Imaging update in arthroplasty

Abstract

Imaging of metal implants has historically been difficult, regardless of the applied modality. The number of primary arthroplasties is increasing over the years. With it, we expect the number of symptomatic complications to increase as well. Acquiring accurate imaging for diagnosis and treatment planning for these cases is of paramount importance. Significant advancements have been made to reduce artifacts, leading to better imaging representation of arthroplasty. This review article would give a background on the current ways of imaging arthroplasty and metal implants, covering recent advances in imaging techniques.

1. Introduction

Imaging of metal implants has historically been difficult, regardless of the applied modality. The number of primary arthroplasties is increasing over the years. With it, we expect the number of symptomatic complications to increase as well.¹ Acquiring accurate imaging for diagnosis and treatment planning for these cases is of paramount importance. Significant advancements have been made to reduce artifacts, leading to better imaging representation of arthroplasty. This review article would give a background on the current ways of imaging arthroplasty and metal implants, covering recent advances in imaging techniques.

2. Modalities for imaging arthroplasty

2.1. Plain radiographs

Plain radiographs are the initial imaging performed for symptomatic joints, and asymptomatic follow-up. Routine radiographs performed immediately after surgery are not useful as they rarely have good diagnostic yield.² Limb-length assessment, as well as position of the prosthesis can be assessed. Complications such as loosening, bone abnormalities, or component failure can be detected.³ The lack of 3D spatial imaging and soft tissue delineation limits the utility of screening. Other modalities are used for further delineation and characterization of pathology.

2.2. CT and dual energy (DE) CT

The use of CT compliments radiographs; 3D images and models can be created, exact measurements of the implant angulation and position can be obtained. Conventional CT is hampered due to metallic artifacts,⁴ mainly due to beam hardening and photon starvation,^5,6 resulting in relevant anatomic structures being obscured. These are not avoidable, given the high atomic number of the metal contained in the implant. The larger the prosthesis, or higher the atomic number of the metal used, greater are the artifacts.⁴ New technologies such as projection-based metal artifact reduction algorithms (MAR) and dual energy CT have been developed to improve image quality.

Conventionally, increasing the tube charge or having a higher peak voltage can reduce photon starvation, but only to a minor degree.⁷ Decreasing the pitch over-samples data improves image quality.⁸ However, this leads to a higher radiation dose. Roth et al. gives an example of imaging parameters used at his centre.⁸

Projection-based MAR seeks to detect and segment the corrupted data, and replaces them with estimates of the correct value, either by using projection data, or reconstruction images^7,9, 10, 11[Fig. 1a]. When corrupted data is discarded, may lead to a certain degree of loss-of-information that cannot be fully recovered. Residual or newly developed streak artifacts have been reported after applying MAR techniques.¹² Beam hardening and scattering artifacts are not fully corrected for by the application of MAR¹²[Fig. 1b]. Underestimation of the implant size and sometimes complete removal of implant on the resultant image can occur as well.¹³ The commercially available MAR algorithms use reconstructed images, applying various technique(s), in succession.^9,11 These can be applied retrospectively to the scan data. As a general rule, image obtained with and without MAR should be reviewed together.

Fig. 1a.

Iterative metal artifact reduction (iMAR) sequence of a proximal femoral nail antirotation implant using proprietary Siemens algorithm. Note the reduction in beam hardening artifacts.

Fig. 1b.

iMAR image of a left total hip arthroplasty performed using the same Siemens algorithm in Fig 1a. Note the increased residual beam hardening artifacts due to the difference in implant type.

Increasing the tube voltage would also reduce beam hardening, but will also increase the radiation dose if other parameters are kept constant. Higher voltage settings reduce tissue contrast on images.¹⁴ Conventional systems use filtration, calibration correction and beam hardening correction software to reduce this artifact.¹⁵ In newer dual energy systems, data is acquired at 2 different energy voltages, and reconstructed to monochromatic images using material decomposition technique. These function on the basis that different materials absorb energy differently at various voltages. In practice, users have to select the optimal energy levels to use in the dataset (usually 95-150 KeV¹⁶ for implants), to ultimately acquire a virtual monochromatic image, that yields the best balance between contrast, noise, and artifact [Fig. 2a, Fig. 2ba and b]. These have to be decided before the actual data acquisition, and cannot be altered retrospectively.

Fig. 2a.

Computed tomography of posterior spinal instrumenta'on from fusion of L2 to S1 without dual energy sequences.

Fig. 2b.

Dual energy computed tomography of the same patient. Note the reduction in beam hardening artifact and better visualization of the interface between implant screws and adjacent bone in comparison to Fig 2a.

For contrast-enhanced studies, the energy absorption characteristics of iodine have been studied extensively. The maximal iodine contrast-to-noise ratio occurs at around 70 KeV.¹⁷ When higher energy is applied to reduce beam hardening artifact, there is a simultaneous reduction in iodine contrast enhancement. Projection-based MAR, by itself, does not cause attenuation of iodine enhancement.

3. Magnetic resonance (MR) imaging

MR imaging yields the benefit of avoiding ionizing radiation, and better soft tissue contrast in musculoskeletal imaging. Metal implants pose a hazard, with their potential for dislodgement, heating, and damage to surrounding tissue. Concerns remain, especially with the increasing field strength of newer magnets. Few joint prostheses bear the mark of “MR safe” or “MR conditional”, and thus majority of implant imaging is performed off-label.¹⁸ In a literature review by Mosher et al.,¹⁹ potential implant displacement occurred at 7.0T, and with ferromagnetic implants. Non-ferromagnetic implants showed no potential displacement at 1.5–3.0T field strength. A few studies showed minimal torque in titanium/titanium alloy implants²⁰ at 7T. Most of the studies show minimal (no more than 1 °C change), but one ex-vivo study showed increase of up to 14.7 °C, at the edge of the implant, when the implant was parallel with the static magnetic field.²¹ Mosher et al. believes these studies are not representative, because the model may not reflect true in vivo properties. Most other studies show that fears of temperature increase and tissue damage from RF heating may be unfounded.^22,23 The specific absorption rate (SAR) is a measure of the amount of power deposited in a certain mass of tissue, and is related to many imaging parameters. At higher field strengths (3T), due to the increased bandwidth, increased signal acquired and prolonged echo train length, SAR time-out may occur, causing the study to be truncated. This is unlikely to occur at 1.5T.

There is lack of consensus regarding the use of MRI immediately postoperatively. Generally, 1.5T field strength is safe for non-ferromagnetic implants in the immediate postoperative period. Imaging of ferromagnetic implants may be performed 6–8 weeks after the procedure.²⁴

Without proper consideration and planning, artifact from metallic prosthesis often results in poor image quality. In general, the more ferromagnetic the metal, or the higher the field strength, the greater the resultant artifact.

Sequences such as spin-echo, fast spin-echo, or turbo spin-echo reduce in-plane distortion artifacts by using multiple radiofrequency pulses to refocus, and reduce the degree of signal loss around the implant.⁷ Gradient-echo based sequences should be avoided, as it amplifies the resultant susceptibility. Swapping the phase and frequency-encoding directions may change the orientation of the artifact, and can shift the signal loss to a less anatomically important location.²⁵ Short tau inversion recovery (STIR) sequences and Dixon-based techniques perform better than spectral-based fat suppression technique around implants.²⁵ Increasing receiver bandwidth can limit the spread of geometric distortion, but causes a decreased signal-to-noise ratio (SNR). The exact bandwidth varies between vendors and machines.²⁵ Obtaining thinner sections can limit the through-plane distortion, and increase the SNR, but the SNR outside the region of artifact will be reduced.²⁵ Increased resolution offers the smallest improvement on image quality, but further reduces SNR, and increases imaging time. As most techniques cause reduced SNR, increasing the number of signals acquired is often the best strategy at regaining SNR, at a cost of a longer scan time. T1-weighted sequences suffer the most from metallic artifacts, as short echo spacing is difficult to achieve. T1-weighted sequences also cause a significant increase in SAR, especially at 3T, due to the low repetition times required.

Advanced metal artifact reduction sequence (MARS) exists for MR as well. Some centers use a non-proprietary MARS package, mainly combining the above techniques to create diagnostic images, Talbot et al.²⁵ has provided a sample of their imaging parameters for use on a 1.5T machine. Other, more advanced, proprietary MARS packages include SEMAC, WARP (SEMAC with view-angle tilting (VAT)), MAVRIC, O-MAR XD, CAIPIRINHA etc, some of which involve a series of techniques, rather than a specific one [Fig. 3]. These are gated behind proprietary vendors, and are machine specific. MAVRIC and SEMAC are the most widely used techniques, both of which have shown similar efficacy,²⁶ but generally results in longer scan times.²⁵ Filli et al.²⁶ showed that, depending on the material, and designated scan time, various strengths of SEMAC or MAVRIC can be used. For example, SEMAC is generally faster, and weak SEMAC may be used for titanium implants due to better SNR. Other combination techniques, such as MAVRIC-Selective (MAVRIC-SL), ultrashort TE (UTE) with MAVRIC are also available, and show promising results on artifact reduction compared with traditional FSE sequences.^26,27

Fig. 3.

MRI of the left ankle with screw fixation of a talus fracture. O-MAR sequencing using a 1.5T Philips scanner reduces the metallic susceptibility artifact of the talar screws and shows the interface between the implant and surrounding bone.

3.1. Nuclear medicine

Traditionally, technetium-99 m labeled diphosphonate was used in bone scintigraphy to detect prosthetic failure. These were sensitive indicators, but not able to indicate the cause of failure. These were generally used as a screening tool, where a normal study effectively rules out prosthesis failure as a cause for symptoms.²⁸ Accuracy was low, quoted as 50–70%, mainly because many normal prosthesis showed persistent uptake even after 1 year post surgery.²⁸

Gallium-67 citrate scintigraphy, was developed in an attempt to pinpoint infection due to increased local blood flow and bacterial uptake of the tracer. Most agree that it is specific, but some have data contending the sensitivity.²⁹ Some centers perform it with bone scintigraphy as a paired test, in an attempt to even out the sensitivity and specificity, but to varying degrees of accuracy (50–95%²⁹).

Most centers have adopted in-vitro labeled leukocyte imaging (WBC scan) over the aforementioned techniques. Neutrophils are usually the majority of the labeled cells. When Indium-111 is used as the label, images are acquired 18–30 h after administration. When technetium-99 m is used, imaging is performed at 4–6 h and repeated 18–30 h after administration. Again, there is mixed data regarding the sensitivity and specificity,²⁹ ranging from 23% to 100% depending on the criteria used for a “positive” study. This is usual since there is no single satisfactory method for interpreting these images. Study images are usually interpreted by comparing intensity of activity of the affected joint, to a reference point at the contralateral limb, or against an adjacent bone; it is considered positive when the intensity of activity exceeds that of the reference point. 100% sensitivity is achieved when any activity is seen at all, at large cost to specificity. Comparing against a reference point is problematic as leukocyte distribution in the bone marrow is grossly unpredictable, with both generalized and localized patterns,³⁰ depending on body constitution and underlying diseases.

Some centers advocate interpreting a WBC scan with a bone scintigraphy, again with mixed results,²⁸ with sensitivity and specificity ranging from 64 to 89%. Fundamentally, bone scintigraphy tracer accumulates in the cortical bone, while leukocytes accumulate in the marrow. Conditions may affect one or the other, and give conflicting results.²⁸

The use of sulfur colloid to image bone marrow, and subsequent comparison of the images with a WBC scan can be performed. Normal marrow should be congruous on both scans, regardless of the underlying disease state, with the exception of osteomyelitis. Most report high sensitivity and specificity (92–100% ²⁸). This can be done sequentially (72 h apart), but if In-111 label is used, dual-isotope acquisition can be performed. However, the in vitro labeling process is labor intensive, poor in immunocompromised individuals, and the need to perform marrow imaging is cumbersome, especially for the less-abled. In-111 labelling often produces lower quality images compared with Tc-99 m labeling.

18F-FDG PET is used in some centers, due to the 3D spatial imaging and relative rapid completion of the study. Sensitivity and specificity were 82% and 87% respectively, based on a metanalysis,³¹ but the large number of inconsistent and contradictory results amongst the studies examined raises concern and leaves the role of 18F-FDG PET to be determined. Overall, most studies quote a lower accuracy, when compared with WBC scan/bone marrow imaging.²⁹

Single-photon emission computed tomography/Computed tomography (SPECT/CT), has been described to be useful for evaluating failed joint arthroplasties,²⁹ and as a potential alternative to the dual isotope procedure listed above. The CT component provides anatomic landmarks for localization, and in general provides better detection and characterization of the problem compared with planar imaging32, 33, 34 with increased sensitivity and specificity. Additional information from the CT component is valuable to diagnose other causes, in a negative study.

3.2. Ultrasound

Ultrasound cannot penetrate bone or metal, and its role is limited in the primary assessment of prosthetic complications. It can however serve to identify and guide drainage of fluid collections [Fig. 4], and delineate sinus tracts, as a comparatively inexpensive investigation.

Fig. 4.

(A) radiograph of the right wrist with periprosthetic lucencies (B) corresponding ultrasound of the same patient demonstrating an irregular thick walled collection with echogenic contentsin the vicinity of the metallic prosthesis likely representing a pseudotumor related to metallosis.

4. Examples of complications

4.1. Infection

It is one of the most devastating complications following arthroplasty, fast diagnosis is key to preserve the joint function, to limit systemic sepsis, and to contain tissue damage. Many joint infections do not have the classical symptoms of fever, chills, joint pain, erythema and discharge,³⁵ with cultures not positive for as long as 2 weeks.³⁶ Prevalence is around 1–2%.^35,37 Most common organisms are staphylococci and streptococci. Risk factors include immunosuppression, inflammatory arthritis, obesity, repeat surgery, dental infections, poor nutrition and distal infections.³⁸ Early infections occur within 2 months of surgery, and may be due to wound complications. Late complications occur because of hematogenous seeding, often with a distant infection. [Fig. 5].

Fig. 5.

Radiograph of the left elbow post internal fixation of a radial fracture with corresponding dual energy computed tomography images illustrating the areas of marrow oedema (depicting osteomyelitis) represented by green colour coding around the implant site at the proximal radius and ulna.

On radiographs, progressive irregular lucency around the implant and periosteal reaction appears only in an advanced stage of the disease.³⁹

CT can be useful, showing osteolysis and lucent lines, but ultimately the more sensitive and specific findings are seen in the surrounding soft tissue (100% sensitive and 87% specific). These include fluid collections, synovitis and sinus tracts⁴⁰[Fig. 6].

Fig. 6.

(A) Initial post-operative radiograph of the left knee following total knee replacement surgery. (B) Radiograph of the same patient 9 years post implantation showing significant implant subsidence (C) Corresponding computed tomography of the lower extremity also demonstrates gas within the cement adjacent to the tibial implant indicative of infective changes with chronic changes of osteomyelitis also shown.

MR imaging is helpful in making the diagnosis. Lamellated hyperintense synovitis seen in 86–92% and 85–87% was a specific finding.⁴¹ It allows delineation of abscesses and effusion, allowing for targeted drainage and cultures to be obtained [Fig. 7].

Fig. 7.

Coronal magnetic resonance imaging of the pelvis with MARS sequencing in a patient post uncemented bilateral hip arthroplasty with suspicion of joint infection due to complaints of hip pain a year post implantation. Oedema is present in the right hip joint and adjoining soft tissue supporting this diagnosis.

In cases where findings are equivocal, nuclear medicine imaging may be performed; various centers examine cases with various combinations of the 3 main tracers,⁴⁰ some using all 3.⁴² These, however, have limited availability and increased costs.

4.2. Alignment assessment, subsidence, instability and dislocation

Especially for the weight bearing joints, good alignment is essential for implant longevity.⁴³ Malalignment contributes to a large proportion of symptomatic replaced joints.⁴⁴ These measurements may be obtained on XR, CT⁴⁵ or MR. Greater inter-observer variation has been observed with more susceptible ferromagnetic implants such as those containing 28cobalt/chrome/molybdenum alloy when compared to zirconium, during assessment with MRI.⁴⁶

Subsidence, occurs when there is component migration of the implant relative to the adjacent bone [Fig. 8]. We still do not have complete understanding regarding the reason for subsidence,^47,48 but certain implant and designs as well as anatomical properties seem to play a role.^49,50 Various criteria to define subluxation have been published depending on the site and type of implant used.

Fig. 8.

(A) Initial radiograph following total ankle replacement (B and C) Subsequent radiograph and computed tomography images demonstrating talar plate subsidence with medial migration resulting in incongruency with the tibial plate.

For hip prosthesis, this is defined as femoral stem distalization, with reference to the greater trochanter. This can be measured using reference lines from anatomical landmarks, or using software such as EBRA and Imagika.⁵¹ Subsidence up to 10 mm can be normal in the first year following uncemented arthroplasty⁵²[Fig. 9]. These tend to occur more in collarless implants compared to collared ones.⁵³ Differences in canal flare index (CFI), and canal fill ratio (CFR) was not identified as a significant factor.⁵³

Fig. 9.

Computed tomography and corresponding plain radiograph of a left total hip replacement implant demonstrating implant subsidence. Depression of the medial tip of the prosthetic shoulder in relation to the lesser trochanter.

Gross assessment of hip implants alignment on radiographs can be performed,⁴⁵ but methods of assessment vary among authors. Park et al.⁵⁴ compared 6 different methods, and found that Liaw et al. method was the most accurate, while the Woo and Morrey method was the most consistently reproducible.

In the knee joint, femoral component rotation can be measured on a single CT slice.⁵⁵ Analysis of the femoral component rotational angle (FCR angle) using the CT derived surgical transepicondylar axis (CTsTEA) and femoral component rotational axis is considered the gold standard for alignment assessment⁵⁶ [Fig. 10a].

Fig. 10a.

(A) CT derived surgical transepicondylar axis (CTsTEA); the line drawn from the most prominent part of the lateral epicondyle to the sulcus in the medial epicondyle (B) Femoral component rotational axis (FCRA) is the common tangent of the two pegs on the inside of the femoral component. Superimposing the CTsTEA on the same axial image we get the femoral component rotational angle (FCR angle) of zero degrees in this case.

Some authors believe MR imaging is more accurate than CT to visualize the medial epicondylar sulcus; a landmark used in the measurement of the surgical transepicondylar axis.⁴⁶

Tibial component rotation measurement is more complicated, and there are various 2D (e.g Berger⁵⁷) and 3D (e.g Mayo⁵⁸) CT techniques described in the literature. The Berger technique uses transposition of 3 CT slices, and measures rotation relative to an axis, defined by the geometric center of the tibial plateau, to the tip of the tubercle [Fig. 10b]. The Mayo technique requires 3D reconstructions of the tibia component, and measures rotation relative to the axis defined by the center of the tibial tray, to the junction of the medial and middle thirds of the tibia tubercle. The 3D Mayo technique is described to be the “gold standard” by a few authors,^58,59 but requires dedicated software and processing. A more recent index described by Saffi et al.,⁶⁰ is the CTTT distance (center of tibial tray to the tip of the tibial tubercle); a rapid 2D technique, requiring 2 slices to be transposed with each other.

Fig. 10b.

Berger method for tibial component rotation measurement. (A) 2D axial CT at the level of the tibial plateau identifying the geometric centre [GC] (B) a line is then drawn from the GC to the tibial tuberosity [TT] (C) the line from the GC to TT is transposed to an axial image at the tibial tray. The angle between this line and the tibial component axis illustrates the extent of external rotation rela've to the tip of the tibial tubercle. Eighteen degrees in this case.

The new technique was found to have a strong correlation with the more accurate and reproducible 3D Mayo technique,⁶⁰ and could be measured within ∼20% of the duration taken for the 3D technique.

Frank dislocations are easy to pick up clinically and, on a radiograph, but 3D imaging of the joint is obtained for implant alignment assessment, and subsequent surgical planning.

4.3. Periprosthetic fractures or stress reaction

These occur in 0.11–21.4% of patients with a mean time of 2–4 years after surgery.⁶¹ Risk factors include trauma, low bone density and prior revision surgery. Fast diagnosis allows for preservation of joint and implant components. XR would be the initial investigation of choice, picking up gross abnormalities with displacement; it is usually sufficient in most cases.⁵⁴ CT allows for surgical planning, especially in complex fracture patterns⁴⁵[Fig. 11]. MR imaging is sensitive, and is useful in a radiographically occult fracture. The expected findings of bone marrow oedema, cortical thickening, periosteal reaction and fracture line are seen. Without a fracture line, these findings indicate a stress reaction.

Fig. 11.

Computed tomography of the left hip following a hemiarthroplasty. Bony fragmentation noted at the anterior acetabular column in keeping with a periprosthetc fracture.

4.4. Heterotopic ossification

This is defined as an abnormal formation of mature lamellar bone within extra-skeletal soft tissues. This occurs early after surgery (45% at 1 year⁶²), and can be well seen on radiographs. CT can provide better anatomical localization for surgical planning⁶³[Fig. 12].

Fig. 12.

(A) plain radiograph of left femur following bipolar hemiarthroplasty (B and C) corresponding computed tomography images. Heterotopic ossification of the adjacent soft tissue lateral to the mid shaft of the femur along the distal femoral implant site.

4.5. Osteolysis and particle disease

Wear of implant components (mainly polyethylene), releases small particles, resulting in intra-articular inflammatory reaction, manifesting as synovitis.⁶⁴ These gradually involve the bone-implant interfaces, and cause geographic osteolysis [Fig. 13]. MR is the most sensitive for lesion detection, and can also demonstrate the characteristic low-to-intermediate signal synovial thickening due to the debris⁶⁵ [Fig. 14].

Fig. 13.

Pseudotumor in the right hip. (A) right hip radiograph post total hip arthroplasty (B) corresponding CT scan of the same patient showing osteolysis in the acetabulum, iliac bone and ischium around the prosthesis along with a well circumscribed homogenous mass lesion in keeping with a pseudotumor.

Fig. 14.

Pseudotumor in the left hip following total hip replacement surgery. (A, C) Axial T1 weighted (C) post contrast and (B, D) STIR (B) with SEMAC (metal reduction sequence) showing a lobulated collection with high signal intensity fluid and intermediate to low signal intensity debris along with osteolysis. This is seen to be in communication with the arthroplasty hardware.

4.6. Aseptic loosening

This is one of the most significant long-term complication of joint replacements.⁶⁶ It can be the result of inadequate initial fixation, or due to particle disease around the implant.⁶⁷ Serial radiographs are the cheapest and most valuable modality in detecting this. Lucency of more than 2 mm around the implant,⁶⁸ or progression of lucency would indicate loosening of the prosthesis⁶⁹[Fig. 15].

Fig. 15.

(A) Initial post-operative plain radiograph of a left hip proximal femoral nail antirotation implant. (B) Subsequent hip radiograph and corresponding computed tomography images of the same patient two years after implantation. Note the implant loosening with increased lucent margin around the lag screw at the femoral head component and increased varus tilt of the intramedullary nail.

CT can be utilized if there is an equivocal radiograph finding. Due to beam hardening artifact around the implant, radiolucent areas may be less apparent, whereas areas >2 mm can be considered as normal if no changes are seen on serial imaging.⁶⁹

5. Conclusion

Currently, radiographs are most commonly used for routine follow-up cases, and for the initial assessment of the painful arthroplasty.

MARS techniques for CT and MRI are usually the next step in the evaluation, and most vendors have their own proprietary protocols. Institutions may fall-back on standard protocols adhering to the basic principles detailed above; several authors have published their protocols. These allow for better characterization of the bone and soft tissues around the joint, allow measurements of implant alignment, and planning for revision surgery. For MRI, the use of a 1.5T machine is generally safe, reduces artifact, and reduces the chances of SAR time-out. Post-processed MAR images should be viewed in conjunction with the source dataset.

Nuclear Medicine, if available, allows for troubleshooting, and acts as a screening tool where a negative study effectively rules out arthroplasty failure as a cause for the pain. Various protocols are used, with most agreeing that WBC/Marrow dual isotope imaging is best at this point in time.

With the advent of Artificial Intelligence (AI), vendors are starting to tap into this technology to develop better MARS images. Future MARS techniques will be increasingly exclusive, vendor-specific and technique selection would be mainly influenced by the choice of machine an institution chooses to purchase.