Artifact reduction strategies for prosthetic heart

Authors: Jesse Habets Petr Symersky Tim Leiner Bas A J M de Mol Willem P Th M Mali Ricardo P J Budde

Publish Date: 2012/04/05

Volume: 28, Issue: 8, Pages: 2099-2108

PDF Link

Abstract

Multislice CT evaluation of prosthetic heart valves PHV is limited by PHVrelated artifacts We assessed the influence of different kV settings a metal artifact reduction filter MARF and an iterative reconstruction algorithm IR on PHVinduced artifacts in an in vitro model A MedtronicHall tilting disc and St Jude bileafet PHV were imaged using a 64slice scanner with 100 kV/165 mAs 120 kV/100 mAs 140 kV/67 mAs at an equal CTDIvol Images were reconstructed with 1 filtered back projection FBP 2 IR 3 MARF and 4 MARF and IR Hypo and hyperdense artifacts volumes mean mm3 ± SD were quantified with 2 thresholds ≤−50 and ≥175 Hounsfield Units Image noise was measured and the presence of secondary artifacts was scored by 2 observers independently Mean hypodense artifacts for the MedtronicHall/St Jude valve FBP were 966 ± 23/1738 ± 21 at 100 kV 610 ± 13/991 ± 12 at 120 kV and 420 ± 9/634 ± 9 at 140 kV Compared to FBP hypodense artifact reductions for IR were 9/8  10/7  and 12/6  respectively for MARF 92 /84  89/81  and 86/77  respectively for MARF + IR 94/85  92/82  and 90/79  respectively Mean hyperdense artifacts for the MedtronicHall/St Jude valve were 5530 ± 48/6940 ± 70 at 100 kV 5120 ± 42/6250 ± 53 at 120 kV and 5011 ± 52/6000 ± 0 at 140 kV Reductions for IR were 2/2  2/3  and 3/4  respectively for MARF were 9/30  0/25  5/22  respectively MARF + IR 12/32  4/27  and 7/25  respectively Secondary artifacts were found in all MARF images Image noise was reduced in the IR images In vitro PHVrelated artifacts can be reduced by increasing kV despite maintaining identical CTDIvol Although MARF is more effective than IR it induces secondary artifactsProsthetic heart valve PHV assessment is a promising new application for multislice CT MSCT 1 2 3 4 5 6 7 Although echocardiography is the mainstay of functional evaluation of prosthetic valves it is hampered by acoustic shadowing and it may not be able to identify periprosthetic obstructive masses or false aneurysms 5 6 7 The visualization of areas considered “offlimits” to echocardiography with MSCT allows detection of obstructive masses and may aid in the management of these patients 2 5 In addition MSCT allows the detection of periprosthetic leaks vegetations and degenerative changes in biological prosthetic valves and allows evaluation of leaflet motion in mechanical valves 1 2 3 4 5 6 7 Despite the excellent spatial and good temporal resolution of current MSCT technology PHV CT images vary in quality For mechanical PHV a variable amount of valveinduced artifacts remains due to the radiopaque and metal components of the PHV 2 3 8 9 Compounding the problem of artifacts are the differences in PHV composition PHV consisting of cobalt chromium components such as the BjörkShiley and Sorin tilting disc valves are associated with severe artifacts that prohibit CT assessment of these valves In contrast modern mechanical PHV that consist of tungsten impregnated carbon leaflets and titanium or nickel alloy rings induce far less artifacts and allow a much more complete visualization of the periprosthetic anatomy 1 2 4 6 Because the masses interfering with normal PHV function and periprosthetic leaks are directly adjacent to the high attenuation components of the PHV further reduction of the PHVrelated artifacts may further improve the diagnostic yield of MSCTThe problem of metal artifacts is ubiquitous in CT imaging and many different strategies have been devised to improve the image quality around metal objects These strategies reflect the multiple mechanisms that cause these artifacts On the one hand physicsrelated interactions such as beam hardening scatter and photon starvation are important On the other hand algorithms used for reconstruction of these faulty raw data may augment artifacts by for example creating windmill artifacts and artifacts related to helical interpolation 9 10 11 12 As mentioned above these interactions are further complicated by the differences in PHV composition which have been related to the severity of artifacts Hence a single intervention such as the increase of beam energy or iterative image reconstruction may decrease some artifacts but not sufficiently eliminate them 10 11Our goal in this study was to evaluate the effectiveness of three ways to reduce PHVrelated artifacts 1 variation of tube voltage beam energy 2 applying a metal artifact reduction filter and 3 iterative image reconstruction By comparing these three approaches for the reduction of PHVrelated artifacts in an in vitro model we sought to determine the effectiveness of each method for optimizing the MSCT image quality of PHVAll scans were performed on a 64slice CT scanner Brilliance 64 Philips Medical Systems Cleveland Ohio USA Because the metal artifact reduction filter can not yet be used with ECGgated scans protocols we adapted a standard scan protocol for thoracic imaging to have the same imaging parameters as a standard retrospectively gated CT of the heart This was done by adjusting the following parameters pitch 02 collimation 64 × 0625 mm 120 kV matrix size 512 × 512 gantry rotation time 420 ms Because a nonECGgated image acquisition results in much lower noise levels we adjusted the mAs setting to a lower value at which the same level of noise defined as the standard deviation SD of CT attenuation was present as that found in earlier ECGgated in vitro experiments 9 using a standard ECGgated cardiac protocol Noise was measured using a circular region of interest diameter 1 cm in a homogenous part of the PMMA structure of the valve chamber not affected by the PHVrelated artifacts For the adapted thoracic protocol at 120 kV a mAs setting of 100 resulted in equal image noiseFor the experiment three different acquisition protocols with different tube voltage settings with identical CTDIvol and DLP values were used For this the adapted thoracic protocol was adjusted to 100 and 140 kV to yield identical CTDIvol and DLP values obtained for the 120 kV 100 mAs scan This resulted in scans performed at 140 kV 67 mAs and 100 kV with 165 mAs Ten scans of each valve were performed with each kV mAs setting A standard reconstruction filter was used because the detailed cardiac filter is not available for nongated scans Images were reconstructed at 09 mm thickness with a 045 mm increment

Keywords: