Skip to main content
. 2023 Mar 16;52(5):20230024. doi: 10.1259/dmfr.20230024

Table 2.

Characteristics of included studies

Author Year & Country Study design Sample Index test Reference standard MRI system MRI sequence parameters Landmarks and measurements Observers Statistical tool Findings/ outcome metrics Conclusion
Eley et al., 201322
UK		 In vivo, prospective Three patients with a mean age of 37 years (range, 21–60 years) and five patients with a mean age of 19 years (range, 16–27 years) 2D MRI cephalogram (Black bone image, T1WI and T2WI of midsagittal plane) Conventional 2D LCR 1.5T MRI system TR: 8.6 ms, TE: 4.2 ms, Scan FOV: 24 cm, Phase encode: 256, Frequency encode: 256, Receive bandwidth: 31.25, NEX: 2, ETL: 1 11 landmarks (nine skeletal and two dental); 17 measurements (11 linear and six angular) One observer repeated the measurements ten times (time interval and blinding not reported) Coefficient of variation to assess repeatability between measures

  1. Discrepancy between black bone images and LCR was 2.1° ± 1.7° for the angles and 2.8 ± 2.7 mm for the distances

  2. Discrepancy between T1W and LCR was 5.0° ± 2.9° for the angles and 3.3 ± 3.2 mm for the distances

  3. Among black bone images, the greatest variability occurred with dental landmarks

 Black-bone MRI sequence showed comparable accuracy to that of LCR in landmark identification
Heil et al., 201728
Germany		 In vitro and in vivo, prospective

  1. Phantom

  2. 20 patients with a mean age of 13.95 ± 5.34 years (Range, 8–26 years)

 2D MRI cephalogram (Postprocessed) Conventional 2D LCR and phantom measurements 3T MRI system with a 16-channel multipurpose coil using T1 weighted, isotropic SPACE sequence TE: 26 ms, TR: 800 ms, bandwidth: 501 Hz/pixel, number of averages: 2, ETL: 63, FOV: 175× 175 mm, acquisition matrix: 256 × 256, voxel size: 0.68× 0.68 mm ×0.68 mm, number of sections: 192, time of acquisition: 6:59 min 18 landmarks (10 skeletal and dental; 10 midsagittal and eight bilateral);
24 measurements (10 linear and 14 angular)		 Two independent observers (one radiologist and one orthodontist) analysed twice in 4 weeks interval (blinding not reported)

  1. Intra- and inter- examiner reliability using ICC

  2. Bland-Altman analysis to assess agreement between two modalities

  3. Equivalence testing (two one-sided test)

  1. Measurements of MRI sequence were equivalent to phantom values

  2. Statistical equivalence between MRI and LCR observed for all measurements except interincisal angle.

  3. Highest bias (−1.33) and widest 95% limits of agreement (−7.22, 4.56) was observed for interincisal angle

  4. Excellent intraobserver and interobservers reliability for both the modalities

 Measurements from LCR derived from high-resolution isotropic MRI datasets has high concordance to the corresponding measurements on conventional LCR.
Juerchott et al., 201824
Germany		 In vitro and in vivo, prospective

  1. Phantom

  2. Three male patients (27 years; 33 years, and 31 years)

 T1WI 3D MRI dataset (Patient data in five different head positions and phantom data in three different head positions) Phantom to validate accuracy 3T MRI system using a 16-channel multipurpose coil with high-resolution T 1-weighted 3D MSVAT-SPACE prototype sequence applied TE: 5.8 ms, TR: 800 ms, bandwidth: 625 Hz/pixel, number of averages: 1, ETL: 100, FOV: 171× 171 mm, acquisition matrix: 320 × 320, voxel size: 0.53× 0.53 mm ×0.53 mm, number of sections: 256, time of acquisition: 7:01 min 27 landmarks (skeletal and dental); 45 measurements (26 linear and 19 angular) One radiologist with 5 year experience (repeated measures and blinding not reported)

  1. Equivalence testing (two one-sided test)

  2. Bland-Altman analysis to assess level of agreement between MRI phantom measurements and true values

  3. One-way ANOVA with Green house -Geisser correction for reproducibility of in-vivo cephalometric measurements

  1. MRI-based phantom measurements showed statistical equivalence and an excellent agreement (bias ranges: −0.090 to 0.044°, −0.220 to 0.241 mm)

  2. In vivo cephalometric analysis was highly reproducible without statistical differences for all angles and distances (average ranges: 0.88°/0.87 mm).

 Accurate and reproducible 3D cephalometric analysis can be performed using MRI
Jency et al., 201932
India		 In vivo, prospective 11 patients (age 18–30 years) 2D MRI cephalogram (Black bone image, T1WI and T2WI of midsagittal plane) Conventional 2D LCR 1.5T MRI system TR: 11 ms, TE: 4.20 ms, FOV: 220 mm Hard tissue and soft tissue landmarks;
18 measurements (12 linear and six angular)		 One radiologist repeated measurements ten times (time interval and blinding not reported) 1. Covariance between LCR and MRI images
2.Paired T-test between mean values of LCR and MRI measurements		 The ease of landmark identification was difficult on T2 weighted images, but on black bone images, it was comparable to LCR. Black bone MRI sequence can be an effective non-ionizing imaging modality over conventional methods.
Maspero et al., 201923
Italy		 In vivo, retrospective 18 subjects (four male; 14 female) with a mean age of 37.8 ± 10.2 years T2WI 3D MRI dataset CBCT (Acquisition parameters- 4 mm slice thickness, 170 × 230 mm FOV, 20 sec scan time, 0.49 × 0.49×0.5 mm voxel size, 120 kVp, and 3–8 mA) 3T MRI system TR: 2500 ms, TE: 280 ms, NEX: 1, ETL: 65, bandwidth: 255 Hz/pixel, flip angle: 90°, FOV: 240 × 240×180 mm, voxel size: 0.49 × 0.49×0.50 mm, section thickness: 0.49 mm, and time of acquisition: 5:27 min 18 landmarks (10 midsagittal and four lateral points)
24 measurements (11 linear and 13 angular)		 Two independent orthodontist performed the analysis twice in 3 weeks interval 1. Intra- and interobserver agreement by ICC

  1. Coefficient of variation (CV) to compare the precision of measurements

  2. Two-sample t-test to evaluate the differences in the mean CV between CBCT and MRI

  3. Bland-Altman analysis to assess the agreement between the two modalities with 95% limits of agreement

  4. Paired t-test to compare CBCT and MRI measurements

 1.Both CBCT and MRI showed good reliability, with mean intraobserver ICCs of 0.977/0.971 for CBCT and 0.881/
0.912 for MRI.

  1. Average interobserver ICCs: 0.965 for CBCT and 0.833 for MRI.

  2. A similar range of agreement between the two modalities with bias range −0.25 to 0.66 mm and −0.41 to 0.54°

 Cephalometric measurements using 3T-MRI possess adequate reliability and repeatability, and satisfying agreement with CBCT measurements.
Juerchott et al., 202026
Germany		 In vivo, prospective 16 patients (8 males; 8 females) with a mean age of 23.3 ± 7.5 years (range, 14–40 years) T1WI 3D MRI dataset None 3T MRI system using a dedicated 15-channel dental surface coil and T1-weighted 3D MSVAT-SPACE prototype sequence applied TE: 5.8 ms, TR: 800 ms, bandwidth: 625 Hz/pixel, number of averages: 1, ETL: 100, FOV: 171× 171 mm, acquisition matrix: 320 × 320, voxel size: 0.53× 0.53 mm ×0.53 mm, number of sections: 256, time of acquisition: 7:01 min 42 landmarks [28 skeletal (14 midsagittal and 14 bilateral) and 16 dental)] Two independent radiologists identified landmarks twice after more than 4 weeks interval (Blinding done)

  1. Intra- and interrater agreement by mean measurement errors and ICC

  2. Measurement error calculated as Euclidean distances

  1. Intra- and interrater ICCs were consistently higher than 0.9.

  2. Intrarater comparisons showed mean measurement differences (ranges) of 0.87 mm (0.41 to 1.63) for rater I and 0.94 mm (0.49 to 1.28) for rater II.

  3. Average interrater difference was 1.10 mm (0.52 to 2.29).

 High-resolution 3D MRI enables reliable determination of 3D cephalometric landmarks with high intra- and interrater reliability
Juerchott et al., 202025
Germany		 In vivo, prospective 12 patients (8 males, 4 females) with a mean of 26.1 ± 6.6 years (range, 17–40 years) T1WI 3D MRI dataset CBCT (Acquisition parameters- tube voltage: 98 kV, tube current: 5 mAs, scanning time: 14 s, FOV: 150 × 150 mm, and isotopic voxel size: 0.25 mm) 3T MRI system using a dedicated 15-channel dental surface coil and T 1-weighted 3D MSVAT-SPACE prototype sequence applied TE: 5.8 ms, TR: 800 ms, bandwidth: 625 Hz/pixel, number of averages: 1, ETL: 100, FOV: 171× 171 mm, acquisition matrix: 320 × 320, voxel size: 0.53× 0.53 mm ×0.53 mm, number of sections: 256, time of acquisition: 7:01 min 27 landmarks (skeletal and dental)
35 measurements
(18 linear and 17 angular)		 Two independent radiologists performed the analysis twice after more than 4 weeks interval (Blinding done) Calculation of Euclidean distances, ICC, Bland- Altman analysis, and equivalence testing (linear mixed effects model) with a predefined equivalence margin of ±1°/1 mm.

  1. Analysis of reliability for CBCT vs MRI (intrarater I/ intrarater II/ interrater) revealed Euclidean distances of 0.86/0.86/0.98 mm vs 0.93/0.99/1.10 mm for landmarks, ICCs of 0.990/0.980/0.986 vs 0.982/0.978/0.980 for angles, and ICCs of 0.992/0.988/0.989 vs 0.991/0.985/0.988 for distances.

  2. High levels of agreement with bias value of 0.03° (− 1.49 to 1.54) for angles and 0.02 mm (− 1.44 to 1.47) for distances.

  3. The mean values of CBCT and MRI measurements were equivalent.

 3T MRI enables reliable 3D cephalometric analysis and excellent agreement with CBCT measurements
Marz et al., 202127
Germany		 In vitro, prospective Three human cadaver head preparations T1WI 3D MRI dataset with no post-processing Conventional 2D LCR 3T MR scanner and a standard 20-channel head-neck-coil was used and T1-weighted multislice Turbo Spin Echo MR images was obtained TE: 9.9 ms, TR: 300 ms, turbo factor: 3, in-plane spatial resolution: 0.88 × 0.88 mm2, slice thickness:1.5 mm, 120 slices, total acquisition time: 19.5 min, three averages 19 skeletal and dental landmarks; 13 angular measurements Five independent orthodontists (Blinding done and repetitive measures not reported)

  1. ICC for agreement between the five raters

  2. Bland-Altman plots to visualise interrater agreement

  3. Linear mixed-effects model for angle specific differences and equivalence testing (two one-sided test) including Bonferroni-Holm correction

  1. The interrater reliability was high for all angles (ICC≥0.74)

  2. Differences between LCR and MRI measurements ranged between –0.91° (–1.88 to 0.07) and 0.97° (–0.63 to 2.57) and were equivalent with respect to a margin of [–2°, 2°] except for lower incisor inclination.

 A reliable method for cephalometric analysis of a 3D MRI dataset with semi-automatic dataset orientation was established
Abkai C et al., 202121
Germany		 In vitro, prospective One patient (male) 7 MRI cephalometric projections (MCPs) with various scan parameters and no post-processing Conventional 2D LCR 3T MRI system with eight-channel head coil Pixel size 0.39 × 0.39 mm or 0.2 × 0.2 mm, FOV 293.3 × 293.3 mm or 300 × 300 mm, scan time ranging from 5 to 154 sec, TE: 358 to 360 μs, TR: 4.2 to 16.9 ms, pixel bandwidth: 816, 517 or 259 Hz and NEX: one or 2 14 skeletal and dental landmarks; 10 angular measurements 40 orthodontists with at least 15 years' experience (randomisation of radiographs and blinding done;repetitive measures not reported) Levene’s test for evaluating homogeneity of variance and two-tailed t-tests

  1. Mean relative distances were 2.4 to 2.7 mm in MCPs and 1.6 mm in LCR, demonstrating the accuracy and level of agreement of assessors

  2. MCP showed on average deviation of 1.2˚ ±1.1˚ in comparison to LCR. Larger differences were found for gonial and interincisal angles.

  3. Images with higher resolution and contrast (154 s) showed mean angular deviations up to 0.9˚ and lower resolution images (5–16 s) showed mean angular deviation up to 1.6˚ in comparison to LCR.

 MCPs can be acquired much faster, and this study demonstrated the potential of this new method.

ANOVA, Analysis of Variance; CBCT, Cone Beam Computed Tomography; 2D, two-dimensional; 3D, Three dimensional; ETL, Echo train length; FH, Frankfort horizontal; FOV, Field of View; Hz, Hertz; ICC, Intraclass Coefficient; LCR, Lateral Cephalometric Radiograph; MRI, Magnetic Resonance Imaging; MSVAT-SPACE, Multiple-Slab acquisition with View Angle Tilting gradient based on Sampling Perfection with Application optimised Contrast using different flip angle Evolution; NEX, Number of excitations; SPACE, Sampling Perfection with Application optimised Contrast using different flip angle Evolution; T, Tesla; TE, echo time; TR, repetition time; T1WI, T1 Weighted Images; T2WI, T2 Weighted Images; cm, centimetres; kV, tube voltage; kVp, kilovoltage Peak; mA, milliampere; min, minutes; mm, millimetres; ms, milliseconds; s, second; μs, microseconds.