Real-time assessment of temporomandibular joint using HASTE sequences: feasibility and comparison with standard static sequences

Abstract

Objective:

To test real-time MRI (rtMRI) using HASTE sequences in patients with suspected internal disk derangement (IDD) of temporomandibular joint (TMJ) and to compare these sequences with standard static sequences.

Methods and materials:

99 TMJ were studied with both standard sequences (fat-saturated proton density) and HASTE sequences with high temporal resolution. Image quality was assessed using a 4-point Likert scale. Two radiologists analysed both standard and rtMRI sequences separately, randomly and blinded (by a third operator) to patients’ names in order to assess inter-observer repeatability. One of the radiologists performed the analysis twice for assessing intra-observer repeatability. The same radiologists evaluated randomly and blinded to the previous assessment both the sequences and decided in consensus which was the most credible. Qualitative scores were compared using Friedman’s test while concordance between radiologists and sequences was evaluated using the concordance correlation coefficient.

Results:

At image quality analysis, mean score was 3.41 for static MRI and 3.82 for rtMRI, with a statistically significant difference in favour of rtMRI (p < 0.0001). Inter-rater concordance between operator 1 (Op1) and operator 2 (Op2) with regard to the same sequence was high for both static and rtMRI sequences (0.824 and 0.888, respectively). Inter-rater variability of Op1 and Op2 between static and rtMRI sequences was lower (0.647 for Op1 and 0.633 for Op2). Among 71 discordances between sequences, 60 were judged in favour of rtMRI, while 11 were in favour of static MRI.

Conclusion:

rtMRI with HASTE sequences is a robust technique and provide additional information in assessing IDD compared to static sequences.

Introduction

Temporo-mandibular joint (TMJ) disorders are widespread within the population. Symptoms vary from pain to functional limitations such as limited mouth opening, blockage, or simple “clicks” during mastication.^1,2 Internal disk derangement (IDD) of the TMJ corresponds to lack of coordination between the movement of the mandibular condoyle and that of the articular disk. MRI is the technique of choice for the study of disk-condyle derangement: it plays a pivotal role in visualizing the relationships and relative position of condyle, disk and articular temporal roof; furthermore, MRI is able to depict morphological changes of condyle and disk, as well as bone marrow oedema and joint effusion.³ This is routinely obtained using different MRI protocols, basically consisting of the acquisition of static sequences with mouth closed and maximally opened mouth; some protocols also include the acquisition of images at mid-opening. Usually, T1 and T2 or Proton-density(PD) sequences are acquired along parasagittal and paracoronal planes, properly aligned with the condyle and the ascending branch of the mandible. Such an approach is time-consuming and prone to motion artefacts (PROPELLER sequences are used in non-compliant patients)^4,5; moreover, it provides limited information on disk-condyle dynamics given their static nature.

Real-time MRI (rtMRI; also referred as “pseudodynamic”,⁶ “cine-“,⁷ “kinematic”,^8,9 or “golden angle”¹⁰ in the literature) is not a new technique, given that the first report dates back to 1987.¹¹ However, most of the studies in the literature used FLASH or T1/T2 sequences obtaining controversial evidence of the clinical utility of the technique.^12,13

The purpose of this paper is to explore the feasibility of rtMRI of the TMJ using a different sequence compared to the studies cited before, T2 Half Fourier Single-shot Turbo Spin-Echo (HASTE)^14,15; the second aim was to compare such sequences with the state-of-the-art static technique and assess the potential impact of rtMRI in reporting TMJ MRI studies in a population of patients scanned with suspected IDD.

Methods and materials

Patients

This study was evaluated by the Internal Review Board (IRB) of our University and was conducted in accordance with the ethical statements of the Helsinki declaration. The specific informed content for the non-standard technique was waived by the IRB because it did not heavily impact on duration and biological impact of the study.

This retrospective study included 50 consecutive subjects submitted of the TMJ from October the first to 15 November 2019, acquired on closed and open mouth with both static and rtMRI sequences; a total of 99 TMJ were assessed (in one patient, only one TMJ was evaluated due to claustrophobia in the MR chamber).

MRI protocol

All of the studies were performed on a 1.5 Tesla Siemens Aera scanner (Siemens Healthcare, Erlangen, Germany) using small single-channel circular surface coils with the diameter of 4 cm applied to the skin on the TMJ region, to minimize noise. After fast scout sequences, an axial high spatial resolution Turbo-Spin-Echo (TSE-T2) sequence was acquired parallel to the skull base including mandibular condyles. Parasagittal Spin-Echo (SE)T1 and fat-saturated PD sequences were oriented on the coronal scout parallel to the vertical mandibular branch and on the axial plane orthogonal to the major axis of the mandibular condyle. Proton Density (PD) sequences were repeated during maximal mouth opening (Figure 1). PD sequences were also acquired on a paracoronal plane, parallel to the condyle on the axial sequence and parallel to the condylar neck on parasagittal images, in order to visualize the position of the disk in a frontal view. PROPELLER (BLADE) sequences were used in uncooperative patients.

Figure 1.

Static MRI images from a normal TMJ. In the right image at closed mouth, the articular disk is interposed between mandibular condyle and glenoid fossa. In the left image, acquired in the maximally opened mouth, the disk remains interposed between condyle and articular eminence.

RtMRI sequences were performed without BLADE after the standard protocol and consisted of a single thick slice (6 mm) T2 HASTE sequence per TMJ, oriented using the same parasagittal plane as in the previous morphological sequences selecting the slice best visualizing the disk; patients were asked to open and close their mouths very slowly and continuously during an acquisition time of around 30–40 s (Figure 2). Sequence parameters are listed in Table 1. Figures 1 and 2 show normal findings in static and real time MRI, respectively (Supplementary Videos 1–3, Supplementary Material 1).

Figure 2.

(from 1 to 20). This panel shows RtMRI frames from the same patient than in Figure 1. A good coordination between mandibular condyle and disk is well depicted. A better assessment can be seen in the first movie in the Supplementary material.

Table 1.

Sequence parameters of the HASTE sequence used in the study

Surface coil	Round 4 cm single-array coil
TR/TE (ms)	1000/84
FOV (mm)	180
Thickness (mm)	6
Matrix	256 × 179
Flip angle (grades)	105
Averages	1
Bandwith (Hz/Px)	781
Acquisition time	105 sec

FOV, Filed of view; TE, Echo time; TR, Relaxation time.

Image analysis

Image quality assessment

Image analysis was performed on high-quality screens used for reporting in the clinical practice (Barco MDCC-64430 with screen resolution 1640 × 2048; Barco Display Systems, Kortrijk, Belgium).

Image quality was assessed on both standard MRI and rtMRI by two expert head and neck radiologists (10 and 20 years of experience) in consensus, using a 4-point Likert scale to quantify the visibility of the articular disk: (1) not visible; (2) visible with artefacts; (3) clearly visible only in the closed or fully opened mouth position; (4) clearly visible in all phases; the observers who performed this analysis were asked to focus only on condyle and disk visibility and not on the anatomical detail, given that the spatial resolution of static sequences is definitely higher than with rtMRI. For rtMRI, the number of frames present in the interval between closed and maximally open mouth was used as a further quality index: class (1) one frame; class (2) two or three frames; class (3) >3 frames.

Inter- and intraobserver agreement

The two observers separately, randomly and blinded to patients’ names (an external operator performed the anonymization) assessed both standard and rtMRI sequences. One of the two observers repeated the same analysis 3 months later to assess intraobserver variability. IDD was classified as follows: Grade 1: normal; Grade 2: anterior displacement of disk recaptured during mouth opening; Grade 3: non-reducible anterior disk displacement; Grade 4: stuck disk.^16,17 Differential diagnosis between Grades 1 and 2 was performed using the criteria proposed by Rammelsberg et al¹⁸.

The following measures were extracted from the resulting database: inter-rater concordance between operator 1 (Op1) and operator 2 (Op2) with regard to each sequence; inter-rater agreement of Op1 and Op2 between static and rtMRI sequences; intraobserver agreement for both static and rtMRI sequences (Op2). Given the well-known challenging discrimination between Grade 1 and Grade 2, a separate subanalysis was also performed.

Credibility decision

When ratings of static and rtMRI were discordant, the two operators decided the most likely diagnosis relying on a side-by-side comparison of the static and rtMRI sequences. This solution was taken in the absence of a valuable gold standard to define the potential impact of rtMRI in reporting MRI studies of TMJ.¹⁹

Statistical analysis

Analysis of qualitative data was performed comparing static and rtMRI using the Friedman test (significance level p = 0.05). Concordance in inter- and intraobserver analysis was assessed using the concordance correlation coefficient (CCC), a parameter which ranges from 0 (null correlation) to 1 (perfect correlation). Statistical tests were performed using Medcalc v18.9 (Medcalc, Mariakerke, Belgium).

Results

Qualitative assessment

At image quality analysis of 99 TMJs, mean score was 3.41 for static MRI and 3.82 for rtMRI, with a statistically significant difference in favour of rtMRI (Friedman test p < 0.0001).

On average, rtMRI was able to capture a reasonably wide range of frames between closed and fully opened mouth (from 1 to 8 frames); using the aforementioned classification, 11 TMJ belonged to class 1, 57 to class 2, and 31 to class 3. The number of frames depicted by rtMRI during mouth opening obviously depended on patient compliance (opening and closing the mouth very slowly). No motion artefacts significantly hampered TMJ assessment. Motion-compensation with PROPELLER (BLADE) sequences were used in 17 out of 50 patients.

Interobserver agreement between operators inter rater concordance between operator 1 (Op1) and operator 2 (Op2) with regard to the same sequence showed a very good global correlation for both static and rtMRI sequences (0.824 and 0.888, respectively). Concordance was lower when considering only grades 1 and 2, which are usually the most difficult to discriminate between each other (CCC of 0.728 for static MRI and 0.55 for rtMRI sequences).

Interobserver agreement between sequences

Inter-rater agreement of Op1 and Op2 between static and rtMRI sequences was low (0.647 for Op1 and 0.633 for Op2), and very low when considering only Grades 1 and 2 (0.153 for Op1 and 0.199 for Op2).

Globally, a total of 297 observations were made, 99 by Op1 and 99 repeated twice by Op2. Seventy-one of them (24%) showed discordance between static and rtMRI. The most frequent discordance concerned Grades 1 and 2 (49/71); in particular, rtMRI tended to classify these as normal (Grade 1) sequences, whereas static MR reported them as Grade 2 (41/49), while the inverse occurred in eight cases. In the remaining group of 22 TMJ, rtMRI downstaged 19 Grades 3 and 4 to Grade 1 or 2. In particular, among six cases judged as Grade 4 IDD in the static images, four were down-graded to Grade 1 (3) or Grade 2 (1) in the RtMRI images.

Intraobserver agreement

Intraobserver repeatability was very high, globally 0.802 and 0.932 for static and rtMRI sequences, respectively; it remained high considering differentiation between Grades 1 and 2 (0.842 for static and one for rtMRI). Even analysing repeatability for Grades 1 and 2, intraobserver concordance remained optimal (0.842 and 1 for static and rtMRI sequences).

Credibility decision

Among the 71 previously reported discordances, 60 (84.5%) were judged in favour of rtMRI, while 11 were in favour of static MRI. Among 60 judged in favour of rtMRI, 49 concerned discrimination between Grades 1 and 2 while 11 involved Grades 3 (Figures 3 and 4 for static and four for rtMRI. Movie in the Supplementary material) and 4 (Figures 5 and 6) for static and six for dynamic. Movie in the Supplementary material). Among the 11 cases judged in favour of static MRI sequences, six concerned the discrimination between Grades 1 and 2, while 5 concerned Grades 3 and 4.

Figure 3.

In this case, a Grade 2 internal disk derangement was identified by static MRI images. In the closed mouth position (upper image), the disk seems to be anteriorly displaced, while in the maximally opened mouth position, the posterior portion of the disk (hyperintense compared the anterior portion) seems to be recaptured.

Figure 4.

(from 1 to 12). RtMRI from the same case than in Figure 3. The classification obtained with the static MRI was finally down-graded to normal. The posterior portion of the disk (equally hypointense than the anterior portion) seems to be interposed between condyle and temporal bone in all the frames. A better depiction of disk-condyle coordination can be seen in the second movie available in the Supplementary material.

Figure 5.

In this case, a Grade 4 internal disk derangement (stuck disk) was identified by static MRI images. The disk position (along the posterior face of the articular eminence) seems to be the same in the closed mouth and the maximally opened mouth images.

Figure 6.

In this paradigmatic case (the same case than in Figure 5), RtMRI unequivocally shows a normal movement of the disk. The final judgement was down-graded from Grade 4 to Grade 1. A better depiction of disk-condyle coordination can be seen in the third movie available in the Supplementary material.

Discussion

The HASTE sequence allowed adequate visualization of the articular disk; as temporal resolution is not very high (1 s), the quality of rtMRI images with this technique is strictly dependent on patient cooperation in opening and closing their mouths very slowly; accurate instruction before scanning is therefore of paramount value. Compared with other protocols tested in the literature, the combination of HASTE and surface coils seems to provide higher image quality, in particular, better than TrueFISP and FLASH sequences.^12,20

In our study, rtMRI with HASTE scored better than static MRI. This is quite surprising because HASTE has lower spatial resolution. However, consistency of disk visualization is improved, probably because of the high slice thickness (6 mm) which allows a large part of the disk to be included along the slice selection direction; additionally, T2W is beneficial, as the disk is visualized as uniformly hypointense and surrounded by a hyperintense signal displayed by the surrounding fat and by the fatty bone marrow within the mandibular condyle and temporal bone; differently from static PD sequences, the posterior portion of articular disk is not hyperintense, a condition which may lead to equivocal interpretation.²¹

Inter- and intraobserver repeatability of the two radiologists was very high; this means that rtMRI gives robust information which minimally depends on the operator. For these reasons, rtMRI with HASTE sequences is a feasible technique.

However, correlation of rtMRI with standard static MRI was low, in particular, when trying to separate Grade 1 from Grade 2. In the majority of discordances, TMJ were classified as normal by rtMRI and Grade 2 by static MRI. Compared to our standard of reference (consensus judgement), RtMRI was able to down-grade a significant number of cases from Grade 2 to Grade 1. RtMRI allows real-time analysis of the movement of condyle and disk and sharper visualization of the posterior limit which, on standard PD fat sat sequences, appears more fuzzy and less recognizable. Furthermore, during the acquisition of a static closed mouth sequence using a Burnett’s device, an isometric contraction of masticator muscles may occur: this might theoretically result in traction on the disk or in retropulsion of the condyle; conversely, with rtMRI, a more physiological motion of the condyle and disk can be seen. When the number of frames is sufficiently high, rtMRI even shows the disk leaping over the posterior part of the disk which indicates Grade 2; such a ‘leap over’ sign might refine the diagnostic accuracy of the technique, but it needs to be validated in a more clinically oriented study.

In the major number of observations, rtMRI proved disk mobility in patients diagnosed as Grade 4 (stuck disk) on static images. In these cases, the motion of the disk seen on real-time imaging is an irrefutable sign of the success of rtMRI and error of static MRI sequences. In other cases, rtMRI down-staged Grade 3 findings seen on static MRI to Grade 2, whereas the opposite occurred in fewer patients.

Our work demonstrates the feasibility of rtMRI sequences and their potential in classifying internal disk derangement; further studies are necessary to validate the clinical application of a protocol based solely on real-time evaluation of disk-condyle dynamics.

This study has limitations. The first and major drawback is the lack of detailed clinical data which precluded the possibility to correlate signs and symptoms with findings provided by the two sequences. Another quite obvious limitation is the lack of a true gold standard; in fact, although the static MRI technique is widely used, this does not imply that it is the most accurate; in this paper, the final judgment of two expert radiologists, who assessed simultaneously and in consensus static and rtMR images in discordant cases, has been considered to be a surrogate standard of reference. However, no diagnostic performance indices were calculated. Using this technique, acquisition is limited to a single slice to maintain good time resolution (for this reason, this must be carefully chosen relying on previous morphological sequences); new developments in sequences will probably present the possibility to acquire more than one slice and study.²² Furthermore, no assessment on pseudocoronal plane has been attempted.

In the future, the findings obtained with these sequences should be correlated with prospectively collected clinical data.

In conclusion, rtMRI with HASTE is technically feasible and requires a limited degree of cooperation that can be reasonably obtained from the majority of patients. With this technique, the main TMJ structures are clearly visible and real-time observation of opening and closing movements could help the radiologist to better evaluate and grade IDD.

Key points

• Compare real-time MRI of temporomandibular joint with standard static sequences.

• Using rtMRI with HASTE sequences in patients with suspected internal disk derangement.

• rtMRI with HASTE sequences provide additional information compared to static sequences in patient with suspected internal disk derangement.