MRI of odontogenic maxillofacial infections: diagnostic accuracy and reliability

The incidence of oral infections is rising, and they are currently the most common cause of deep neck infections that require hospitalization [1, 2]. Advanced infections in these patients can lead to life-threatening conditions, including airway compromises, vascular complications, septicemia, and mediastinitis. In addition, clinical factors such as older age, smoking, diabetes, multi-space infection, and the demonstration of a causative molar tooth have been associated with a more extended hospital stay among patients with odontogenic infections [3]. Odontogenic infections have well-established management protocols involving antimicrobial drugs and incision and drainage of abscesses. Death from odontogenic infections is rare and is often associated with prolonged hospital care/mechanical ventilation and secondary pneumonia [4]; early diagnosis is the key to preventing severe complications.

In clinical practice, determining whether the symptoms and findings reflect tissue cellulitis or phlegmon rather than actual abscess formation can be challenging. Emergency imaging is often required in severe cases to determine the exact diagnosis and extent of the abscess in deep neck infections and for accurate and timely intervention [5, 6]. Multidetector computed tomography (CT) of the neck soft tissues or cone-beam computed tomography (CBCT) of bony structures is considered the standard imaging method for acute odontogenic infections [6, 7]. Although MRI is considered time-consuming and challenging for acutely ill patients, emergency neck MRI has been shown to be feasible and to have better diagnostic accuracy than that previously reported for CT [8]. In a prospective head-to-head comparison, MRI was also considered superior to CT in the initial evaluation of acute neck infections [9, 10]. MRI can also reliably detect abscesses [8] and visualize specific edema patterns associated with more severe illnesses [11, 12]. Despite the growing incidence of odontogenic infections and the increasing awareness of the utility of emergency MRI imaging, detailed MRI imaging patterns in serious odontogenic infections and their clinical significance, as well as how accurately MRI can pinpoint the causative tooth, are unknown.

This retrospective cohort study aimed to characterize MRI findings in clinically confirmed odontogenic infections. The purposes of this study were to assess the diagnostic accuracy of MRI in odontogenic abscess detection relative to surgical findings as a reference standard, to study whether MRI findings can predict the extraoral surgical approach and to examine whether MRI can accurately identify the causative tooth.

The authors obtained permission from the hospital district board for this retrospective cohort study in a single academic tertiary care referral center. Reviews were not sought from the institutional review board (IRB) (approval or waiver) because the national legislature does not require this for retrospective studies of existing data. The inclusion criteria were as follows: (1) emergency MRI for suspected neck infection between April 1, 2013, and December 31, 2018; (2) MRI evidence of infection: high signal of fat-suppressed T2-weighted Dixon images suggesting edema, or high signal of fat-suppressed post-Gadolinium (post-Gd) T1-weighted Dixon images suggesting abnormal tissue enhancement; (3) final clinical diagnosis of odontogenic infection in the medical records (as reference standard); and (4) diagnostic image quality deemed by the radiologist reading the study. Using standard neck MRI codes, cases were identified using picture archiving and communication systems (PACS) and radiological information systems (RIS), and the data were cross-referenced with patient medical files.

MRI was performed with a Philips Ingenia 3 Tesla system using a dS HeadNeckSpine coil configuration (Philips Healthcare) and Gd-based contrast agent (Dotarem^®, Guerbet). The supplemental material provides details of the MRI protocol (approximately 30 min). All MRI data were retrospectively reviewed by two fellowship-trained neuroradiologists (J.Hi. and J.He.). MRI criteria for infection and abscess and surgical methods for abscess detection have been previously described [8, 11, 12]. Both the detection [8] and measurement [11] of an abscess have substantial interobserver reliability. For accuracy of abscess detection, the surgical records were reviewed for the presence of pus and were considered the reference standard for abscess diagnosis.

Abscess size (maximal diameter) and edema patterns were evaluated to study whether MRI findings predict the surgical approach (intraoral vs. extraoral). On the basis of the fat-suppressed T2-weighted Dixon images, the authors evaluated edema as a high signal in the submandibular space (SMS; inferior to the mylohyoid muscle), sublingual space (SLS; superior to the mylohyoid muscle), visceral space (VS; infrahyoid soft tissue space including the larynx, surrounding strap muscles, and the thyroid; also including edema in the anterior cervical space between the sternocleidomastoid muscle and the carotid space [12]), retropharyngeal space edema (RPE), and mediastinal edema (ME) [11]. These edema patterns have a substantial interobserver agreement [11, 12]. Surgical records were retrospectively reviewed to determine whether the surgical approach was intraoral or extraoral.

To study whether MRI can accurately demonstrate the causative tooth, three board-certified and fellowship-trained specialist radiologists (M.N., a neuroradiologist with 15 years of experience; V.J., an emergency radiologist with ten years of experience; and J.N., a musculoskeletal radiologist with nine years of experience) independently evaluated a subsample of patients without a previous dental procedure, in whom the causative tooth was unknown at the time of imaging but could later be reliably inferred from clinical notes (clinical reference standard). All readings were blinded to other imaging data, ratings from other radiologists, and clinical data, including referrals. Radiologists were asked to rate the presence or absence of bone marrow changes (in T1-weighted, fat-suppressed T2-weighted, fat-suppressed contrast-enhanced T1-weighted, and DW images), cortical erosions, periapical abnormalities, and artifacts. They were also asked to propose the causative tooth, for their opinion on the most useful MRI sequence to evaluate this and to rate the subjective level of confidence of this assessment on a scale of 1 to 5.

The results are expressed as percentages, means, and standard deviations (SDs). The authors used independent samples T-tests to compare continuous variables and Chi-square (Χ²) tests to compare ordinal data. The sensitivity, specificity, and accuracy of MRI vs. surgery were calculated as previously described [8]. For the multivariate prediction model for extraoral surgery, a binary logistic regression model with two methods of variable selection was used: enter all and forward stepwise. In the multi-reader study on the causative tooth, percentage agreement was calculated for both accuracy (MRI vs. clinical reference standard) and interobserver agreement. The Fleiss multi-reader Kappa was also used to assess the agreement on the causative tooth. Data were analyzed using IBM SPSS Statistics for Mac (version 26, copyright IBM Corporation 2019). P-values less than 0.05 were considered statistically significant.

Previous studies have shown that MRI is a feasible imaging method for diagnosing deep neck infections in an emergency care setting [8] and that MRI has higher diagnostic accuracy for maxillofacial infections than CT [10]. We found that MRI has high diagnostic accuracy for odontogenic abscesses, that MRI findings can predict clinical severity and surgical approach and that MRI can point to the causative tooth. Together, these results add to the growing knowledge on the utility of emergency MRI in acute neck infections.

Accurate delineation of odontogenic abscesses is important for choosing the optimal type and extent of treatment (e.g., surgery). Sensitivity, specificity, and accuracy values of 0.95, 0.84, and 0.92 were found, respectively, for the MRI diagnosis of an odontogenic abscess. The high diagnostic accuracy is consistent with prior reports in larger samples of various types of neck infections [8]. The PPV of 0.95 is also markedly higher than that previously reported for CT (approximately 0.80) [13, 14]. All the false positives (four) were small abscesses (7–22 mm) and had imaging characteristics of an abscess, but no pus had been identified in surgery. However, small abscesses can also be missed in surgery, and pus can emerge between MRI and surgery. These issues may explain some of the false MRI findings in this study.

As expected, extensive infection findings (e.g., parapharyngeal or multi-space abscess) on MRI were associated with a more severe course of the disease (Supplementary Data Table 6). Previously, ME, RPE, and abscess diameter have been shown to predict more severe illnesses in deep neck infection patients with multiple etiologies [11]. Here, ME was a significant predictor of extraoral surgery in patients with odontogenic infections. SLS edema was more common here in odontogenic infections (77%) than was previously shown for tonsillar infections (24%) [12]. In contrast, RPE is less common in odontogenic than in tonsillar infections [11]. Thus, different etiologies of neck infection are associated with distinct soft tissue edema patterns, each with its own clinical significance.

Bone marrow signal changes were present in almost all cases. Based on recent consensus on the nomenclature for MRI of musculoskeletal infection outside the spine, these changes are consistent with acute osteomyelitis [15], although they do not necessarily indicate bone destruction, sequestration, or pus formation inside the bone, as is often associated with this term in the jaw [16]. In patients with suspected deep neck infection or abscess, these bone signal changes may be a valuable indicator of the odontogenic origin of infection because MRI with fat-saturated sequences is very sensitive to bone marrow edema. However, this analysis was restricted to patients already known to have an odontogenic infection, and the authors did not study the differentiation between odontogenic and non-odontogenic infections.

CT and CBCT imaging are considered reference standards for assessing dental emergencies, mostly due to their widespread availability and ability to depict bony structures [6, 7, 17, 18]. Although MRI is considered superior for evaluating soft tissues in deep neck infections [8, 9], whether it can also demonstrate the causative tooth in odontogenic infections has been unclear. Although there was variability among agreement between raters (kappa 0.66), it was found that MRI can pinpoint the causative infected tooth (within a margin of error of one tooth) with good accuracy. The fat-suppressed, contrast-enhanced, T1-weighted sequence was considered the most useful in this regard, and periapical enhancement is particularly suggestive of significant infection (Fig. 4). Gd is recommended for use in odontogenic infections, like in other soft-tissue or musculoskeletal infections [15]. A potential clinical implication of this study is that MRI alone may be sufficient to detect the odontogenic origin of severe neck infections needing medical imaging for suspected abscess, but lack of direct comparison with CT precludes strong practical recommendations. In addition, longer scanning times compared with X-ray or CT may be an issue in clinical practice when imaging the causative tooth.

Artifacts from dental hardware are a common limitation of CT in evaluating the oral cavity [19]. In general, MRI can also suffer from artifacts (such as distortion and signal loss), but these are considered less significant than those of CT in dental imaging [20]. In the subcohort, artifacts from dental hardware were rare and not considered significant in the decision-making.

The strengths of this study are its large sample size, high-quality 3 Tesla MR imaging with a Gd-based contrast agent and DWI, systematic neuroradiological evaluation of MRI findings, blinded multi-reader assessment thorough clinical characterization, and surgical confirmation of abscesses. However, as the study was retrospective in nature, the medical and surgical records may have been incomplete or imprecise. As indications for imaging may vary, the current results may be biased. Circumferential reasoning may also have biased the investigation. The MRI images might have had an impact on the final clinical decision regarding the causative tooth in this study. However, the authors did not consider this likely to significantly bias our results because, in clinical practice, MRI is not considered an established imaging method for indicating the causative tooth in odontogenic infections, and the exact tooth was not usually mentioned in the original reports. Interobserver agreement on the causative teeth was not perfect (Kappa of 0.66 indicates substantial agreement) but much improved when disagreement of one tooth was allowed. While this may indicate difficulties for radiologists in accurately numbering individual teeth (e.g., when some teeth are missing), MRI can pinpoint the region of the infected tooth with acceptable precision. A further limitation of this analysis is that only a subset of patients was included in whom the causative tooth was not known. The surgical results were also sometimes unclear, although in very few cases. The interpretation of some of the MRI findings may be subjective, although a substantial percent agreement was found for the bone marrow signal changes, as has also been found for soft tissue edema patterns [11, 12] and detection of abscesses [8]. The most important limitation is the lack of head-to-head comparisons between MRI and CT, although such data already exist in the literature [10]. The evidence seems to be accumulating that MRI can more accurately show the extent and origin of infection and abscess formation [8, 11, 12]. It should be recognized that MRI may not always be suitable or available for all patients, so these results may not apply to all facilities.

In conclusion, MRI provided clinically meaningful information in patients with odontogenic infections. These results add to our understanding of the clinical utility of MRI in acute neck infections.