallic implants (P<0.002). Subjective image quality (P<0.005, shoulder P = 0.038/P = 0.046) and overall diagnostic confidence were higher in Smart MAR (all regions P<0.002). Signal-to-noise ratio (SNR; P = 0.72/P = 0.96) was not improved. Compared to standard ASIR V new, artificial metal extinctions (up to 50%) or periprosthetic hem lines (48%–73%) were introduced by Smart MAR.

Conclusion

Smart MAR improved image quality of the hip, spine, and shoulder CT scans resulting in higher diagnostic confidence in evaluation of periprosthetic soft tissues. As shown for spine implants, it should be used with caution and as a complementary tool for evaluation of periprosthetic loosening or integrity of metal implant, as in many cases it introduced new artifacts.

Keywords

Metal artifact computed tomography image reconstruction hip/shoulder prosthesis hip replacement

Introduction

Computed tomography (CT) is still the most important diagnostic tool to evaluate post-surgical periprosthetic complications as displacement and loosening of screws or arthroplasty (1). Metal artifacts caused by osteosynthetic material still impair image quality and hence degrade the diagnostic yield of CT scans. Smaller structures such as periprosthetic abscesses, bleedings, or small tumors can be obscured. For both hip and shoulder regions, but especially for spine surgery, evaluation of adjacent neurovascular and musculoskeletal tissues is most important as well as the detection of prosthetic breaks or loosening (2).

Besides conventional techniques (increasing tube current and increasing kV and peak voltage, narrow collimation), metal-related artifacts can be reduced on conventional CT scanners to some extent in different image reconstruction algorithms such as iterative and projection-based approaches (3 –5) . Other options including dual-energy techniques have become available more recently (6,7). Recent studies of raw-data-based metal artifact reduction (MAR) with iterative or projection-based algorithms reported a significant reduction of metal artifacts and improved evaluation of periarticular soft tissues compared to iterative projections alone (3,4,8 –12). Some have even been used to improve dual-energy CT (DECT) raw data (13,14).

The purpose of the present study was to investigate whether the new, commercially available Smart MAR tool, a projection-based metal artifact reduction algorithm, improves image quality and diagnostic confidence in patients with orthopedic metallic implants in the hip, spine, or shoulder region compared to standard, well-established iterative reconstruction technology (GE ASIR V) (5 –7,15).

Material and Methods

Patients and study design

Seventy-two consecutive patients with orthopedic screws or partial/total replacement of the hip (n = 23) and shoulder (n = 23) or spondylodesis of the spine (n = 26) were included in this study. All patients had a clinical indication for a CT scan; indications were post-surgical pain, position control, suspicion of hematoma, abscess, or loosening of implant. No examinations were performed only for study purposes.

Inclusion criteria were: (i) age > 18 years; (ii) clinical indication for a CT of respective region; (iii) presence of unilateral hip or shoulder implant; and (iv) written consent. Exclusion criteria were contraindication for CT scan, e.g. pregnancy, claustrophobia.

Only non-contrast CT scans were included. Raw data were retrieved retrospectively. The image quality of Smart MAR-based reconstruction series was evaluated both quantitatively (six regions of interest [ROI]) and qualitatively (two independent raters) and compared to standard reconstructions. As the comparative standard for artificial prosthetic breaks or loosening of prosthesis, clinical and radiological follow-up examinations within one year were used.

The local ethics committee approved the study; all patients were aged > 18 years and written informed consent was obtained. The study was approved by the local ethics committee (decision number: (EA4/035/17)). No funding was received to conduct this study.

CT scan parameters

All CT scans were obtained with a single-source, 64-row multislice CT scanner (MSCT; Revolution Evolution, GE Healthcare, Milwaukee, WI, USA) using a single-source X-ray tube with the following scan parameters: collimation = 64 × 0.625 mm; rotation time = 0.4 s; pitch = 0.9; due to automatic dose modulation tube voltage and current were determined by the patient’s body mass index (in the range of 120–140 kV and 100–430 mAs).

CT image reconstruction

For iterative reconstruction of the raw data, adaptive iterative reconstruction technique (40% adaptive statistical iterative reconstruction, ASIR V, GE Healthcare, Milwaukee, WI, USA) was used in both scan series with the same iterative reconstruction settings (15,16). The Smart MAR post-processing tool (GE Healthcare, Milwaukee, WI, USA) is approved for the reconstruction of iterative primary reconstructions. It automatically generates projection-based, reconstructed images from the acquired CT data in three steps. First, corrupted areas in the projection that correspond to metallic objects are identified and metal traces are segmented from the sinogram. Second, inpainted data are generated by interpolation, which serves to replace metal-corrupted projections where metal sinogram is blended out. The corrected data are generated using the forward projection of the classified image. Finally, the final corrected projection is generated using a combination of the original projection data and the inpainted projections (17). However, detailed information on the (validated) algorithms, such as the exact interpolation method or projection thresholds, is not available to users. Thin (0.625 mm) and thick (2.5 mm) axial slices were reconstructed and multiplanar reconstructions (coronal and sagittal) were generated in slice thickness of 2.5 mm with the standard reconstruction kernel. Images could be displayed using a soft tissue and bone window chosen by the rater.

CT image analysis

Image quality was evaluated both quantitatively and qualitatively using the picture archiving and communication system (PACS) workstation (Centricity Radiology RA1000; GE Medical Systems). Image quality of the routine ASIR V CT datasets and the datasets generated using Smart MAR reconstruction was compared.

Quantitative image analysis

Quantitative measurements of mean CT number (in Hounsfield units [HU]) and the SD (i.e. image noise) were recorded in circular ROIs (Fig. 1). ROIs were placed in the corresponding soft tissue window and in thick reconstructions (slice thickness of 2.5 mm) for both the standard ASIR V and the Smart MAR datasets: ROI_1, area including implant and most severe artifact for best representation of metal artifacts (11,13,18); ROI₂, adjacent vessel; ROI₃, adjacent muscle; ROI₄, fatty tissue; ROI₅, surrounding air for measuring background noise by recording SD; and ROI₆, adjacent bone marrow. In both dataset series, ROIs were placed as large as possible and were identical in shape and area.

Fig. 1.

Quantitative measurements of mean CT number (in HU) and the SD (i.e. image noise) were recorded in identical, circular ROIs. (a–c) Standard ASIR V; (d–f) Smart MAR, projection-based reconstructions. CT, computed tomography; HU, Hounsfield Unit; ROI, region of interest.

Signal-to-noise ratio (SNR), defined as the mean attenuation of ROI₁ (area including implant and most severe artifact) divided by SD of ROI₅ (placed in air) was calculated using the following formula: SNR = ROI₁/SD_ROI5. ROI measurements and calculated SNR were considered as parameters for objective image quality.

Qualitative image analysis

Two radiologists, with 14 and 5 years of experience in reading CT scans, reviewed the two series of datasets independently and in random order. Both readers were blinded to the presence of significant pathology (i.e. hematoma, screw/implant fracture) and to the applied reconstruction technique. They selected the axial slices with the most severe artifacts and compared the corresponding slices in both series. Axial slices were evaluated in the soft tissue window, which could be adapted (standard level = 40, width = 350). Item 2 and 3, regarding evaluation of adjacent trabecular and cortical bone, were evaluated in bone window (level = 700 HU, width = 2700 HU).

The following features were evaluated subjectively and independently: qualitative noise (range 1 = unacceptable to 5 = minimal); delineation of periprosthetic trabeculae (1 = very poor to 5 = excellent); visualization/delineation of periprosthetic cortical bone (1 = unacceptable to 5 = excellent); visualization of periprosthetic soft tissue (1 = insufficient to 5 = fully diagnostic evaluation); and extent of artifacts (1 = major streak artifacts to 5 = no artifact). Furthermore, readers rated their diagnostic confidence as a summarizing item (1 = no confidence to 5 =completely confident). The distribution and volume of metal artifacts were evaluated subjectively using a 4-point rating scale: 3 = extensive (> 75% artifacts of axial slice); 2 = moderate (25%–75% implant artifacts of axial slice); 1 = minimal (< 25% implant bridge and/or < 5 pin implants); and 0 = no artifacts. In a second step, in a semi-quantitative evaluation, both raters evaluated artificial periprosthetic loosening or artificial hem lines/perihardware lucency (0 = no, 1 = 1–2 mm, 2 = > 2 mm) and prosthetic screws/implant breaks (0 = no, 1 = probable break, 2 = clear break) in both series in a 3-point rating scale.

Dose analysis

Radiation dose was analyzed in terms of dose length products (DLP) and CT dose indices (CTDIvol), which were taken from the dose protocol for each patient.

Statistical analysis

Statistical analysis was performed using SPSS 24.0 software (SPSS, Chicago, IL, USA). Continuous variables are provided as mean ± SD. Differences in all objective image quality parameters were analyzed by means of the paired-sample T test. Subjective image quality was analyzed by averaging the two 5-point Likert scores and using the non-parametric Wilcoxon signed-rank test. Interrater agreement was analyzed by Cohen’s kappa (κ) analysis (coefficient interpretation: κ: 0.21–0.40 = fair agreement, κ: 0.41–0.60 = moderate agreement; κ: 0.61–0.80 = substantial agreement; κ: 0.81–1.0 = almost perfect agreement) (19). A P value < 0.05 was considered to indicate a statistically significant difference. The given alpha value (α = 0.05) was adjusted using Bonferroni correction for multiple comparison of several (in our case, n = 106) statistical tests. Multiple alpha (α`) was set to α` = α/25 = 0.00047 (21). All P values in the text, figures, and tables are reported without Bonferroni correction.

Results

Patient and scan parameters

For patient characteristics, details of implant distribution, and dose parameters see Table 1. In all 72 patients, CT examinations were performed technically successful without differences in acquisition.

Table 1.

Patient characteristics, CT findings, and radiation doses.

	Hip (n = 23)	Shoulder (n = 23)	Spine (n = 26)
Mean age (years)	63.9 ± 21.6	64.4 ± 14.4	67.0 ± 17.2
Male/female	12/11 (48% women)	14/9 (39% women)	14/12 (46% women)
Screws	12 (48)	9 (39)	Lumbar	14 (54)
Hemi endoprosthesis	5 (22)	7 (30)	Thoracic	10 (39)
Total endoprosthesis	6 (26)	7 (30)	Cervical	2 (8)
Dose (mean ± SD)
CTDIvol pelvis (mGy)	15.2 ± 5.8	17.6 ± 6.5	11.3 ± 3.0
DLP total (mGy*cm)	518.5 ± 188.8	365.5 ± 143.3	372.6 ± 187.2

Values are given as n (%) or mean ± SD.

CT, computed tomography; CTDI, CT dose index; DLP, dose length product.

Quantitative analysis

Table 2 gives an overview of the results of the objective image analysis, for examples see Figs. 2 and 3. ROI-based analysis showed significantly decreased mean attenuation for all three anatomical regions only regarding main artifact area including the implant (ROI₍₁₎). Image background noise decreased significantly only for the shoulder region, whereas there was a slight tendency for reduced background noise in the hip and spine series. In contrast, most SD values of adjacent tissues of interest (main artifact area, adjacent vessel, muscle, or fat) as parameters for artifact caused heterogeneity were significantly lower in all regions with Smart MAR reconstructions (all P < 0.002).

Table 2.

Results of quantitative ROI-based analysis.

	Hip (n = 23)					Shoulder (n = 23)					Spine (n = 26)
	Standard		Smart MAR			Standard		Smart MAR			Standard		Smart MAR
	HU	SD	HU	SD	P	HU	SD	HU	SD	P	HU	SD	HU	SD	P
ROI1	472.1	205.4	345.3	152.5	<0.001	713.3	357.7	542.5	282.1	<0.001	577.3	201.6	395.1	152.0	<0.001
ROI2	54.7	66.7	43.9	17.4	0.41	23.0	89.8	37.2	54.9	0.80	29.4	93.1	40.0	65.6	0.57
ROI3	78.8	67.3	43.0	23.1	0.01	61.6	195.6	46.3	36.6	0.07	41.2	181.2	35.0	44.0	0.85
ROI6	398.4	388.9	257.0	190.7	0.02	563.5	483.5	333.1	276.4	0.07	473.9	475.8	373.1	503.8	0.12
SD ROI1	864.5	189.4	705.6	215.0	<0.001	976.3	303.6	870.8	352.0	<0.001	865.4	157.3	715.5	190.7	<0.001
SD ROI2	51.5	37.4	21.0	6.3	<0.001	80.6	50.6	40.4	24.0	<0.001	71.5	49.1	46.1	30.8	<0.001
SD ROI3	51.7	30.1	21.6	7.7	<0.001	87.8	62.9	35.8	13.7	0.001	102.7	96.9	53.9	36.1	0.002
SD ROI4	46.6	21.3	20.7	4.8	<0.001	76.9	50.1	34.9	19.5	<0.001	94.1	112.1	58.6	51.3	0.06
SD ROI5	35.9	19.2	30.2	29.6	0.15	43.6	25.8	24.8	13.3	<0.001	83.9	49.5	73.0	45.3	0.09
SD ROI6	217.7	145.0	101.4	82.2	0.001	284.5	201.5	158.9	174.7	<0.001	350.7	195.1	206.7	170.8	<0.001
SNR	19.8	18.5	20.8	17.4	0.6	21.8	17.5	33.7	36.5	0.09	11.6	11.1	8.1	6.4	0.016

Values are given as mean HU and SD.

P < 0.05 was considered to indicate a statistically significant difference (values in bold).

HU, Hounsfield Unit; SNR, signal-to-noise ratio.

Fig. 2.

Case reports (SL, 2.5 mm, respectively; non CE; soft-tissue kernel). (a–c) Standard; (d–f) Smart MAR reconstructions show reduced noise and improved visualization of periprosthetic tissues. (a, d) A 42-year-old man with implant after fracture of humeral head, still suffering from pain. No prosthesis-related complications were detected. (b, e) A 67-year-old woman with spondylodesis after compression fracture of vertebra, no large paravertebral abscess was detected. (c, f) A 72-year-old women, improved visualization of pelvic structures, no lymphatic nodes or abscesses.

Fig. 3.

Case reports (SL, 2.5 mm, respectively; non CE; bone and soft-tissue kernel). Circles outline artificially generated, induced artifacts as hem lines/periprosthetic lucency and artificial distortion in Smart MAR reconstructions. Hence, both background noise is reduced and visualization of periprosthetic tissues is improved. (a, d) A 68-year-old woman with pain after spondylodesis. No prosthesis-related complications were detected. (b, e) A 71-year-old man with suspected loosening of femoral endoprosthesis. New periprosthetic lucency in Smart MAR. (c, f) A 62-year-old man with limited mobility after joint replacement without prosthesis-related pathology but periprosthetic artifacts in the CT scan.

SNR was not increased in any of the anatomical regions, for spine region standard ASIR V even showed a higher SNR compared to Smart MAR (P = 0.016).

Qualitative analysis

The median values of subjective rating with related P values and Kappa coefficients of the reader´s image quality scores are summarized in Tables 3 and 4.

Table 3.

Results of qualitative analysis by two readers using a 5-point Likert scale.

	Hip (n=23)					Shoulder (n=23)					Spine (n=26)
	Standard		Smart MAR			Standard		Smart MAR			Standard		Smart MAR
	Ø	SD	Ø	SD	P	Ø	SD	Ø	SD	P	Ø	SD	Ø	SD	P
Noise
R1	3.1	0.4	3.9	0.5	<0.001	2.8	0.4	3.1	0.3	0.038	3.0	0.3	3.3	0.5	0.005
R2	3.0	0.2	3.8	0.5	<0.001	2.9	0.3	3.1	0.2	0.046	3.0	0.0	3.4	0.5	0.001
Trabecular bone
R1	2.7	0.9	3.7	0.6	<0.001	2.0	1.1	3.2	1.0	<0.001	2.4	0.6	3.3	0.7	<0.001
R2	2.8	0.9	3.7	0.6	<0.001	2.1	1.1	3.1	1.1	<0.001	2.6	0.7	3.7	0.7	<0.001
Cortical bone
R1	2.7	1.0	4.3	0.6	<0.001	2.2	1.2	3.3	1.1	<0.001	2.7	0.6	3.5	0.8	<0.001
R2	2.8	1.0	4.5	0.7	<0.001	2.1	1.1	3.1	1.1	<0.001	2.7	0.7	3.9	0.9	<0.001
Soft tissue
R1	3.2	1.0	4.3	0.6	<0.001	2.6	0.9	3.6	0.9	<0.001	3.1	0.5	3.9	0.7	<0.001
R2	3.4	1.0	4.5	0.7	<0.001	2.5	0.9	3.5	0.9	0.001	3.3	0.6	4.4	0.6	0.004
Artifacts
R1	3.0	0.7	4.0	0.5	<0.001	2.3	0.8	3.6	0.8	<0.001	2.9	0.4	3.4	0.8	<0.001
R2	3.0	0.6	4.1	0.5	<0.001	2.5	0.9	3.5	0.9	<0.001	3.0	0.2	3.9	0.3	<0.001
Diagnostic confidence
R1	3.6	0.7	4.6	0.6	<0.001	3.0	1.1	4.4	0.7	<0.001	3.2	0.6	4.0	0.6	<0.001
R2	3.7	0.6	4.7	0.5	<0.001	3.1	1.2	4.1	1.0	0.002	3.5	0.6	4.3	0.7	<0.001

Values are given as means based on the ratings of both readers (R1 and R2) and their SDs. Interrater agreement analyzed by Cohen’s kappa (see Table 4). Evaluation of all five items was significantly improved by the Smart MAR algorithm for both readers.A P value <0.05 was considered to indicate a statistically significant difference.

Table 4.

Results of interrater agreement analyzed by means of Cohen’s kappa (κ) analysis.

Inter-observer agreement Cohen’s Kappa
	Hip		Shoulder		Spine
	κ	P	κ	P	κ	P
Interrater artifact grading	0.81	<0.001	0.65	<0.001	1.00
Noise standard	0.36	0.01	0.16	0.435	1.00
Noise Smart MAR	0.81	<0.001	0.47	0.008	0.59	0.002
Adjacent trabecular bone standard	0.66	<0.001	0.75	<0.001	0.64	<0.001
Adjacent trabecular bone Smart MAR	0.77	<0.001	0.83	<0.001	0.49	0.001
Cortical bone standard	0.56	<0.001	0.88	<0.001	0.74	<0.001
Cortical bone Smart MAR	0.63	<0.001	0.83	<0.001	0.59	<0.001
Adjacent soft tissue standard	0.75	<0.001	0.81	<0.001	0.54	<0.001
Adjacent soft tissue Smart MAR	0.71	<0.001	0.74	0.005	0.34	0.005
Artifacts standard	0.65	<0.001	0.81	<0.001	0.65	<0.001
Artifacts Smart MAR	0.76	<0.001	0.73	<0.001	0.33	0.005
Diagnostic confidence standard	0.75	<0.001	0.79	0.002	0.47	0.002
Diagnostic confidence Smart MAR	0.70	<0.001	0.53	<0.001	0.52	<0.001

Coefficients were interpreted as follows: κ < 0.20 = slight agreement; κ 0.21–0.40 = fair agreement; κ 0.41–0.60 = moderate agreement; κ: 0.61–0.80 = substantial agreement; κ 0.81–1.0 = almost perfect agreement.A P-value <0.05 was considered to indicate statistical significance.

There was substantial interrater agreement (Table 4) regarding the distribution and volume of artifacts in all three regions (Cohen’s kappa, κ = 0.65–1.00, P < 0.001).

For all six parameters assessed (noise, delineation of trabecular/cortical bone, soft tissue contrast, extent of artifacts and the summarizing parameter “overall diagnostic confidence”), the Smart MAR reconstructions received significantly higher ratings (all P < 0.005) except from evaluation of noise in the shoulder region with P values of 0.038 and 0.046, respectively (Tables 2 and 3).

Except for the evaluation of noise (κ = 0.16–0.59), soft tissue (κ = 0.34, P = 0.005), and artifact reduction in the spine series after Smart MAR reconstruction (κ = 0.33, P = 0.005), interrater agreement was moderate to almost excellent (κ = 0.65–1.00, all P values < 0.005; Table 4).

New artifacts

Nearly all complementary Smart MAR reconstructions show possible or unambiguous metal breaks or loosening signs in comparison to standard ASIR V series as reference (Table 5 and Fig. 4). These potential breaks or loosening was in no case confirmed to be relevant in the comparative clinical and radiological follow-up assessments the following year.

Table 5.

Results of subjective, semi-quantitative evaluation of periprosthetic hem lines and (artificial) fractures in Smart MAR series in comparison to standard ASIR V series as reference.

	Rater 1	Rater 2	Rater 1	Rater 2	Rater 1	Rater 2
Hip (n = 23)
Prosthetic screws/ implant breaks	No break	No break	Probable	Probable	Certain break	Certain break
Prosthetic screws/ implant breaks	8 (35)	10 (44)	9 (39)	5 (22)	6 (26)	8 (35)
Artificial periprosthetic loosening hem	No	No	1–2 mm	1–2 mm	>2 mm	>2 mm
Artificial periprosthetic loosening hem	9 (39)	11 (48)	9 (39)	7 (30)	5 (22)	5 (22)
Spine (n = 26)
Prosthetic screws/implant breaks	No break	No break	Probable	Probable	Certain break	Certain break
	11 (24)	13 (50)	3 (12)	0	12 (46)	13 (50)
Artificial periprosthetic loosening hem	No	No	1–2 mm	1–2 mm	>2 mm	>2 mm
Artificial periprosthetic loosening hem	5 (19)	4 (15)	19 (73)	19 (73)	2 (8)	3 (12)
Shoulder (n = 23)
Prosthetic screws/implant breaks	No break	No break	probable	probable	Certain break	Certain break
	16 (70)	16 (70)	3 (13)	2 (9)	4 (17)	5 (22)
Artificial periprosthetic loosening hem	No	No	1–2 mm	1–2 mm	>2 mm	>2 mm
Artificial periprosthetic loosening hem	1 (4)	0	13 (57)	12 (52)	9 (39)	11 (48)

Values are given as n (%).

Fig. 4.

Case report (SL, 2.5 mm; non CE; bone and soft-tissue kernel). A 63-year-old man after femoral neck fracture and total endoprosthesis, presenting with reduced adduction of left thigh after surgery. Arrow indicating small lipoma/seroma only visible in reconstructed series. Both background noise is reduced and delineation/visualization of periprosthetic tissues is improved.

Especially for spine scans, a large number of newly introduced artifacts occured compared to standard ASIR V scans (R1: 12 [46%], R2: 13 [50%]). In contrast, new, additionally inserted artifacts were comparatively rare in shoulder scans (n = 6 [26%]). Periprosthetic lucency especially occurred in reconstructed shoulder scans with up to 48% (n = 11). Probable or certain periprosthetic loosening was most frequently introduced to spine implants by Smart MAR reconstruction compared to ASIR V (1–2 mm in up to 73%, > 2 mm in 8%/12%).

Discussion

Objective, ROI-based measurements, as parameters for reduction of both image noise and artifact-caused heterogeneity, showed a reduced SD and partially reduced attenuation compared to ASIR V images alone. The missing effect of Smart MAR on SNR is probably related to a counterbalancing effect of beam hardening and streak artifacts on mean HU in a particular ROI. We attribute the superior SNR in the standard reconstructions of the spine region to the overall pronounced artifacts in this region and the small sample size. Using similar quantitative variables, comparable projection-based O-MAR algorithms (Philips) were successful in reducing mean attenuation and SD resulting in improved CT image quality by reducing noise levels in the pelvic region or streak artifacts and background noise for CT of the hip or spinal prostheses (18,21,22).

For the evaluation of artifacts, however, these objective parameters should be interpreted with caution because lower median attenuation or SD in ROIs as well as higher SNR are also consistent with heterogeneous artifacts (e.g. white and black streaks) adjacent to implants. For this reason, our objective results may simply indicate a different distribution of artifacts rather than improved image quality.

The relatively high proportion of screws in the hip and shoulder region (48% and 39%, respectively) in our study group in relation to large implants might be another reason for only moderate improvement of objective parameters; hence, other MAR algorithms were more successful in large implants (3,4,11,23).

The scores of all six parameters assessed subjectively, including the overall parameter “diagnostic confidence,” were significantly improved in all three anatomical regions after Smart MAR reconstruction instead of using ASIR V reconstructions alone (most P < 0.005, Figs. 2 and 3).

Using a similar, 5-point rating scale for evaluation of projection based SEMAR compared to iterative reconstructions alone, recent studies found higher diagnostic confidence (scores 3.3–3.4 and 4.7 for IR-only and SEMAR, respectively) and improved detection of smaller anatomical structures (11). Similar reconstruction algorithms such as O-MAR showed success in reducing metal artifacts for radiation therapy planning due to improved overall quality score (1.35 vs. 3.25, P = 0.002) or for shoulder regions (10,18,21).

Comparison with more numerous results obtained with MAR algorithms for DECT is difficult due to the fundamentally different technologies underlying artifact reduction. Nevertheless, a recent study showed the advantage of a combined use of virtual monochromatic imaging and GE MARS, resulting in a reduction of periprosthetic artifacts and achieving attenuation levels comparable to implant-free tissue (24). For GE Smart MAR algorithms applied on single source and DECT, improved visualization of periprosthetic tissues was described by recent studies, whereas image quality for use with DECT scans varied with prosthesis composition (12,14,25). Drawbacks of DE-based algorithms regarding post-processing and workflow are well described (4,26,27). For clinical routine, the simple and time and dose saving acquisition of automatically and additionally generated, post-processed Smart MAR reconstructions without previous consideration of prosthesis material may represent an advantage in addition to the comparably low radiation exposure (11).

Besides the positive results, it was found that Smart MAR introduced artifacts compared to standard ASIR V (Table 5 and Fig. 4) as it was recently shown in the evaluation of spine implants (25). For Smart MAR, both raters described newly introduced changes in integrity and size with partially increasing periprosthetic lucency in all anatomical regions, especially for the spine and shoulder regions. For similar projection-based O-MAR technology, distortion of size in hip implants or introduction of several new artifacts as a “penumbra region” still hampers quality of MAR algorithms (18,28). According to recent studies, the occurrence of additional artifacts is very common in projection-based MARS algorithms, especially in inhomogeneous environments with massive differences of CT density between metallic implants, periprosthetic soft tissues, and extracorporeal air (9,23). The high amount of Smart MAR-related artifacts in the spine region (new artifacts in 47%) could be at least partly related to this fact (location of thoracic/cervical spine near lung tissue or extracorporeal air).

Regarding the evaluation of implant size and integrity, Smart MAR should be used very cautiously and a comparison to standard series is still recommended.

Other limitations of the present study should be mentioned. First, CT images reconstructed using Smart MAR have a different visual appearance, which precludes full blinding and might have biased subjective rating. Second, the fact that impact of different alloys and the respective types of implanted materials (e.g. screw, partial or full prosthesis) were not considered further in the evaluation might have biased our results. Third, the number of patients was limited. Furthermore, the mean age of the study population was high, which might be associated with decreased bone mineralization and biased evaluation of bone structures.

A comparison of the Smart MAR tool with other promising single-source or DECT MAR technologies, especially regarding the introduction of new artifacts, might be an interesting topic for further prospective studies.

In conclusion, Smart MAR improved image quality of hip, spine, and shoulder CT scans resulting in higher diagnostic confidence in the evaluation of periprosthetic soft tissues. As shown for spine implants, it should be used with caution and as a complementary tool for evaluation of periprosthetic loosening or integrity of metal implant, as in many cases it introduced new artifacts.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Felix W Feldhaus

Georg Böning

Uli Fehrenbach

References

Roth

Maertz

Parr

, et al. CT of the hip prosthesis: appearance of components, fixation, and complications. RadioGraphics 2012; 32:1089–1107.

Douglas-Akinwande

Buckwalter

Rydberg

, et al. Multichannel CT: Evaluating the spine in postoperative patients with orthopedic hardware. RadioGraphics 2006; 26:S97–S110.

Axente

Paidi

Von Eyben

, et al. Clinical evaluation of the iterative metal artifact reduction algorithm for CT simulation in radiotherapy. Med Phys 2015; 42:1170–1183.

Morsbach

Bickelhaupt

Wanner

, et al. Reduction of metal artifacts from hip prostheses on CT images of the pelvis: value of iterative reconstructions. Radiology 2013; 268:237–244.

Huang

Kerns

Nute

, et al. An evaluation of three commercially available metal artifact reduction methods for CT imaging. Phys Med Biol 2015; 60:1047–1067.

Nakamoto

Kim

Hori

, et al. Clinical evaluation of image quality and radiation dose reduction in upper abdominal computed tomography using model-based iterative reconstruction; comparison with filtered back projection and adaptive statistical iterative reconstruction. Eur J Radiol 2015; 84:1715–1723.

Zhu

Zhao

, et al. Feasibility study of using gemstone spectral imaging (GSI) and adaptive statistical iterative reconstruction (ASIR) for reducing radiation and iodine contrast dose in abdominal CT patients with high BMI values. PLoS One 2015; 10:e0129201.

Lee

M-J

Kim

Lee

S-A

, et al. Overcoming artifacts from metallic orthopedic implants at high-field-strength MR imaging and multi-detector CT. RadioGraphics 2007; 27:791–803.

Boas

Fleischmann

Evaluation of two iterative techniques for reducing metal artifacts in computed tomography. Radiology 2011; 259:894–902.

10.

Subhas

Polster

Obuchowski

, et al. Imaging of arthroplasties: Improved image quality and lesion detection with iterative metal artifact reduction, a new CT metal artifact reduction technique. AJR Am J Roentgenol 2016; 207:378–385.

11.

Gondim Teixeira

Meyer

Baumann

, et al. Total hip prosthesis CT with single-energy projection-based metallic artifact reduction: Impact on the visualization of specific periprosthetic soft tissue structures. Skeletal Radiol 2014; 43:1237–1246.

12.

Feldhaus F, Böning G, Jonczyk M, et al. Metallic dental artifact reduction in computed tomography (Smart MAR): improvement of image quality and diagnostic confidence in patients with suspected head and neck pathology and oral implants. Eur J Radiol. Epub ahead of print 2019. DOI: 10.1016/j.ejrad.2019.07.015.

13.

Weiß

Schabel

Bongers

, et al. Impact of iterative metal artifact reduction on diagnostic image quality in patients with dental hardware. Acta Radiol 2017; 58:279–285.

14.

Lee

Park

Song

, et al. Metal artefact reduction in gemstone spectral imaging dual-energy CT with and without metal artefact reduction software. Eur Radiol 2012; 22:1331–1340.

15.

Schäfer ML, Lüdemann L, Böning G, et al. Radiation dose reduction in CT with adaptive statistical iterative reconstruction (ASIR) for patients with bronchial carcinoma and intrapulmonary metastases. Clin Radiol 2016;71:442–449.

16.

Kaul D, Kahn J, Huizing L, et al. Dose reduction in paediatric cranial CT via iterative reconstruction: a clinical study in 78 patients. Clin Radiol 2016;71:1168–1177.

17.

GE Healthcare. Smart MAR White Paper. Waukesha, WI: GE Healthcare, 2013. Available at: https://www.gehealthcare.co.uk/-/jssmedia/3b5c4c1c695e45a9b41309ae3f4ce22b.pdf?rev=3d026220228343d68f887c18005bf0b9.

18.

Shim

Kang

Ahn

, et al. Metal artifact reduction for orthopedic implants (O-MAR): Usefulness in CT evaluation of reverse total shoulder arthroplasty. AJR Am J Roentgenol 2017; 209:860–866.

19.

Cohen

A coefficient of agreement for nominal scales. Educ Psychol Meas 1960; 20:37–46.

20.

Bland

Altman

DG.

Multiple significance tests: the Bonferroni method. BMJ 1995; 310:170.

21.

Noel

Chen

, et al. Clinical evaluation of a commercial orthopedic metal artifact reduction tool for CT simulations in radiation therapy. Med Phys 2012; 39:7507–7517.

22.

Jeong

Kim

Hwang

, et al. Usefulness of a metal artifact reduction algorithm for orthopedic implants in abdominal CT: phantom and clinical study results. AJR Am J Roentgenol 2015; 204:307–317.

23.

Brook

Gourtsoyianni

Brook

, et al. Spectral CT with metal artifacts reduction software for improvement of tumor visibility in the vicinity of gold fiducial markers. Radiology 2012; 263:696–705.

24.

Reynoso

Capunay

Rasumoff

, et al. Periprosthetic artifact reduction using virtual monochromatic imaging derived from gemstone dual-energy computed tomography and dedicated software. J Comput Assist Tomogr 2016; 40:649–657.

25.

Wayer

Kim

Otto

, et al. Unintended consequences: review of new artifacts introduced by iterative reconstruction CT metal artifact reduction in spine imaging. AJNR Am J Neuroradiol 2019; 40:1973–1975.

26.

Wellenberg

RHH

Boomsma

van Osch

JAC

, et al. Quantifying metal artefact reduction using virtual monochromatic dual-layer detector spectral CT imaging in unilateral and bilateral total hip prostheses. Eur J Radiol 2017; 88:61–70.

27.

Srinivasan

Hoeffner

Ibrahim

, et al. Utility of dual-energy CT virtual kev monochromatic series for the assessment of spinal transpedicular hardware-bone interface. AJR Am J Roentgenol 2013; 201:878–883.

28.

Wang

Qian

, et al. Metal artifacts reduction using monochromatic images from spectral CT: Evaluation of pedicle screws in patients with scoliosis. Eur J Radiol 2013; 82:e360–366.