Chairs:
bence tamás szabó
ruben pauwels
117: INTERFERENCE OF TITANIUM AND ZIRCONIA IMPLANTS ON DENTAL-DEDICATED MAGNETIC RESONANCE IMAGE QUALITY: EX VIVO AND IN VIVO ASSESSMENT
R. Spin Neto1, K. Johannsen1, J. Christensen1, L. Hauge Matzen1, B. Hansen2
1Aarhus University, Department of Dentistry and Oral Health, Aarhus, Denmark, 2Aarhus University, Center of Functionally Integrative Neuroscience, Aarhus, Denmark
Aim: To assess the impact of titanium and zirconia implants on dental-dedicated MR image (ddMRI) quality ex vivo (magnetic field distortion, MFD) and in vivo (artefacts).
Material and methods: ddMRI were acquired (Magnetom Free.Max, 0.55T, Siemens Healthcare, Germany) using a dental-dedicated coil (Rapid-Biomedical, Germany). Ex vivo: three phantoms were manufactured: one with an agar-embedded titanium implant, one with an agar-embedded zirconia implant, and one with agar 1.5% (control). Field-map analysis of images acquired at 0.55T, 1.5T (Magnetom Sola, Siemens Healthineers, Germany) and 3T (Magnetom Lumina, Siemens Healthineers, Germany) was done to illustrate the extent and severity of MFD caused by the implants. In vivo (0.55T only): a splint was designed to serve as implant carrier allowing diverse implant positions (0, 1, 2, or 5 implants of each material). With this, a volunteer was imaged using multiple pulse sequences. Three blinded observers scored the images twice for the presence, severity, and type of artefacts, illustrated by descriptive statistics and inter- and intra-observer reproducibility (kappa statistics).
Results: Ex vivo: titanium produced more severe MFD than zirconia. MFD extent and amplitude increased with field strength (0.55T < 1.5T < 3.0T). In vivo: titanium and zirconia produced artefacts, generally as signal voids in tooth crowns close to implants. Inter- and intra-observer reproducibility ranged from 0.22-0.70, and 0.27-0.57, respectively.
Conclusion: Titanium generated larger MFD than zirconia. For both materials, artefacts were visible mainly in the crown area. Observer reproducibility needs improvement by ddMRI training.
126: AUTOMATIC VIEW PLANNING IN DENTOMAXILLOFACIAL MAGNETIC RESONANCE IMAGING
C. Hayes1, A. Greiser1, M. Keil1, M. Van de Stadt-Lagemaat1, J. Thöne1, L. Lauer1, R. Spin-Neto2, D. Nixdorf3, K. Iyer4
1Siemens Healthineers AG, Erlangen, Germany, 2Aarhus University, Department of Dentistry and Oral Health, Aarhus, Denmark, 3University of Minnesota School of Dentistry, Division of TMD & Orofacial Pain, Minneapolis, United States, 4Siemens Medical Solutions USA, Malvern, United States
Aim: To facilitate ease of use and to enhance efficiency in dental-dedicated MRI scanning procedures, an automatic section planning software was developed for a selection of dentomaxillofacial workflows at 0.55 Tesla.
Material and Methods: A suitable set of anatomical landmarks were annotated in scout images (n=70 subjects) covering the entire dentomaxillofacial region and acquired on a dental-dedicated 0.55 Tesla MRI system (MAGNETOM Free.Max, Siemens Healthineers AG, Germany). An AdaBoost-based machine learning algorithm was trained to detect these landmarks automatically and to calculate standard dental views, thereby allowing for automatic view planning as part of the examination workflow. Section planning was automated for longitudinal, cross-sectional. and axial sextants of the dental arches and for standard views in TMJ imaging and orthodontics. The automated planning performance was evaluated retrospectively in an additional 133 subjects, including instances where dental material produced artefacts, e.g., retainers, titanium implants, and prospectively on a scanner.
Results: Overall, the performance of the automatic planning algorithm was adequate and provided results ca.10 seconds following scout scan completion. Even in subjects with metal prostheses, view planning was often acceptable; in 3 cases the algorithm failed, and manual planning was performed. Quality of landmark detection was rated very good in 88.4%, good in 8.1%, marginal in 1.3% and failed in 2.2%. In some cases automatic planning was suboptimal due to sporadic misdetection of molar roots.
Conclusion: Automated section planning based on landmark detection in dentomaxillofacial MRI scout images is feasible and will enable time-efficient scan planning even with limited operator experience.
154: IN VIVO RELIABILITY AND ACCURACY OF DENTAL MRI FOR IDENTIFICATION OF CEMENTO-ENAMEL-JUNCTION
M. Kranz1, M. Sohani1, A. Juerchott1, S. Heiland1, M. Bendszus1, T. Hilgenfeld1
1Universitätsklinikum Heidelberg, Neuroradiologie, Heidelberg, Germany
Aim: To evaluate the reliability and accuracy of high-resolution dental MRI (dMRI) for in vivo identification of cemento-enamel junction (CEJ) using cone-beam computed tomography (CBCT) as reference imaging modality.
Methods: Three-Tesla dMRI and CBCT were prospectively acquired in 23 participants (mean age: 52,9 ± 10,6 years). A fat-saturated T2 weighted sequence with isotropic resolution of 0,6 mm was used in dMRI. In 319 teeth longitudinal axis for each root (389 roots) were reconstructed in CBCT and dMRI in the mesio-distal (M-D) and vestibulo-oral (V-O) direction and anonymized. Next, distance from CEJ to apex was measured altogether three times by a blinded dentist and a radiologist. Reliability/accuracy were assessed using intraclass correlation coefficients (ICCs)/Bland–Altman analysis.
Results: The reliability of dMRI was excellent and comparable to CBCT (intra-/inter-rater; dMRI vs. CBCT) M-D: 0,963/ 0,914 vs. 0,988/ 0,982; V-O: 0,969/ 0,868 vs. 0,99/ 0,984). Accuracy analysis revealed a mean error (95% confidence interval) of dMRI of 0,019 (-0,178/ 0,215) / 0,0005 (-0,169/ 0,170) cm in the M-D / V-O direction.
Conclusion: The reliability of dMRI for CEJ identification in vivo is comparable to CBCT. Mean accuracy of dMRI was high and slightly better in the V-O than in the M-D direction. Due to no distinction in signal between enamel and dentin in standard MRI sequences, however, a 95% confidence intervall up to 0.2 cm limits the accuracy of dMRI for CEJ identification in vivo.
198: COMPARISON OF 0.55T AND 1.5T MRI IMAGES OF TEMPOROMANDIBULAR JOINTS
B. Groenke1, D. Nixdorf1, A. Greiser2, C. Hayes2, L. Gaalaas1, C. Herman1, S. Kaimal1, E. Moana-Filho1, M. Mulet1, C. Ozutemiz1, R. Spin-Neto3
1University of Minnesota, Minneapolis, United States, 2Siemens Healthineers, Erlangen, Germany, 3Aarhus University, Aarhus, Denmark
Aim: MRI is the gold standard for imaging soft tissues of the temporomandibular joint (TMJ). Our aim was to compare rater agreement and image quality from a dental-dedicated 0.55T scanner with images produced from a conventional medical 1.5T scanner.
Materials and Methods: Six raters evaluated 80 image stacks of TMJs from 5 participants. Images were acquired on a 0.55T scanner (Magnetom Free.Max, Siemens Healthineers, Germany) using a dental-dedicated receive coil and a 1.5T scanner (Magnetom Sola, Siemens Healthineers, Germany) using a head-neck coil. Images included: parasagittal T2-weighted and proton-density with/without fat suppression and paracoronal proton-density. Stacks were presented to raters blinded and randomized. Hard and soft tissue components of the TMJs and image quality parameters (noise, resolution, contrast) were assessed using a 3-point scale (0=unacceptable, 1=acceptable, 2=excellent). No calibration was performed: raters scored images based on professional expertise. Data was pooled across participants, pulse sequences, and side of the body. Intra- and inter-rater kappas were computed. Chi-square tests determined differences between scanners with p≤0.05.
Results: Kappa values for intra-rater agreement were 0.49-0.70 for 0.55T and 0.42-0.71 for 1.5T. Inter-rater kappa values were 0.02-0.29 and 0.02-0.23 for 0.55T and 1.5T respectively. Three raters preferred images from the 0.55T system for visualizing hard and soft tissue anatomy (p<0.05). One rater preferred 0.55T image contrast.
Conclusions: Rater agreement and image quality were not different between the dental-dedicated 0.55T scanner and the conventional medical 1.5T scanner. The 0.55T scanner produced images of comparable diagnostic quality to a 1.5T scanner for evaluation of TMJs.
227: DEEP-LEARNING BASED IMAGE RECONSTRUCTION TO ACCELERATE MRI SCAN TIME
D. Nixdorf1, C. Hayes2, M. Hellinger2, M. Nickel2, L. Gaalaas3, R. Spin-Neto4, A. Greiser2
1Univesiyt of Minneosta, TMD & Orofacial Pain, Minneapolis, United States, 2Siemens Healthineers, Erlangen, Germany, 3University of Minnesota, Oral Medicine & Radiology, Minneapolis, United States, 4Aarhus University, Dentistry and Oral Health, Aarhus, Denmark
Aim: Image reconstruction methods using deep-learning (DL)-based reconstruction can be used to enable shorter MRI scan times. We assessed image quality parameters in MRI volumes reconstructed with and without DL.
Materials and methods: Twenty-seven MRI stacks from 9 healthy volunteers were reviewed by one oral radiologist in a blinded randomized fashion. TMJ images were obtained on a 0.55T scanner (MAGNETOM Free.Max, Siemens Healthineers, Erlangen, Germany) equipped with a dental-dedicated surface coil using proton density-weighted TSE sequence in the parasagittal plane with three different scan techniques: 2:38min duration accelerated with DL reconstruction at 0.24mm in-plane resolution (DL), 4:32min duration with conventional acquisition (CQ) at 0.31mm in-plane resolution (reference with similar quality), and 2:38min duration with conventional acceleration (CD) at 0.24mm in-plane resolution (reference with similar scan duration). Anatomical components of the TMJ and image quality parameters were assessed using a 3-point scale (0=unacceptable, 1=acceptable, 2=excellent). Chi-square tests determined differences between techniques (significance set at ≤0.05).
Results: No difference between DL and CQ were observed when scoring hard and soft tissue anatomy (p=1,00; p=1.00 respectively), while all CD images were judged unacceptable when compared to DL and CQ (p<0.001, p<0.001 respectively). All images were judged as acceptable/excellent regarding resolution and artefacts, while all CD images were rated as having unacceptable noise.
Conclusions: DL based image reconstruction enables shorter (-42%) scan times with comparable subjective image quality when compared to conventional imaging. Robustness and preservation of sensitivity in detecting pathologies of this approach must be further investigated.