Quantitative evaluation of artefact reduction from metallic dental materials in short tau inversion recovery imaging: efficacy of syngo WARP at 3.0 tesla

Abstract

Objectives:

To evaluate the effects of syngo WARP on reducing metal artefacts from dental materials.

Methods:

Short tau inversion recovery (STIR) with syngo WARP [a dedicated metal artefact reduction sequence in combination with view-angle-tilting (VAT)] was performed using phantoms of three dental alloys: cobalt–chromium (Co–Cr), nickel–chromium (Ni–Cr), and titanium (Ti). Artefact volumes and reduction ratios of black, white and overall artefacts in the standard STIR and syngo WARP images with several different parameter settings were quantified according to standards of the American Society for Testing and Materials F2119-07. In all sequences, the artefact volumes and reduction ratios were compared. The modulation transfer function (MTF) and contrast-to-noise ratio (CNR) were also measured for evaluation of image quality.

Results:

In standard STIR, the overall artefact volume of Co–Cr was markedly larger than those of Ni–Cr and Ti. All types of artefacts tended to be reduced with increasing receiver bandwidth (rBW) and VAT. The effect of artefact reduction tended to be more obvious in the axial plane than in the sagittal plane. Compared with standard STIR, syngo WARP with a matrix of 384 × 384, receiver bandwidth of 620 Hz/pixel, and VAT of 100 % in the axial plane obtained reduction effects of 30 % (white artefacts), 45 % (black artefacts), and 38 % (overall artefacts) although MTF and CNR decreased by 30 and 22 % compared with those of standard STIR, respectively.

Conclusions:

syngo WARP for STIR can effectively reduce metal artefacts from dental materials.

Introduction

In MRI of the head and neck region, the presence of metallic dental materials often causes metal artefacts including signal-loss (black artefacts), signal-pileup (white artefacts), geometric distortion, and failure of fat suppression.¹² These artefacts can result in obstacles to obtaining optimal images and can complicate image interpretation.

Based on the physical phenomena associated with artefact formation, metal artefact reduction sequences (MARS) have been developed by optimizing conventional MR sequences, and are already available as product sequences that are widely applied.³

Recently, MR technical innovations in hardware and software have significantly reduced metal artefacts through the use of several advanced techniques, such as view-angle-tilting (VAT), slice-encoding for metal artefact correction (SEMAC), and multiacquisition variable-resonance image combination (MAVRIC). syngo WARP, commercialized by Siemens Healthcare, is a dedicated MARS technique in combination with VAT.^3–5 VAT uses a gradient applied on the slice-select axis during readout; therefore, in-plane artefacts can be corrected without increasing imaging time.⁶ Many publications have shown that syngo WARP has been successfully applied in orthopaedic radiological contexts, including hip and knee replacement.^7,8 To our knowledge, only a few studies have investigated the application of syngo WARP for reducing metal artefacts caused by commonly used dental materials.^9,10 However, no studies have investigated quantitative evaluation of the artefact reduction effects of the syngo WARP technique on reducing metal artefacts caused by metallic dental materials.

The aim of this study was to evaluate the effects of syngo WARP on reducing artefacts caused by metallic dental materials in head and neck MRI at 3 T.

Methods and materials

Phantoms

Three types of dental materials—cobalt–chromium (Co–Cr), nickel–chromium (Ni–Cr) and titanium (Ti) alloys—were cast into three cylindrical samples, 4 mm in diameter and 20 mm in height. Characteristics of the sample materials are listed in Table 1. Each metal sample was sealed in a cylindrical acrylic vial with agar. A vial containing only agar was used as the reference object. Each of these vials and another vial containing pure water were inserted into a box-shaped dedicated phantom (160 × 160×160 mm³) filled with polyvinyl alcohol gel (93-402S; Nikko Fines Industries Co, Tokyo, Japan). Furthermore, a vial containing pure water and a vial containing Gadolium 0.4 mM were used to measure the contrast-to-noise ratio (CNR).

Table 1.

Characteristics of dental materials used in phantom study

Dental casting alloys	Trade name	Manufacturer	Composition with mass (%)
Co–Cr	Realloy-C	Realloy e.K. Germany	Co 63%, Cr 25%, Mo 3.8%, W 9.0%, Si 1.0%
Ni–Cr	Verabond	Aalba Dent, Inc. USA	Ni 77.9%, Cr 12.6%, Mo 5.0%, Al, Be, Co
Ti	Neotitan	Neodontics, Inc. USA	Ti 5.0–6.0%, Ni 71%, Cr 12%, Mo 9.0%, Co 1.0–2.0%

Co, cobalt; Cr, chromium; Mo, molybdenum; W, tungsten; Si, silicon; Ni, nickel; Al, aluminium; Ti, titanium

MRI techniques

A 3 T MR scanner (Magnetom Spectra; Siemens Healthcare, Erlangen, Germany) and a 16-channel head and neck coil were used to obtain all MR images. syngo WARP, a dedicated MARS technique in combination with VAT, was used for reduction of metal artefacts. Artefact reduction effects via reducing through-plane and in-plane distortion can be obtained by increasing the bandwidth of both the excitation pulse and the signal readout; this was implemented in the WARP package.⁵ Additionally, artefacts can be reduced with a VAT technique, which uses an additional gradient in a slice-select direction during readout to reduce in-plane distortions caused by susceptibility gradients.⁶ To evaluate the artefact reduction effect, conventional short tau inversion recovery (STIR) images and STIR images combined with WARP were obtained in the axial and sagittal planes with different parameter settings (Table 2). The common image parameters were as follows: repetition time 5000 ms, inversion time 210 ms, echo train length 11, number of signals averaged 1, field of view 230 × 230 mm, slice thickness 3 mm with no gap, number of slices 25. Parallel imaging acceleration was not used in this study.

Table 2.

Sequences and imaging parameters

Number		Matrix	TE (ms)	rBW [Hz/pixel]	WARP	VAT (%)	Acquisition time (minute:second)
1	Standard	384 × 384	84	296	Off	0	6:00
2	M384-WARP-rBW296-VAT0	384 × 384	87	296	On	0	6:00
3	M384-WARP-rBW296-VAT50	384 × 384	87	296	On	50	6:00
4	M384-WARP-rBW296-VAT100	384 × 384	87	296	On	100	6:00
5	M384-WARP-rBW620-VAT0	384 × 384	85	620	On	0	6:00
6	M384-WARP-rBW620-VAT50	384 × 384	85	620	On	50	6:00
7	M384-WARP-rBW620-VAT100	384 × 384	85	620	On	100	6:00
8	M256-WARP- rBW296-VAT0	256 × 256	84	296	On	0	4:10
9	M256-WARP- rBW296-VAT50	256 × 256	84	296	On	50	4:10
10	M256-WARP- rBW296-VAT100	256 × 256	84	296	On	100	4:10
11	M256-WARP- rBW610-VAT0	256 × 256	84	610	On	0	4:10
12	M256-WARP- rBW610-VAT50	256 × 256	84	610	On	50	4:10
13	M256-WARP- rBW610-VAT100	256 × 256	84	610	On	100	4:10
14	M256-WARP- rBW930-VAT0	256 × 256	83	930	On	0	4:10
15	M256-WARP- rBW930-VAT50	256 × 256	83	930	On	50	4:10
16	M256-WARP- rBW930-VAT100	256 × 256	83	930	On	100	4:10

TE, echo time; rBW, receiver bandwidth; VAT, view angle tilting; M384, matrix of 384 × 384; M256, matrix of 256 × 256

According to the American Society for Testing and Materials standard (ASTM-F2119-07), for images containing metal devices, two sets of images must be acquired using both possibilities for designation of readout and phase-encoding directions in each plane, and for images containing a reference object, one image may be acquired using either possibility for designation of readout and phase-encoding directions in each plane. In this study, phase-encoding directions were set at right-to-left (R–L), left-to-right (L–R), anterior-to-posterior (A–P), and posterior-to-anterior (P–A) in the axial plane, and A–P, P–A, head-to-foot (H–F), and foot-to-head (F–H) in the sagittal plane for the metal samples, and at R–L and A–P in the axial plane, and A–P and H–F in the sagittal plane for images of the agar vial.¹¹

For each metal sample, the STIR sequence series was performed for the metal and agar vials. During the scanning procedure for the metal samples and the agar vial, the phantom and MRI tables were not moved in order to retain the same spatial position. Furthermore, a pre-scan was performed only once for every set of metal sample/agar vial (the first sequence of the STIR series for the metal sample), and the other sequences for the metal samples and agar vial were performed with no pre-scan mode. Therefore, for each metal sample, all images of the metal samples and agar vial were obtained under the same conditions of B₀ field homogeneity, coil tuning/matching, transmitter attenuation/gain, receiver attenuation/gain, and dummy cycles.

The sequence series for each metal sample was performed three times with resetting of the phantom position to quantify the individual measurement variation and to confirm the reproducibility.

For measurement of the CNR in each sequence, STIR images and STIR images combined with WARP were also obtained in the axial plane with the same settings for the dental materials (Table 2).

Measurement of artefact volume

According to ASTM-F2119-07, a pixel is considered as a part of the artefact if its signal intensity (SI) is changed by at least 30% when the device is present compared with a reference image in which the device is absent.¹² The areas enclosing the pixels exceeding 30 % SI change above or below are considered to be white or black artefacts, respectively, and the sum of all white and black artefacts is considered as the overall artefacts. These artefact volumes were measured quantitatively by the following image processing and measurement method (Figure 1a):

Figure 1.

(a) Diagram demonstrating image processing and the method to quantitatively measure the artefact volume. The flowchart summarizes the process of artefact volume measurement. (b) The pin hole section embedded inside the phantom for evaluation of the effects of syngo WARP on image sharpness. A line (red dash) was drawn across the pin hole section (five pins with 2.0 mm diameter and interval) to produce a pin pattern profile curve. ROI, region of interest.

1. Creating divided images (metal images/reference images) for each metal sample image for every STIR sequence.

2. Setting rectangular regions of interest (ROIs) at the centre of each image enclosing the whole artefact (approximately 70 x 70 mm) and trimming the divided image data.

3. Binarizing the divided image (values of 0 and 255) by using the threshold of 1.3 or more for white artefacts, and 0.7 or less for black artefacts.

4. Counting the number of pixels automatically on the binarized image for white and black artefacts.

5. Converting the number of pixels into volume because metallic artefacts occur as in-plane and through-plane artefacts on the two-dimensional slice, and these artefacts are difficult to separate.^13,14 In this study, metal objects, which showed low signal intensity on the MR images, were not counted as artefact. Therefore, black artefact volume was obtained after excluding the volume of metal objects (251.3 mm³) from the ROIs. Overall artefact volume was calculated by combining the volume of white and black artefacts.

To measure the artefact volume, the ROIs were placed on the images in consensus by two radiologists (LTTX and JS with 4 and 17 years of experiences in head and neck MRI).

Artefact volume data in each sequence consisted of 12 measurements (four phase-encoding directions × 3 times) in each plane. Furthermore, reduction ratios (%) for white, black, and overall artefact volumes in syngo WARP for those of standard STIR were calculated for each material. All image processing and measurement methods were performed with software ImageJ 1.50i for Windows (National Institute of Health, Bethesda, MD).

The effects of syngo WARP on the image quality

To evaluate the effects of syngo WARP on the image sharpness and contrast, the modulation transfer function (MTF) and CNR were measured in this study. SI profiles were displayed as a two-dimensional graph of the pixel along a line going through the pin hole section within the images were obtained from standard STIR and syngo WARP images of the agar vial (Figure 1b). The MTF values at 0.25 cycle/mm in all sequences were calculated from SI profiles. These measurements were performed with the ImageJ plugin Plot Profile.

The CNR was calculated between the two vials (pure water and Gadolium 0.4 mM) and the background air according to the following equation:

$${\displaystyle CNR=\frac{\mathbf{M}\mathbf{e}\mathbf{a}\mathbf{n}\mathbf{P}\mathbf{W}-\mathbf{M}\mathbf{e}\mathbf{a}\mathbf{n}\mathbf{G}\mathbf{D}}{\mathbf{S}\mathbf{D}}}$$

where Mean PW is the signal intensity of the ROI placed in the centre of the the vial containing pure water, Mean GD is the signal intensity of the ROI placed in the centre of the the vial containing Gadolium 0.4 mM, and SD stands for the standard deviation of background air.¹⁵ To quantify CNR effects, all sequences were repeated three times and CNR in each sequence was expressed as mean ± standard deviation.

Statistical analysis

One-way analysis of variance (ANOVA) and a Tukey–Kramer multiple comparison test were used for each material to detect significant differences among the white, black, and overall artefact volumes in all sequences. These tests were also used to detect significant differences among the artefact volumes of materials in the standard STIR sequences. Furthermore, a paired t-test was used to detect significant differences in the reduction ratios between white and black artefacts of every syngo WARP sequence and between the axial and sagittal planes of every syngo WARP sequence in each material. A p-value of less than 0.05 was considered to indicate statistical significance. All analyses were performed with the statistical software package R 2.9.2 for Macintosh.¹⁶

Results

Figure 2 and Table 3 show the three types of artefact volumes and the reduction ratios in the standard STIR and syngo WARP sequences with several different parameter settings. In the standard STIR sequence, the overall artefact volume of Co–Cr was markedly larger than those of Ni–Cr and Ti, regardless of the planes (p < 0.05).

Figure 2.

Bar charts of artefact volumes in the STIR sequences with several different parameter settings in the axial and sagittal planes. Sequence numbers in the y-axis correspond to the numbers in Table 2. Each bar shows the white, black, and overall artefact volumes for each sequence. a: Co-Cr. b: Ni-Cr. c: Ti. Each bar consists of white and black portions, which represent white and black artefact volumes, respectively. Asterisks (*) indicate significant differences with respect to sequence 7, which had the maximum reduction effect. STIR, Short tau inversion recovery.

Table 3.

Reduction ratios (%) for Co–Cr

Sequence	Axial plane			Sagittal plane
	White artefacts	Black artefacts	Overall artefacts	White artefacts	Black artefacts	Overall artefacts
1	−	−	−	−	−	−
2	−17.37 ± 4.27	1.34 ± 0.93a	−5.96 ± 2.11b	−18.25 ± 4.67	6.68 ± 1.04a	−3.9 ± 1.85
3	−5.35 ± 5.65	8.5 ± 1.42a	3.14 ± 3.10	−10.13 ± 8.30	11.60 ± 3.04a	2.43 ± 4.70
4	6.87 ± 3.98	15.73 ± 1.79a	12.30 ± 1.77b	−2.85 ± 4.71	14.27 ± 1.26a	7.14 ± 2.05
5	5.07 ± 4.91	22.02 ± 1.24a	15.46 ± 2.25b	2.21 ± 6.22	20.39 ± 1.83a	12.73 ± 3.03
6	11.56 ± 4.07	25.52 ± 1.24a	20.12 ± 1.66b	5.47 ± 7.89	22.26 ± 1.94a	15.23 ± 3.89
7	16.41 ± 2.05	27.85 ± 1.05a	23.42 ± 1.57b	7.24 ± 6.21	23.34 ± 4.11a	16.65 ± 2.65
8	−25.96 ± 2.32	−21.41 ± 1.96a	−23.16 ± 1.92b	−33.62 ± 8.29	−17.65 ± 4.12a	−24.2 ± 3.38
9	4.27 ± 3.77	−10.10 ± 2.04a	−7.82 ± 2.50	−22.33 ± 6.76	−6.45 ± 2.39a	−13.1 ± 2.89
10	10.67 ± 7.50	2.18 ± 2.26	5.48 ± 4.87b	−13.54 ± 8.59	0.21 ± 3.77a	−5.46 ± 5.05
11	−15.30 ± 12.1	−4.43 ± 3.50a	−8.69 ± 5.14	−20.77 ± 7.73	−0.80 ± 2.29a	−9.10 ± 2.69
12	−2.23 ± 3.90	0.94 ± 1.44a	−0.26 ± 1.65	−15.41 ± 9.08	3.16 ± 1.69a	−4.51 ± 3.40
13	5.04 ± 5.08	5.82 ± 1.42a	5.53 ± 2.24	−11.62 ± 9.40	5.65 ± 2.00a	−1.48 ± 4.30
14	−3.82 ± 3.20	4.16 ± 1.87a	1.09 ± 1.36	−11.38 ± 8.22	5.94 ± 2.08a	−1.21 ± 2.98
15	2.97 ± 4.36	7.87 ± 1.24a	5.99 ± 2.01	−8.80 ± 9.23	8.81 ± 1.99a	1.54 ± 3.90
16	6.66 ± 3.64	10.34 ± 1.77a	8.93 ± 1.54	−6.81 ± 9.94	10.28 ± 2.26a	3.23 ± 4.75

The maximum reduction ratios are shown in sequence 7 (bold face).

^aSignificant differences were detected between the reduction ratios of white artefacts and black artefacts in each sequence.

^bSignificant differences were detected between the reduction ratios of the overall artefacts in the axial and sagittal planes in each sequence.

The overall artefact volumes of all materials tended to reduce with increasing receiver bandwidth (rBW) and VAT in both planes (Figure 2). However, the highest rBW setting of 930 Hz/pixel at a matrix of 256 × 256 did not record the best reduction. In each material, the sequence showing the minimum overall artefact volume was syngo WARP with a matrix of 384 × 384, rBW of 620 Hz/pixel, and VAT of 100 %, and this volume was significantly smaller than the volumes in many other sequences, regardless of the planes (Supplementary Material 1).

Regarding the influence of different spatial planes (axial and sagittal) on the overall artefact reduction, our results showed that in every sequence, the effect in the axial plane tended to be higher than that of the sagittal plane in Ni–Cr and Ti (Tables 3–5).

Table 4.

Reduction ratios (%) for Ni–Cr

Sequence	Axial plane			Sagittal plane
	White artefacts	Black artefacts	Overall artefacts	White artefacts	Black artefact	Overall artefact
1	−	−	−	−	−	−
2	−6.58 ± 3.25	1.19 ± 1.96a	−2.3 ± 2.25b	−8.32 ± 3.47	−1.08 ± 2.00a	−4.53 ± 2.05
3	4.64 ± 5.06	15.51 ± 3.06a	10.61 ± 3.30b	0.27 ± 10.1	5.68 ± 7.00a	3.10 ± 8.23
4	21.62 ± 4.30	27.83 ± 2.05a	25.06 ± 2.64b	5.17 ± 7.06	11.75 ± 4.26a	8.67 ± 5.05
5	16.26 ± 4.95	34.44 ± 3.44a	26.24 ± 3.65b	11.93 ± 8.64	20.15 ± 4.19a	16.29 ± 5.72
6	23.82 ± 6.43	39.85 ± 4.41a	32.60 ± 4.94b	14.46 ± 10.5	24.93 ± 4.21a	20.00 ± 6.67
7	29.97 ± 4.09	44.17 ± 5.01a	37.76 ± 3.24b	14.28 ± 7.36	25.41 ± 4.58a	20.28 ± 5.11
8	−12.52 ± 3.89	−44.18 ± 3.77a	−29.88 ± 2.41b	−36.94 ± 8.18	−44.61 ± 4.07a	−40.71 ± 5.22
9	4.66 ± 5.12	−19.07 ± 4.08a	−8.38 ± 3.22b	−27.16 ± 8.57	31.96 ± 4.94a	−29.36 ± 5.74
10	23.57 ± 6.40	6.10 ± 6.49a	14.02 ± 5.14b	−23.72 ± 10.3	−22.15 ± 11.0a	−22.74 ± 10.04
11	1.08 ± 3.11	−17.78 ± 3.21a	−9.29 ± 2.13	−16.62 ± 7.98	−21.93 ± 10.3a	−18.73 ± 6.83
12	9.18 ± 4.07	−4.34 ± 4.99a	1.72 ± 3.48b	−13.04 ± 9.01	−13.19 ± 5.53a	−12.94 ± 6.60
13	18.41 ± 4.98	5.1 ± 6.73a	11.05 ± 5.34b	−10.75 ± 11.0	−9.00 ± 6.18a	−9.63 ± 8.08
14	10.21 ± 3.50	0.43 ± 3.65a	4.83 ± 3.18	−4.54 ± 9.37	−6.14 ± 5.46a	−5.10 ± 6.27
15	17.30 ± 4.09	9.57 ± 6.02a	13.04 ± 4.71b	−1.36 ± 10.3	−0.14 ± 6.24a	−0.55 ± 7.84
16	24.19 ± 5.99	13.97 ± 8.82a	18.54 ± 7.46b	0.29 ± 12.3	0.88 ± 6.62a	0.84 ± 9.03

The maximum reduction ratios are shown in sequence 7 (bold face).

^aSignificant differences were detected between the reduction ratios of white artefacts and black artefacts in each sequence.

^bSignificant differences were detected between the reduction ratios of the overall artefacts in the axial and sagittal planes in each sequence.

Table 5.

Reduction ratios (%) for Ti

Sequence	Axial plane			Sagittal plane
	White artefacts	Black artefacts	Overall artefacts	White artefacts	Black artefacts	Overall artefacts
1	−	−	−	−	−	−
2	−4.15 ± 3.94	0.71 ± 1.80a	−1.53 ± 2.33b	−6.39 ± 4.70	−0.49 ± 1.86a	−3.40 ± 3..0
3	6.47 ± 5.49	14.45 ± 5.61a	10.76 ± 4.87b	1.18 ± 11.7	6.29 ± 8.67a	3.76 ± 10.08
4	20.4 ± 3.05	27.44 ± 1.86a	24.19 ± 2.15b	6.11 ± 5.33	11.11 ± 3.44a	8.64 ± 3.79
5	16.6 ± 3.49	35.24 ± 3.52a	26.68 ± 3.32b	13.3 ± 6.98	21.57 ± 3.87a	17.48 ± 4.90
6	23.2 ± 5.40	41.50 ± 4.53a	33.09 ± 4.92b	15.1 ± 9.33	25.37 ± 5.17a	20.31 ± 6.90
7	27.7 ± 5.67	44.96 ± 6.76a	37.01 ± 5.21b	15.77 ± 10.8	26.33 ± 8.21a	21.19 ± 9.08
8	−11.91 ± 6.78	−42.91 ± 7.38a	−28.64 ± 6.29b	−27.96 ± 12.8	−41.30 ± 8.93a	−34.32 ± 10.25
9	3.66 ± 7.72	−16.14 ± 5.82a	−7.02 ± 5.43b	−20.01 ± 12.5	−29.61 ± 8.05a	−24.41 ± 9.67
10	19.65 ± 5.17	6.74 ± 9.40a	12.72 ± 7.00b	−17.35 ± 15.0	−22.42 ± 14.4a	−19.65 ± 14.28
11	1.37 ± 4.58	−15.08 ± 5.21a	−7.49 ± 3.77b	−9.21 ± 10.5	−18.69 ± 9.5a	−13.48 ± 7.95
12	8.48 ± 4.76	−1.53 ± 6.05a	3.11 ± 2.80b	−6.16 ± 10.3	−12.17 ± 5.65a	−8.92 ± 7.71
13	14.46 ± 3.64	6.63 ± 8.27a	10.28 ± 4.89b	−4.63 ± 12.5	−8.66 ± 8.23a	−6.41 ± 10.17
14	11.14 ± 3.63	2.63 ± 4.71a	6.58 ± 2.91b	2.95 ± 10.5	−4.14 ± 5.39a	−0.28 ± 7.47
15	16.57 ± 2.96	11.28 ± 6.58a	13.75 ± 3.84b	5.39 ± 10.3	0.08 ± 6.57a	2.99 ± 7.95
16	20.08 ± 4.70	15.85 ± 10.2a	17.85 ± 6.75b	6.37 ± 12.5	1.31 ± 7.69a	4.08 ± 9.92

The maximum reduction ratios are shown in sequence 7 (bold face).

^aSignificant differences were detected between the reduction ratios of white artefacts and black artefacts in each sequence.

^bSignificant differences were detected between the reduction ratios of the overall artefacts in the axial and sagittal planes in each sequence.

Regarding the reduction effect of black and white artefacts based on the reduction ratios, the effect for black artefacts was more marked than that for white artefacts in every sequence regardless of the material (Tables 3–5). In most of the sequences, a significant difference was detected between the reduction ratio of white artefacts and that of black artefacts in each material. The best reduction compared with standard STIR in all materials was achieved with syngo WARP with a matrix of 384 × 384, rBW of 620 Hz/pixel, and VAT of 100 % in the axial plane. The greatest reduction effects, with rates up to 23–38 %, were obtained in comparisons with the standard sequences (Tables 3–5 and Figure 3).

Figure 3.

Phantom images in the standard STIR and the best reduction syngo WARP sequences. Axial images (with A–P and R–L readout direction) and sagittal images (A–P and H–F readout direction) are shown in the upper and lower rows, respectively. In all readout directions, the images with the best reduction sequence (WARP-M384-rBW620-VAT100) show a great improvement in the overall artefact volume. In sagittal images of the standard sequence, the presence of metallic alloys has caused a serious distortion (most obvious in Co–Cr) which was partly corrected by the best reduction sequence. STIR, short tau inversion recovery.

Evaluations of the effects of syngo WARP on the image quality are summarized in Figures 4 and 5, and Table 6. Images obtained with a matrix of 384 × 384 show less loss of sharpness across the pin hole section with increasing VAT in comparison with the images obtained with a matrix of 256 × 256. An increase in VAT (VAT50, VAT100) produced images with more blurring (Figure 4a), and markedly decreasing amplitudes were evident on the plot profiles (Figure 4b and c). Images obtained at a matrix of 384 × 384 with a rBW of 620 Hz/pixel and VAT100 (Sequence No. 7), which was the optimal parameter settings for the reduced artefact volume on the syngo WARP images, showed only a little loss of sharpness in comparison with the image of the standard STIR sequence (Figure 4a). However, the plot profile for the optimal parameter settings showed decreasing amplitudes in comparison with those of standard STIR (Figure 4b) and MTF value of the optimal parameter settings decreased by 30 % compared with that of standard STIR (Table 6). Images of a matrix of 384 × 384 showed lower CNR in comparison with that of a matrix of 256 × 256. CNR of images obtained with VAT100 showed no remarkable difference in comparison with that of images with VAT0 and VAT50 (Figure 5). Images acquired from the optimal parameter settings showed approximately 22 % loss of CNR in comparison with the image of the standard STIR sequence.

Figure 4.

Evaluation of the blurring effects of syngo WARP. (a) Standard STIR and syngo WARP images of the pin hole section with R–L readout direction. Marked blurring was evident on the syngo WARP images with a lower matrix, lower rBW, and higher VAT. (b) The graphs show the signal intensity profiles in the pin hole section from standard STIR and syngo WARP images. Sequence numbers in the legend box correspond to the numbers in Table 2.

Figure 5.

Evaluation of CNR in syngo WARP. Sequence numbers in the y-axis correspond to the numbers in Table 2. CNR, contrast-to-noise ratio.

Table 6.

MTF values of standard and WARP sequences

Number	MTF value at 0.25 cycle/mm
1	Standard	0.80
2	M384-WARP-rBW296-VAT0	0.73
3	M384-WARP-rBW296-VAT50	0.64
4	M384-WARP-rBW296-VAT100	0.38
5	M384-WARP-rBW620-VAT0	0.67
6	M384-WARP-rBW620-VAT50	0.60
7	M384-WARP-rBW620-VAT100	0.56
8	M256-WARP- rBW296-VAT0	0.47
9	M256-WARP- rBW296-VAT50	0.23
10	M256-WARP- rBW296-VAT100	0.02
11	M256-WARP- rBW610-VAT0	0.44
12	M256-WARP- rBW610-VAT50	0.40
13	M256-WARP- rBW610-VAT100	0.19
14	M256-WARP- rBW930-VAT0	0.44
15	M256-WARP- rBW930-VAT50	0.38
16	M256-WARP- rBW930-VAT100	0.28

MTF, modulation transfer function.

Discussion

In the current phantom study, we investigated the effects of the syngo WARP technique on reducing metal artefacts caused by metallic dental materials. In previous studies, various image acquisition and evaluation methods have been applied to study the effects of several advanced MARS techniques for metal artefact reduction.⁷ In this study, an image acquisition method based on ASTM F2119-07 was used and the artefact volumes were quantitatively measured. Additionally, to evaluate metal-induced artefacts as precisely as possible, the parallel imaging or undersampling techniques, which are commonly used to shorten the scanning time, were not applied in this study. Regarding the quantitative measurement of artefact volumes, black (signal loss) and white (signal pile-up) artefact volumes were measured separately.¹⁶ Therefore, reliable artefact volume data could be obtained and the effects of syngo WARP on artefact reduction could be evaluated more accurately. In this study, this technique was combined with STIR, although this technique can also be applied to other (turbo) spin-echo sequences. STIR, which is less susceptible to the inhomogeneous magnetic field, is performed as an alternative to T₂ weighted imaging with fat suppression in cases of adequate fat suppression caused by metal artefacts from large metallic devices in the maxillofacial region.¹⁰ However, the disadvantage of STIR is that its signal-to-noise ratio (SNR) is lower than that of other (turbo) spin-echo sequences. Therefore, the application of STIR combined with WARP would require immediate and strict adjustment.

In dental practice, metallic materials are commonly used in the fabrication of dental restorations, crowns, dentures, orthodontic devices, and dental implants. The presence of metallic materials can lead to a major change in the static magnetic field, resulting in the presence of artefacts on MR images.^17,18 According to Murakami et al,¹⁶ Co–Cr, Ni–Cr, and Ti alloys can produce large artefacts. It is also recognized that artefacts from these metallic materials can hamper image interpretation in clinical head and neck MRIs.^17,18 In this study, we investigated three common types of metallic dental materials and found that the overall artefact volume of Co–Cr was the largest, followed by those of Ni–Cr and Ti, in both planes. Among the metallic materials used in this study, the magnetic susceptibility of Cr–Co is the highest, followed by Ni–Cr, while that of Ti is extremely low.¹⁸ These different magnetic susceptibilities may have led to the differences in artefact size. syngo WARP, commercialized by Siemens, is a dedicated MARS technique in combination with VAT.⁵ In this study, the relationship between the reduction effects and the imaging parameters, including the BW of the signal readout (i.e. rBW), was evaluated quantitatively with the use of metallic dental materials. The overall artefact volume could be reduced by using a higher rBW. The maximum reduction effects were obtained at an rBW of 620 Hz/pixel, not 930 Hz/pixel. This result indicated that setting a higher rBW with a smaller matrix did not necessarily heighten the artefact reduction. However, the matrix for the clinical settings should be determined taking into consideration the advantages of a smaller matrix, such as a higher SNR, a lower specific absorption rate (SAR), and a shorter acquisition time.

Another characteristic of syngo WARP is VAT, which applies a replay of the section selection gradient during the signal readout, leading to a cancellation of the signal displacement in the readout direction for off-resonance spins. Therefore, an effect on artefact reduction can be obtained.⁶ To our knowledge, almost all previous studies applied VAT of 100 % (VAT100) to reduce metal artefacts without mentioning the effect of other values of VAT.^7,10,19 In the present study, we used sequences without VAT (VAT0), and with VAT (VAT50, VAT100) to determine whether increasing the value of VAT could gain more benefit in artefact reduction. The results showed that a higher VAT leads to a greater reduction in artefacts with not much change in CNR compared with images of lower VAT (Figures 2 and 5). For example, in an axial STIR sequence with the same parameter settings (WARP, rBW 620 Hz/pixel, and matrix of 384 × 384), VAT100 can reduce artefact volume by an average of 10 and 5 % in comparison with VAT0 and VAT50, respectively. However, VAT may lead to a blurring effect along the readout direction, and may sometimes compromise image interpretations. In this study, there were blurring effects that might hamper image interpretation on syngo WARP images obtained at a 256 × 256 matrix with low rBW and VAT50 or 100, and markedly decreasing MTF values were shown in Table 6. According to Butts et al,²⁰ image blurring due to VAT can be reduced by shortening the readout time. In this study, image blurring resulting from VAT may have been offset by the parameter settings of a higher rBW and a relatively high matrix of 384 × 384.

Black artefacts result from signal loss caused by T₂ dephasing and encoding errors in the position. White artefacts result from a signal pile-up caused by an encoding error in the position.¹³ In syngo WARP, the increasing BW of the excitation pulse and signal readout can reduce T₂ dephasing mediated by the reduced echo spacing, resulting in a reduction in signal loss. Therefore, it was considered that black artefacts might be improved by a reduction in signal loss. In other words, the reduction effect for black artefacts might be more effective than that for white artefacts, as shown in this study. However, VAT is thought to exhibit the same reduction effects for black and white artefacts because this technique is a cancellation of the signal displacement in the readout direction.

The effects of artefact reduction in the axial plane tended to be higher than that in the sagittal plane in every syngo WARP sequence excluding Co–Cr. In this study, all cylindrical materials were located with the long axis parallel to the main magnetic field (B₀). The local magnetic field occurred near the metallic material. In this study, the magnetic field variation in the direction of B₀ was higher than that of the vertical direction, and larger resonant frequency changes were caused in the direction B₀. Therefore, the artefact reduction in the sagittal plane parallel to B₀ may be greater than that in the axial plane. In Co–Cr, however, there were no significant differences in most of the syngo WARP sequences, possibly because the artefact size caused by Co–Cr might be larger than the ROI in our study.

In this current study, we found that a syngo WARP STIR at a matrix of 384 × 384 with a high rBW of 620 Hz/pixel and VAT100 in the axial plane obtained the best metal artefact reduction compared with standard STIR (Ti: 32–42 %, Ni–Cr: 34–37 %, Co–Cr: 21–25%). This optimal parameter setting may not be applicable in a clinical setting because SAR and acquisition time were not determined in this study. However, the characteristics of WARP–VAT for metal artefact reduction might be obtained as new knowledge from quantitative evaluations.

There were several limitations in this study. First, we did not obtain coronal sequences and did not apply a range of different values of VAT due to the limited experimental time. The total acquisition time for one metal material and agar vial set was almost 8 h (16 sequences × 4 phase directions × 2 vials × 1 direction). Murakami et al found that there were no obvious differences between sagittal images and coronal images.¹⁶ According to ASTM F2119-07, cylindrically symmetrical devices may be tested parallel to the static field and in just one direction perpendicular to the static field. Sagittal images, relative to the static field direction, should be acquired in all cases. Therefore, we obtained axial and sagittal images. To evaluate metal-induced artefacts as precisely as possible, the parallel imaging or undersampling techniques, which are commonly used to shorten the scanning time, were not applied in this study. Second, we did not use more advanced techniques such as SEMAC or MAVRIC which have been proven to provide a strong artefact reduction effect in orthopedic imaging.^21–23 SEMAC and MAVRIC require special software which is not available in many institutions. Finally, the phantom conditions in this study might be quite different from clinical situations in the maxillofacial region. In clinical situations, the size and shape of artefacts may depend on the shape, position, and orientation of the metal, and the number of objects.^9,16 However, the use of a simple box-shaped phantom and cylindrical metal samples was appropriate for the quantitative evaluation based on ASTM F2119-07, and our results could be used as a baseline in a clinical setting.

In conclusion, the present study proposed MR pulse sequence parameters for reducing artefacts arising from metallic dental materials at three tesla. By using syngo WARP with optimal parameter settings (a high rBW of 620 Hz/pixel, a matrix of 384 × 384 and VAT of 100 %), a significant reduction in metal artefacts can be obtained in comparison with a standard STIR sequence. However, blurring effect and the loss of CNR from using syngo WARP should be considered carefully to balance the artefact reduction effect and the image quality.