Quantitative analysis of metal artifact reduction in total hip arthroplasty using virtual monochromatic imaging and orthopedic metal artifact reduction, a phantom study

Computed tomography (CT) is commonly used as imaging modality during postoperative follow-up after total hip arthroplasties (THA) [1]. CT is a relatively inexpensive, readily available modality and it is very effective in visualizing bone defects [2]. However, imaging of the pelvis in patients with large metal hip prostheses remains one of the biggest challenges in CT imaging since metallic prosthetic components lead to metal artifacts that make the assessment of surrounding bone and soft tissues difficult. Metal artifacts also hamper the detection of pathological capsular reactions of THA components [2]. Metal artifacts arise from beam hardening, scatter and photon starvation [2].

Several methods are available to reduce these metal artifacts. This includes basic steps such as increasing tube voltage and proper positioning of the patient. However, increasing tube voltage results in a higher required dose. More advanced methods include the use of model-based image reconstruction, iterative image reconstruction and metal artifact reduction (MAR) software [2, 3]. The artifact reduction ability of such software in conventional CT imaging of THA can be further increased by the addition of iterative model-based reconstruction (IMR) [1]. Another option to reduce artifacts is the use of dual-energy CT (DECT) with reconstruction of virtual monochromatic images at high energy levels. Prior work by Wellenberg and co-workers showed that the optimal monochromatic energy depends on the material of the prostheses. In general, virtual monochromatic imaging (VMI) at 130 keV was the optimal monochromatic energy level for metal artifact reduction in THA [4]. Several patient studies showed that the combination of MAR software and virtual monochromatic dual-energy CT imaging reduces metal artifacts even more effectively than one technique alone [5‐7].

However, quantification of metal artifacts in patient studies is limited by anatomical differences between patients. By using a THA phantom, scans without hip prostheses can serve as a direct reference for scans with hip prostheses. This allows systematic assessment of metal artifacts and quantitative analysis which is not possible in patients with THA. Hitherto, a thorough, quantitative analysis of metal artifact reduction by combined use of DECT and orthopedic metal artifact reduction (O-MAR) in a THA phantom has not been performed. Hence, the aim of this study was to quantitatively compare the extent to which metal artifacts in a THA phantom can be reduced by 130 keV VMI with O-MAR, 130 keV VMI without O-MAR, conventional polychromatic imaging with O-MAR and conventional imaging without O-MAR.

Image quality parameters of conventional and VMI without O-MAR were different without insertion of prostheses (Table 2). Reconstructions with O-MAR resulted in the exact same values for these parameters as O-MAR does not change CT values when no metal is present.

Conventional images of the phantom with bilateral protheses showed more streak artifacts compared to conventional images of the phantom with unilateral prosthesis. Conventional imaging with O-MAR resulted in a decrease of streak artifacts for both unilateral and bilateral images. VMI of 130 keV without O-MAR resulted in a decrease of streak artifacts, in particular for the unilateral prosthesis. However, bright streak artifacts were introduced on the 130 keV VMI. VMI of 130 keV with O-MAR reduced dark streak artifacts without introducing bright artifacts (Fig. 6).

The highest CT values, CNR and SNR were found in conventional images with O-MAR for both mildly as severely affected pellets. The difference between both mildly and severely affected pellets and reference in CT values, CNR and SNR were lowest for 130 keV images with O-MAR. Also, the lowest noise was measured on the 130 keV images with O-MAR. VMI of 130 keV without O-MAR showed similar results to 130 keV images with O-MAR in CT values, CNR, SNR and noise for mildly affected pellets. However, CT values, CNR, SNR and noise were much lower and noise was much higher on the 130 keV images without O-MAR in comparison to 130 keV images with O-MAR for severely affected pellets (Fig. 7, see Additional file 1 for data of all individual pellets).

Conventional imaging with O-MAR resulted in significant metal artifact reduction in CT values, CNR, SNR and noise for mildly affected pellets (45%, 34%, 33% and 71%, respectively) and severely affected pellets (77%, 36%, 26% and 84%, respectively) in comparison to conventional imaging without O-MAR (Table 3). VMI of 130 keV without O-MAR resulted in significant metal artifact reduction for mildly affected pellets in CT values, CNR, SNR and noise (95%, 81%, 58% and 86%, respectively) and a significant metal artifact reduction of 47% in noise in severely affected artifacts in comparison to conventional imaging without O-MAR (Table 3). VMI of 130 keV with O-MAR also resulted in significant metal artifact reduction in CT values, CNR, SNR and noise for mildly affected pellets (81%, 83%, 66% and 91%, respectively) and severely affected pellets (79%, 46%, 32% and 90%, respectively) in comparison to conventional imaging without O-MAR (Table 3).

VMI of 130 keV without O-MAR resulted in significant metal artifact reduction in CT values, CNR and SNR in mildly affected pellets (92%, 72% and 38%, respectively) in comparison to conventional imaging with O-MAR but resulted in a significant decrease in metal artifact reduction in HU, SNR and noise in severely affected pellets (− 331%, − 202% and − 223%, respectively; Table 4). It should be noted that a negative metal artifact reduction indicates aggravation of the artifacts. VMI of 130 keV with O-MAR resulted in significant metal artifact reduction in CT values, CNR, SNR and noise for mildly (67%, 74%, 48% and 68%, respectively) in comparison to conventional imaging with O-MAR (Table 4).

In this study, we quantified metal artifact reduction in a THA phantom using high energy VMI with and without O-MAR in comparison to conventional imaging with and without O-MAR. We found that 130 keV VMI with O-MAR resulted in the strongest metal artifact reduction of mild and severe metal artifacts.

A tube voltage of 140 kVp is often used to reduce metal artifacts and results in lower CT number inaccuracies, lower noise and higher SNR and CNR [11] in comparison to 120 kVp. However, increasing tube voltage may result in only a slight reduction of metal artifacts [12‐17]. Moreover, a higher tube voltage leads to a higher required dose whereas extraction of VMI and application of O-MAR are not leading to a higher required dose.

In previous THA phantom studies, it was already reported that the use of O-MAR in conventional imaging reduces metal artifacts [1, 2, 18‐21]. Our results confirmed the value of O-MAR in conventional imaging for metal artifact reduction. Although 130 keV VMI without O-MAR showed stronger metal artifact reduction for mildly affected pellets than conventional imaging with O-MAR, metal artifact reduction for severely affected pellets worsened in comparison to conventional imaging with O-MAR in our study. Besides, the metal artifact reduction in CT values of 92% when comparing 130 keV VMI without O-MAR and conventional imaging with O-MAR is probably heightened by the bright streak artifacts visible on the 130 keV VMI without O-MAR. Our finding that 130 keV VMI in only mildly affected pellets showed metal artifact reduction in comparison to conventional imaging may be explained by reduction of metal artifacts caused by beam hardening. VMI of 130 keV is less prone to beam hardening artifacts, which are dominant in mildly affected pellets, decreasing metal artifacts.

Severely affected pellets were only present in the images of the phantom with bilateral prostheses where photon starvation is dominant over beam hardening. These severe metal artifacts caused by bilateral implants were not reduced on 130 keV VMI without O-MAR, which is supported by previous studies [6, 11, 22]. On the other hand, conventional imaging with O-MAR did result in a reduction of metal artifacts in severely affected pellets. Similarly, 130 keV VMI with O-MAR resulted in a decrease of metal artifacts in severely affected pellets. In addition, 130 keV VMI with O-MAR resulted in a decrease of metal artifacts in mildly affected pellets compared to conventional imaging with O-MAR and therefore resulted in the strongest metal artifact reduction for both mildly as severely affected pellets.

Our quantitative analysis using a THA phantom shows that high energy VMI with O-MAR results in the strongest metal artifact reduction, and this is supported by studies incorporating subjective assessment by radiologists [5‐7] and by studies incorporating quantitative assessment of metal artifact reduction [5, 7, 23]. However, these studies used General Electric (GE) or Siemens systems to extract VMI with metal artifact reduction software, while we used a Philips system to extract VMI with O-MAR. Although metal artifact reduction techniques of the different vendors show comparable results regarding metal artifact reduction using VMI and VMI with metal artifact reduction software, it remains difficult to compare the results. A phantom study comparing metal artifact reduction of the different vendors would provide more insight in the performance of the different DECT techniques of the vendors and their metal artifact reduction software.

Although 130 keV VMI with O-MAR resulted in the strongest relative metal artifact reduction, it should be noted that the CT values, CNR and SNR of the unaffected, mildly affected and severely affected pellets were lower in the 130 keV VMI with O-MAR in comparison to conventional images with O-MAR. This can be explained by the high virtual monochromatic energy of the 130 keV images, causing a reduction of the CT values in the images in comparison to conventional images. Although low SNR and CNR can affect the detectability of the pellets and therefore diligence is important, subjective analysis of patients shows that high energy VMI with O-MAR still can be assessed reliable [22]. Furthermore, the Rose criterion states that an object’s CNR must exceed 3–5 to be detectable [24, 25]. Both conventional imaging with O-MAR as well as 130 keV imaging with O-MAR exceeded this criterion for both mildly affected and severely affected pellets.

On our 130 keV images, bright artifacts were induced. Likewise, several studies described that VMI and metal artifact reduction software can induce secondary artifacts [7, 26, 27]. Hence, the radiologist should also review conventional images in addition to the images with metal artifact reduction. Furthermore, relatively high negative metal artifact reduction in CNR (− 69%) and SNR (− 124%) was observed in severely affected pellets while high p values (0.889 and 0.889, respectively) were found when comparing 130 keV VMI without O-MAR and conventional imaging without O-MAR, which seems contradictory. This may be explained by the large dispersion of CNR values and SNR values of the individual severely affected pellets in the 130 keV VMI without O-MAR, caused by the introduction of bright artifacts.

Our study has limitations. First, we categorized the pellets in the images with the unilateral and bilateral prostheses in unaffected, mildly affected and severely affected which allowed comprehensible analysis of all pellets. However, mildly affected pellets were caused by unilateral and bilateral prostheses. Assignment of these pellets to one category may not be in accordance with clinical practice because a unilateral prosthesis causes different artifacts than bilateral prostheses Furthermore, pellet L9 in the bilateral image was categorized as unaffected despite bright streaking artifacts, and a more appropriate category may be mildly affected instead. Third, our results did not quantify metal artifact reduction in the close vicinity of the implant. Fourth, the pellets only represent one bone density, comparable to trabecular bone. Bone structures with higher densities such as cortical bone may alter the results. Finally, we chose to place the background ROI in a part of the image where no metal artifacts were present. Although this avoids metal artifacts in the background ROI, the calculated CNR does not quantify contrast between pellets and the background in the vicinity of the pellets.

We conclude that 130 keV VMI with O-MAR shows stronger metal artifact reduction over 130 keV without O-MAR and conventional imaging with O-MAR. It should be encouraged to reconstruct high energy VMI with O-MAR besides conventional polychromatic images after THA.