Role of metal artifact reduction software in computed tomography angiography of lower limb metallic prosthesis: A retrospective study

Abstract

BACKGROUND

Metal artifacts significantly limit the diagnostic quality of computed tomography (CT) algorithm in patients with lower limb metallic prostheses. Single-energy metal artifact reduction (SEMAR) algorithms have been developed to address this limitation.

AIM

To evaluate the effects of SEMAR algorithm on image quality characteristics of lower limb CT algorithm (CTA) (artifact reduction, signal-to-noise ratio [SNR] and inter-observer agreement) in patients with metallic prostheses.

METHODS

Twenty-two patients (14 males, 8 females; median age 54 years) with unilateral or bilateral metal lower limb prostheses or screws underwent lower limb CTA on a 128-slice multi-detector CT scanner. Images were reconstructed with and without SEMAR algorithm. Three radiologists independently assessed subjective image quality using standardized scoring systems. Quantitative analysis included measurement of SNR and contrast-to-noise ratio in arterial and soft tissue regions of interest.

RESULTS

Application of SEMAR significantly improved SNR in two of three observers (P = 0.009 and P = 0.004), with values approaching those in the contralateral reference limb. Overall artifact reduction was statistically significant (median difference 1.5 Hounsfield unit, P < 0.001), as was improvement in overall image quality (median difference 1 Hounsfield unit, P < 0.001). Good inter-observer agreement (intraclass correlation coefficient > 0.75) was demonstrated for SNR and contrast-to-noise ratio assessment with metal artifact reduction (MAR) and reference images, while poor agreement (intraclass correlation coefficient < 0.5) was noted for non-MAR images. Kendall’s W demonstrated significant concordance among observers (W = 0.899, P < 0.001 for study quality with MAR).

CONCLUSION

SEMAR algorithm significantly reduces metal artifacts from lower limb fixation prostheses without compromising vessel contrast, improving visualization of periprosthetic vascular structures and enhancing diagnostic capability of CTA examinations.

Key Words: Metal artifact reduction; Computed tomography angiography; Lower limb prosthesis; Image quality; Single energy metal artifact reduction

Core Tip: Single-energy metal artifact reduction (SEMAR) is a computed tomography (CT) algorithm designed to reduce metal-induced artifacts while preserving vessel contrast. In this retrospective single-center study, lower limb CT angiography examinations in 22 patients with metallic fixation prostheses were reconstructed with and without SEMAR to assess arterial signal-to-noise ratio, contrast-to-noise ratio, artifact burden, and diagnostic confidence. SEMAR significantly reduced metal artifacts, improved signal-to-noise ratio and overall image quality, and restored periprosthetic arterial visualization to levels comparable with the contralateral reference limb, without compromising vascular contrast, thereby enhancing the diagnostic performance of lower limb CT angiography in this challenging patient population.

INTRODUCTION

Although conventional digital subtraction angiography (DSA) technique is the gold standard examination for evaluating blood vessels, it has disadvantage of being an invasive procedure[1,2]. Magnetic resonance imaging (MRI) and computed tomography (CT) represent less invasive imaging modalities for vascular assessment compared to DSA. However, the presence of metallic prosthetic material presents substantial imaging challenges. In MRI, metallic hardware causes magnetic field inhomogeneity, resulting in image artifact and signal distortion[3,4]. MRI feasibility is further restricted by artifacts from patient motion and contraindications related to non-MRI-compatible metallic objects[5]. Recently, magnetic resonance angiography has become a viable option for non-invasive postembolization surveillance due to its high flow sensitivity[6,7]. However, visualization with magnetic resonance angiography may be influenced by susceptibility artifacts, motion artifacts due to the required long acquisition time, or T1 shortening of surrounding residual subacute hematomas[6,8].

In CT angiography (CTA) studies, the combination of photon starvation and beam hardening effects results in dark and bright streaks across the images[9], which reduce the whole image quality of the adjacent tissues and the metallic material itself[10]. Anatomical structures close to the metallic implant are often obscured by these dark and bright streaks, leading to increase in the risk of missing important data and markedly limiting the diagnostic value such examinations[11,12]. Because metal artifacts markedly limit the diagnostic value of imaging studies as CTA after clipping or coiling, post-prostheses insertion and also in other radiology fields[13], the development of algorithms for the metal artifact reduction (MAR) was an important step forward in providing images of good diagnostic value in patients with any metallic implants[14,15].

X-ray attenuation and resulting physical effects vary according to the material composition and dimensions of metallic implants[16]. Materials with lower atomic numbers, such as those used in surgical clips, generate only mild beam hardening effects. With higher atomic number metallic hardware, as internal fixation prostheses, artifacts are more likely to occur due to photon starvation as well as other factors[9].

Dual-energy CT (DECT) techniques generate virtual monochromatic images at varied photon energies to reduce beam hardening[17,18], yet DECT inadequately removes severe streaking artifacts from high atomic number metals such as platinum coils due to photon starvation effects[19-21]. Additionally, increased keV settings diminish iodine contrast visualization, complicating clinical diagnosis, particularly for vascular assessment in patients with metallic fixation devices[20]. The single-energy MAR (SEMAR) algorithm was developed specifically to mitigate photon starvation-related artifacts[22]. By targeting metal-identified pixels through Hounsfield unit (HU) thresholding while preserving non-metallic and contrast-enhanced structures, SEMAR provides selective artifact reduction without compromising diagnostic information[23].

Many MAR tools have been improved for use in clinical practice, as: DECT and MAR algorithms[24]. DECT obtains data at two energy settings (80 kVp and 135 kVp or 140 kVp). The monochromatic reconstructions at such higher energy levels reduce the beam hardening effects and artifacts by eliminating lower energy levels data. Additional MAR algorithms may be used to further decrease metal artifacts. DECT-imaging has disadvantages as it might complicate the acquisition protocol and may increase radiation dose significantly[25]. Algorithm of SEMAR is a CT MAR algorithm that reduces metal artifacts in either single-energy or DECT images[22,26]. As conventional CT scanners acquire images at settings of single-energy, no adjustments of the acquisition data are required and there is no increase in the radiation dose.

SEMAR suppresses photon starvation artifacts by identifying metallic components through “HU threshold” segmentation method, localizing corrupted projection data corresponding to these metals, and replacing the degraded data with corrected values[23].

MATERIALS AND METHODS

Study objective

The objective of this study was to comprehensively evaluate the effects of SEMAR algorithm on quantitative and qualitative image quality characteristics of lower limb CTA in patients with metallic prostheses or internal fixation hardware used for bone fixation.

Study design and ethical approval

This was a single-center, retrospective study approved by Medical Research Ethics Committee of Mansoura Faculty of Medicine Mansoura University (Approval No. R.22.11.1945). Patient informed consent was waived due to the retrospective design. All patients were over 18 years of age at the time of imaging.

Patient selection and inclusion criteria

Twenty-two patients with unilateral or bilateral metal lower limb prostheses or screws for internal fixation at thigh or leg who underwent lower limb CTA were included in this study. Inclusion criteria were as follows: (1) Age more than 18 years; (2) Ability to undergo CT in the supine position; (3) Ability to remain still for the duration of CT acquisition; and (4) The absence of other foreign materials located close to the prostheses possibly affecting the artifact analysis.

CT data acquisition protocol

All studies were performed with a MDCT scanner (GE Revolution EVO 128; GE Healthcare, IL, United States). For CTA, contrast medium (iopamidol, iopamiron 370 mg/mL, 1.5-2 mL/kg of body weight) was injected via the antecubital vein using a power injector with fixed injection duration of 30 seconds (at a flow rate of 5 mL/second followed by a 30-mL saline chaser using the same flow rate). A bolus-tracking technique was used, and images were obtained after the attenuation measurement reached 100 HU in the abdominal aorta. The CT parameters were as follows: Tube voltage, 80 kVp; tube current, 150 mA to 350 mA for CTA; gantry rotation time, 1.0 second; detector collimation, 320 mm × 0.5 mm; field of view, 240 mm; and matrix, 512 × 512. Radiation dose was recorded as volume CT dose index and dose-length product. Effective dose was estimated using k-factor 0.015 for lower extremity imaging. SEMAR processing occurred only during reconstruction without acquisition parameter modification, incurring no additional radiation exposure.

Imaging data were sent to a workstation (GE Advantage Workstation server; GE Healthcare, IL, United States) and images were reconstructed with and without the use of the MAR algorithm with 0.5-mm thick sections in 0.25-mm increments using the adaptive iterative reconstruction. MAR software (MARS) using CT machine data was obtained by using native MAR protocols in the console. After single scan data acquisition, two series of images were generated, with and without MAR algorithms. In first stage of data processing, corrupted samples in the projection that correspond to metallic objects are identified. Stage two involved generation of in-painted data by replacing the metal corrupted projections with the corrected data. The corrected data is generated using the forward projection of the classified image. Final Stage corrected projection is generated using a combination of the original projection data and the in-painted projection, revealing anatomic details hidden beneath the artifacts. Finally, image sets were sent to a picture archiving and communication system (PACS) system for image review. Subjective image analysis and quantitative image analysis were performed with PACS software by three radiologists (one with 9+ years of experience and the other two with 6+ years of experience in radiology).

Subjective image quality assessment

Three observers independently reviewed the two image sets at random ordering and at 2 weeks interval to avoid recall bias. Images were evaluated in the same window settings (window level, 60 HU; window width, 340 HU). The effect of metallic artifacts on the visualization of periprosthetic vascular structures was graded as follows: 1 = no vascular structure obscured; 2 = mild artifacts; 3 = moderate artifacts obscuring vascular anatomy; 4 = sever artifacts near totally obscuring vessels; 5 = vessels completely obscured by artifacts. Overall study diagnostic confidence was graded as follows: 1 = excellent diagnostic confidence; 2 = medium diagnostic confidence; 3 = poor confidence/images non diagnostic.

Arteries were evaluated on CTA images for visibility of the artery lumen, sharpness of artery margins, and the contrast degree between the artery and surrounding tissues. Image quality characteristics assessed in this study have been described in the “European Guidelines on Quality Criteria for Computerized Tomography”[27] and have been used in a similar manner in multiple previous radiology studies[28,29].

Quantitative image quality assessment

Levels of maximum vascular-related artifacts were determined through consensus among the three observers, with subsequent independent quantitative analysis by each reader using PACS workstation software. Arterial regions of interests (ROIs) were drawn as single circle positioned within the center of related major artery at the ipsilateral and contra-lateral sides of metallic prosthesis. Mean density and SD attenuation values in HUs were measured. Second ROI measurements were done in subcutaneous fat in the same direction of the metallic artifact, as close as possible to skin surface without including adjacent organs or structures. All arterial and soft tissue ROIs areas were between 0.02-0.03 cm². Signal-to-noise ratio (SNR) was calculated using formula (SNR = average pixel values in signal ROI/SD background ROI) and contrast-to-noise ratio (CNR) was calculated using formula (CNR = average signal ROI-average background ROI/SD background ROI). Three independent measurements were performed by each observer to assess intra-observer variability.

Statistical analysis

Data were entered and analyzed using both IBM-SPSS software (IBM Corp. Released 2019. IBM SPSS Statistics for Windows, Version 26.0. Armonk, NY, United States) and MedCalc^® Statistical Software version 20 (MedCalc Software Ltd, Ostend, Belgium). Subjective data were expressed as n (%). Quantitative data were initially tested for normality using Shapiro-Wilk’s test with data being normally distributed if P > 0.050. Presence of significant outliers was tested for by inspecting boxplots. Quantitative data were expressed as median (25^th percentile, 75^th percentile) or median (minimum-maximum) as data was not normally distributed. Friedman’s test was used to compare three related non-normally distributed quantitative data. The intraclass correlation coefficient (ICC) was used to determine intra- and inter-rater reliability (absolute agreement) in assessment of a continuous data. Based on the 95% confident interval of the ICC estimate, values less than 0.5, between 0.5 and 0.75, between 0.75 and 0.9, and greater than 0.90 are indicative of poor, moderate, good, and excellent reliability, respectively[30]. Kendall’s coefficient of concordance, also known as Kendall’s W, was used as a measure of agreement in assessing inter-rater reliability between multiple raters. For any of the used tests, results were considered as statistically significant if P ≤ 0.050. Appropriate charts were used to graphically present the results whenever needed.

RESULTS

Patient demographics and prosthesis characteristics

The study population consisted of 22 patients [14 males (63.6%), 8 females (36.4%)] with a median age of 54 years (range 19-90 years). Different types and sites of metallic fixations as well as clinical vascular Pathologies were noted as shown in the following Tables 1, 2, and 3. Regarding prothesis material, Titanium alloy was used in 12 patients (54.5%), stainless steel was used in 6 patients (36.4%), and for the remaining 4 patients (18.2%) material documentation was unavailable or lost. mean radiation dose parameters were as follow; volume computed tomography dose index: 6.1 ± 1.4 mGy (range 3.6-7.2 mGy), dose-length product: 1282 ± 263 mGy-cm (range 1016-1643 mGy-cm). Mean estimated effective dose: 19.2 ± 3.9 mSv (range 15-25 mSv).

Table 1 Distribution of fixation device types.

Fixation type	n	%
DHS	6	27.3
External fixation	3	13.6
Medullary nail	5	22.7
Plate and screws	6	27.3
Screws	1	4.5
THR	1	4.5

DHS: Dynamic hip screw; THR: Total hip replacement.

Table 2 Vascular pathology in study population.

Pathology	n	%
Atherosclerosis	9	40.9
Normal	10	45.5
Tumor compression	3	13.6

Table 3 Location and laterality of prostheses.

Parameter	n	%
Location
Femur	17	77.3
Tibia	5	22.7
Side
Right	11	50
Left	11	50

Quantitative image quality analysis

Friedman’s test demonstrated statistically significant differences in SNR across the three imaging conditions (with MAR, without MAR, and reference limb) in two of three observers (P = 0.009 for observer 2; P = 0.004 for observer 3). Observer 1 demonstrated a trend toward significance (P = 0.057).

Notably, SNR values obtained with MAR application closely approximated those from the contralateral reference limb in all observers. In contrast, SNR values from images without MAR were substantially reduced compared to both MAR-reconstructed images and reference limb measurements (Table 4).

Table 4 Signal-to-noise ratio with and without metal artifact reduction, median (25^th percentile, 75^th percentile).

Observer	With MAR	Without MAR	Reference	P value
One	17.3 (8.5-30.2)	8.03 (1.99-26)	17.6 (6.65-42.5)	0.057
Two	12.98 (7.68-41.8)	5.74 (2.88-23.6)a	19.1 (8.2-36.5)	0.009
Three	16.5 (5.5-40.6)	6.5 (3.0-15.8)a	16.8 (7.1-43.0)	0.004

^aP < 0.05 signal-to-noise ratio without metal artifact reduction vs signal-to-noise ratio with metal artifact reduction.

Test of significance is Friedman’s test. MAR: Metal artifact reduction.

Inter-observer reliability for quantitative measurements

Inter-observer reliability demonstrated important differences based on reconstruction method. Good absolute agreement (ICC 0.75-0.9) was achieved among all three observers for SNR and CNR assessment when evaluating MAR-reconstructed images and reference limb images (Table 5). In marked contrast, poor inter-observer agreement (ICC < 0.5) was observed for SNR and CNR measurements from non-MAR reconstructed images.

Table 5 Intraclass correlation coefficient for inter-observer agreement.

Measurement	ICC	Absolute agreement	95%CI for ICC	P value
SNR
With MAR	0.751	Good	0.570-0.877	< 0.001
Without MAR	0.460	Poor	0.205-0.697	< 0.001
Reference	0.790	Good	0.628-0.898	< 0.001
CNR
With MAR	0.764	Good	0.488-0.809	< 0.001
Without MAR	0.487	Poor	0.230-0.717	< 0.001
Reference	0.810	Good	0.684-0.916	< 0.001

SNR: Signal-to-noise ratio; MAR: Metal artifact reduction; CNR: Contrast-to-noise ratio; ICC: Intraclass correlation coefficient; CI: Confidence interval.

Subjective image quality assessment

On subjective evaluation of the studies, there was a statistically significant difference in overall artifact and image quality between the studies with MAR vs those without MAR, indicating a lower value of artifact and higher image quality when using MAR, as shown in Table 6. Kendall’s W was run to determine if there was agreement between three radiologists’ judgment on the overall quality of studies and the overall artifact. All three radiologists demonstrated highly statistically significant concordance (P < 0.001) in their subjective assessments of both study quality and artifact burden, indicating robust inter-observer agreement regardless of reconstruction method (Table 7).

Table 6 Overall artifact burden and image quality with vs without metal artifact reduction, median (minimum-maximum).

Characteristic	With MARS	Without MARS	MD1 (95%CI)	P value
Overall artifact	2 (2-5)	4 (3-5)	1.5 (1-2)	< 0.001
Overall image quality	1 (1-3)	2.5 (1-3)	1 (0.5-1)	< 0.001

¹MD = Hodges-Lehmann median difference.

Test of significance is Wilcoxon’s signed ranks test. MARS: Metal artifact reduction software; CI: Confidence interval.

Table 7 Inter-observer agreement of subjective assessment.

Image type	Study quality assessment (W)	Study quality (P value)	Artifact assessment (W)	P value
MAR-reconstructed	0.899	< 0.001	0.918	< 0.001
Non-MAR	0.847	< 0.001	0.922	< 0.001

Kendall’s Coefficient of Concordance (W). MAR: Metal artifact reduction.

Wilcoxon signed-ranks testing revealed highly statistically significant improvements with MAR application. Overall artifact scores were substantially reduced with MAR (median 2 vs 4 without MAR, P < 0.001), and overall image quality was significantly improved (median 1 vs 2.5 without MAR, P < 0.001).

Case presentations

Case 1: 42-year-old male with left femoral fracture treated with medullary nail fixation. CTA assessment to evaluate lower limb vasculature was requested as pre-operative workup for knee arthroplasty (Figure 1).

Figure 1 Improved femoral artery visualization after metal artifact reduction software processing. A: Pre-metal artifact reduction (MAR), lower limbs axial computed tomography angiography without MAR software demonstrates streak and bloom artifacts originating from left femoral intra-medullary nail. Three regions of interest (ROI) were placed; ROI 1 in the contralateral (right) femoral artery shows mean Hounsfield unit (HU) of 553.75, SD = 2.062, ROI 2 in the ipsilateral (left) femoral artery demonstrates suboptimal contrast opacification, mean HU of 499.0, SD = 7.072. ROI 3 was placed in left subcutaneous fat with mean HU of -76, SD = 7.072; B: Post-MAR, same anatomic level post MAR software processing demonstrates marked artifact reduction with improved visualization of femoral artery. ROI 1 in the contralateral (right) femoral artery shows mean HU of 580.0, SD = 6.481, ROI 2 in the ipsilateral (left) femoral artery demonstrates suboptimal contrast opacification, mean HU of 562.5.0, SD = 7.072. ROI 3 was placed in left subcutaneous fat with mean HU of -95.3, SD = 8.387. ROI: Regions of interest.

Case 2: 58-year-old male with history of right femoral neck fracture treated with open reduction and internal fixation using dynamic hip screw. Patient presented with claudication symptoms and underwent lower limb CTA to assess for atherosclerotic disease affecting the right lower extremity (Figure 2).

Figure 2  Metal artifact reduction in iliofemoral computed tomography angiography: A comparative visualization. A and B: Pre-metal artifact reduction (MAR), coronal reformatted computed tomography angiography maximum intensity projection and volume rendering images at ilio-femoral level without MAR showing artifact-induced degradation of image quality and pseudo-stenoses at right superficial femoral and profunda femoris arteries; C and D: Post-MAR, same reformats with single-energy MAR reconstruction shows significant artifact reduction and clear visualization of femoral arteries.

DISCUSSION

This study analyzed the influence of the SEMAR algorithm on various image quality metrics in lower limb CTA studies performed to assess vascular patency in patients with prostheses for fixation. In patients with metallic prostheses, metal artifacts frequently cause pseudo-stenoses caused by artifacts rather than true pathology. Optimization of image quality represents a clinically meaningful intermediate endpoint in the validation pathway. There was a significant improvement of SNR with application of MARS. Where, evaluated SNR in examinations with MAR was close to that in the reference point in opposite limb. While SNR in examinations without MAR is significantly lower than the previous two examinations. Demonstration of SEMAR efficacy across different titanium alloys and stainless-steel hardware compositions indicates adequate algorithm generalizability. Additionally, there was a good absolute agreement between the three observers when assessing SNR with MAR and in opposite limb as a reference, and poor absolute agreement between the three observers in assessing SNR without MAR. In the overall agreement for all cases between the two observers, however, was about 80.7%, indicating a substantial agreement[31].

Also, there is agreement with results of Shinohara et al[32], who concluded that there is significant degree of MAR and better blood vessel visualization around the platinum coil using dual energy technique and MARS in CTA studies after cerebral aneurysm coil embolization, they found that another important factor for beam hardening artifact reduction is the “CT window setting”. Wide settings of window width on CTA are important for reduction of the beam hardening artifact from a metallic stents, surgical clips and platinum coils[33,34]. However, if the window width was too wide, the vessel contrast decreases[33,34]. For the vessel contrast, the suitable energy level for monoenergetic imaging with MARS was about < 75 keV, and wide window width setting was required as the energy level decreased. This is a point of compromise for obtaining a suitable balance between the vessel contrast and the beam hardening artifact. They suggested that “monoenergetic imaging” with MARS and adequate intravascular enhancement with a test bolus or bolus tracking technique, a high injection rate, and a high concentration of contrast medium are important for proper evaluation of intracranial arteries around the coil on dual energy CTA studies[32].

Also, we agreed with Pjontek et al[31] who concluded that when an intravenous contrast protocol is used, MARS significantly enhances the assess ability of brain parenchyma, vessels, and treated aneurysms in patients with intracranial coils or clips, also agreed with who found improved image quality with image noise reduction, less artifacts[29], and an improved visualization of stent patency on the CT images with application of SEMAR reconstruction in all target visceral arteries. As regard CNR, in study[29], they detected improved CNR, in our study, there was good agreement between the three observers in assessing CNR both with MAR.

In the study of Pan et al[35] who assessed the influence of SEMAR on brain CTA images quality in patients with intra-cranial metallic implants. They evaluated effect of SEMAR as regard image noise, depending on HU measurements and SD in ROIs surrounding the metallic implants, these values were measured on contrast enhanced CTA studies. However, quantitative assessment of the effect of SEMAR on vessel contrast on CTA was not done. Also, in study they did not perform quantitative evaluation of the effect of SEMAR on the image quality and metallic artifact reduction[31]. In our study quantitative assessment after MAR revealed improvement of both SNR and CNR.

The quantitative investigation by Katsura et al[22] evaluated SEMAR’s efficacy in reducing noise surrounding intracranial platinum coils and its effects on vessel contrast visibility in CTA protocols. Results demonstrated artifact reduction while maintaining vascular contrast enhancement.

This preservation of contrast signal is clinically significant, distinguishing SEMAR from alternative MAR techniques including DECT, which reduce iodine contrast attenuation and thereby compromise diagnostic accuracy in assessing vascular patency following endovascular procedures[9]. A subsequent study of post-coil embolization cerebral angiography employed both quantitative measurement of vessel contrast and qualitative assessment of peri-coil vascular structures and aneurysm configuration. The findings demonstrated that SEMAR significantly reduced platinum coil-related artifacts without affecting vessel contrast, while simultaneously improving SNR, anatomical structure visualization, peri-coil arterial assessment, and overall diagnostic image quality and clinical acceptability[22].

Another important factor to consider for the SEMAR in the practical settings is the relatively simple workflow. SEMAR does not require necessary adjustments to the acquisition protocol or careful planning before the study because it can be conducted retrospectively on routine single energy CT data without additional radiation dose[9]. This differs from the DECT technique, which needs a CT system capable of dual-energy data acquisition and pre-scan planning. Radiation doses for DECT can be slightly higher than those for conventional single energy CT[36,37]. Because most types of virtual monochromatic images cannot be generated once the patient is already scanned with conventional single energy CT, the decision of DECT application usually needs to be made before the image acquisition. SEMAR has an advantage that it does not requires complicated CT acquisition protocols or extra radiation dose[23].

Published literature has documented technical challenges associated with SEMAR algorithm[9,19]. Metal segmentation accuracy is compromised by substantial artifacts present in uncorrected baseline images, and information loss during the correction phase, resulting from exclusion of corrupted projection data which cannot be fully compensated through interpolation. Processing errors occurring in both segmentation and interpolation phases may produce new artifacts in SEMAR-corrected images: Localized metal density increases, high-density streaking adjacent to metallic implants, and low-density nodules between metal components[9,19]. Prior investigations recommend side-by-side evaluation of pre-SEMAR and post-SEMAR reconstructions to reduce interpretive error[9,19].

Our observations are consistent with previous investigations demonstrating that MAR algorithms improve image quality[29], both subjectively and objectively, in non-enhanced CT examinations following angiographic or surgical aneurysm management. Prior research did, however, reveal a significant limitation: Suboptimal visualization of fine vascular structures, particularly intracranial vessels, with approximately 30% of patients demonstrating reduced perivascular contrast (pseudo-stenosis), creating potential for diagnostic error through false-positive stenosis or occlusion interpretations. MAR methodologies are known to introduce software-generated artifacts, documented particularly in spinal and head-neck imaging[38-40]. In our investigation, SEMAR-related newly-generated artifacts did not substantially compromise diagnostic acceptability; nevertheless, larger prospective evaluations are necessary to fully characterize such potential artifacts.

Similarly, our results align with prior research by Morsbach et al[41] showing that MAR algorithms enhance carotid CTA image quality without substantially compromising arterial contrast visualization. Also, in agreement with many authors who have proved the significant impact of SEMAR on MAR in CT studies of patients with implanted orthopedic prostheses[42,43], embolization coils[43-45], and different dental prostheses[26].

There were some limitations in this study, one of them, this was a retrospective pilot study design and on a relatively small patient cohort. In the future, long-term longitudinal prospective studies may be conducted. The hardware composition in our study was heterogeneous and un-evenly distributed due to modest sample size; however, this can reflect real-world clinical diversity. While individual hardware categories and material compositions are limited in this retrospective pilot study, SEMAR demonstrated overall statistically significant artifact reduction across all prosthesis types evaluated (P < 0.01). A prospective adequately powered study stratified by hardware type and material composition is recommended to establish hardware-specific performance metrics. Also, this study was conducted at a single institution using a 128-slice GE Revolution EVO scanner with General Electric Company’s proprietary iterative reconstruction algorithms (GE Healthcare, IL, United States). Results may differ on alternative scanner platforms (Siemens, Philips, Toshiba). Cross-platform and multicenter study validation as well as comparative studies with other imaging modalities, such as conventional DSA are recommended before broad clinical recommendations.

CONCLUSION

In summary, the SEMAR algorithm represents a significant advancement, substantially reducing metal artifacts from orthopedic fixation prostheses without compromising vessel contrast. This algorithm improves the depiction of anatomical structures and arteries around the prostheses with improved SNR and CNR hence improving the diagnostic capability of CTA examinations in patients with fixation prostheses. While image quality improvements are necessary, prospective DSA correlation is required as a follow-up investigation.