Detection of Coalescent Acute Mastoiditis on MRI in Comparison with CT

This study was conducted retrospectively at an academic tertiary hospital with a referral population of approximately 1.6 million. Study permission was obtained from the institutional ethics committee with a waiver of informed consent.

All patients hospitalized at the otorhinolaryngology department between 2003 and 2016 due to AM infection (International Classification of Diseases 2010 code H70.0) with both contrast-enhanced CT and MRI scans of temporal bones from the same day or consecutive days during hospitalization were enrolled in the study. Exclusion criteria were the following: ipsilateral mastoidectomy prior to either of the imaging studies, AM in a chronically infected ear with cholesteatoma, underlying temporal bone tumor and middle ear sarcoidosis proven by intraoperative biopsy. The final study group had 32 patients.

Clinical data regarding patient age, gender, time of AM diagnosis and imaging studies, clinical symptoms and treatment were collected retrospectively from electronic patient records. All diagnoses were re-validated by a dedicated otorhinolaryngologist according to previously described criteria [20‐22] and categorized into classical AM (symptom duration ≤14 days) or latent AM (symptoms persisting >14 days) in order to more specifically describe our patient cohort.

The CT scans were performed with a preset 50 s delay after intravenous injection of 1.5 ml/kg iohexol 350 mg/ml contrast agent at 3 ml/s injection rate on various 16-row or 64-row multidetector CT scanners. Scans were acquired with straight gantry, a 120 kV tube voltage and tube current of 90–150 mA or quality reference 220mA. Images were reconstructed in the axial plane of lateral semicircular canals and the coronal plane perpendicular to it, with bone and soft tissue kernels and a section thickness of 0.4–0.6 mm (30 patients) or 1 mm (2 patients).

The MRI scans were performed on various 1.5 T scanners in 30 patients and on a 3 T scanner in 2 patients with head and neck coils. The standard protocol consisted of axial and coronal T2 turbo spin echo (TSE) and axial T1 spin echo; axial echo-planar diffusion-weighted imaging (EPI DWI) and apparent diffusion coefficient (ADC); axial isotropic T2-weighted constructive interference in steady state (CISS); and isotropic T1 magnetization-prepared rapid acquisition with gradient echo (MPRAGE) images with intravenous gadoterate meglumine (Dotarem; Guerbet, Aulnay-sous-Bois, France) 0.1 mmol/kg, obtained in the sagittal plane and reconstructed in axial and coronal planes parallel to and perpendicular to the anterior skull base. Details of the protocol are shown in Table 1. Duration of the whole protocol was 30 min.

The anonymized CT and MRI scans were independently assessed from picture archiving and communication system (PACS) by a board-certified neuroradiologist and a board-certified head and neck radiologist with 7 and 8 years of subspecialty working experience, respectively, during reading sessions several months apart. Discrepancies between opinions were resolved by additional joint reading sessions, separately for CT and MRI scans and CT was the reference standard for MRI findings.

On MRI, bone destruction was estimated as lack of definition of bony septa or cortical bone on T2 TSE and T1 Gd MPRAGE images and classified into the following categories: (1) no destruction or (2) suspected or definite destruction. In only partially fluid-filled mastoids septa are not directly assessable on MRI and in these cases septa were assumed to be intact. Additionally, intense mastoid enhancement (approaching the signal intensity of the sigmoid sinus on T1 Gd MPRAGE) and intramastoid diffusion restriction (defined as signal intensity hyperintense to cerebellar parenchyma in DWI, b = 1000 and a signal drop in ADC) were recorded as previously described [17]. For ADC measurement, three different regions of interest (ROI) were placed, each in a different section, on an area with the most diffusion restriction within the mastoid, avoiding the cortical bone and aerated regions and areas with visible artifacts (Fig. 1). Mean values were calculated for each patient.

In CT, bone destruction was estimated from images reconstructed with the bone kernel separately at three anatomical subsites: the mastoid septa, the inner cortical table and the outer cortical table. Bone destruction was divided into the following categories: (1) no destruction, (2) demineralization or suspicious destruction or (3) definite destruction (Fig. 2). Mastoiditis was called coalescent if definite bone destruction at one or more of the subsites was present on CT.

Complications were recorded both in CT and MRI according to a following structure: perisinuous lesions; epidural abscess outside the perisinuous region; venous sinus thrombosis outside the perisinuous region; extracranial abscess; pachymeningitis or leptomeningitis; and other intracranial complications, if any. Perisinuous lesions (contrast filling defects between the sigmoid sinus and the mastoid) were divided into the following four categories as described previously [23]: grade I: no halo, grade II: linear halo, grade III: nodular halo of <4 mm thickness and grade IV: nodular halo of ≥4 mm thickness (the last two categories likely referring to a mural thrombus or an epidural abscess at the sigmoid sinus wall). Other imaging findings were simply recorded as present or absent.

Additionally, a receiver operating characteristic (ROC) analysis with multiple logistic regression was conducted to calculate the area under the curve (AUC) for ROC to determine the optimal ADC threshold for predicting bone destruction in AM, along with corresponding accuracy measurements.

The final study group consisted of 32 pediatric and adult patients (13 male, 19 female) with consecutive contrast-enhanced CT (CECT) and MRI scans. The mean age was 33 years (median 35 years, range 4–81 years) and 9 patients (28%) were children. Anesthesia was required for one MRI scan of a 4-year-old child but the rest of the children (age range 5–14 years) were scanned without anesthesia. Infection was on the left in 14 (44%) and on the right side in 18 (56%) patients. The mean time from AM diagnosis to MRI was 0.53 days (median 0, range 0–5 days). The mean time from CT to MRI was 0.03 days (median 0; range −1 to 1 days). Classical AM was present in 24 (75%) and latent AM in 8 (25%) patients. Mastoidectomy was performed on 20 (63%) patients.

The prevalence rates for bone destruction at different anatomical subsites (Fig. 3) with corresponding kappa values in CT and MRI are presented in Table 2.

The sensitivity, specificity and positive (PPV) and negative predictive values (NPV) in detecting definite bone lesions on MRI (T2 TSE or T1 Gd MPRAGE) were as follows: mastoid septa (100%, 78%, 46%, 100%), outer cortical table (100%, 82%, 44%, 100%), inner cortical table (14%, 76%, 14%, 76%) and coalescent mastoiditis in general (100%, 54%, 42%, 100%).

Intense mastoid enhancement (kappa = 0.69) was present in 17 (53%) patients and diffusion restriction in 21 out of 29 (72%) patients. Both had low specificity when predicting definite bone destruction (intense mastoid enhancement 52–57%; diffusion restriction 32–38%, depending on the anatomical subsite); however, the sensitivity and NPV were 100% for DWI in predicting definite bone destructions regardless of anatomical subsites. For intense mastoid enhancement, the sensitivity and NPV were 100% in predicting septal and outer cortical destructions. Mastoid enhancement was less sensitive in predicting IC destruction (sensitivity 71%; NPV 87%) and accordingly, bone destruction in general (sensitivity 78%; NPV 87%).

The ADC values from the obliterated mastoid ranged from 0.64 to 1.70 × 10⁻³ mm²/s. The mean ADC values differed in groups with (0.93, 95% confidence interval, CI 0.61–1.25 × 10⁻³ mm²/s) and without (1.29, 95% CI 1.19–1.39 × 10⁻³ mm²/s) definite septal destruction and with (0.86, 95% CI 0.48–1.24 × 10⁻³ mm²/s) and without (1.29, 95% CI 1.20–1.38 × 10⁻³ mm²/s) definite outer cortical destruction. Differences were not significant regarding the inner cortical destruction (1.07 versus 1.27 × 10⁻³ mm²/s) and coalescent mastoiditis in general (1.05 versus 1.30 × 10⁻³ mm²/s, p = 0.067).

The optimal ADC cut-off threshold for coalescent mastoiditis was 1.174 × 10⁻³ mm²/s (AUC ROC = 0.72). An ADC value of ≤1.2 × 10⁻³ mm²/s had 86% sensitivity, 71% specificity and 44% PPV for coalescent mastoiditis. An ADC value higher than 1.2 × 10⁻³ mm²/s had 92% NPV for coalescent AM. The ADC values were available from 29 patients; 25 were scanned with the same MRI unit. Discarding the remaining four patients from ROC analysis did not change the threshold values.

Perisinuous lesions were detected more frequently in CT than in MRI. Discrepancies occurred in four patients with grade I and II findings. The MRI missed the linear halo (grade II) in one and small nodular halo <4 mm (grade III) in two patients, likely due to higher contrast differences in CT, which more easily depicts low-density lesions between enhancing sinus and high-density bone, whereas in MRI the adjacent bone remains signal free. Linear halo is a low-risk lesion that probably corresponds to edematous dura and does not associate with epidural abscesses [23, 32]. Previous results are discordant regarding the clinical importance of small nodular halos. A recent study showed correlation with surgically verified intracranial pathology only in grade IV lesions (nodular halo ≥4 mm thick) [32], whereas in children, epidural abscesses have also been found in association with grade III lesions [23]. The clinically most significant grade IV lesions were detected uniformly in both modalities. There were no differences between modalities in detecting epidural abscesses or venous sinus thrombosis outside the perisinuous area. Pachymeningitis and leptomeningitis, which are beyond the scope of CT [33], and one temporal lobe cerebritis were only detected in MRI. In our study, MRI also detected more extracranial complications (6 in MRI versus 4 in CT), likely due to larger imaging volume in MRI.

The results of current study did not support the hypothesis that bone destruction could be directly estimated from MRI in the same way as from CT. The inner cortical table especially remained problematic in MRI; however, based on several different MR characteristics (delineation of bony structures on T2 TSE and T1 Gd MPRAGE, intramastoid enhancement and diffusion restriction), patients could be classified into three different risk categories, according to the imaging findings (Table 3).

MRI can identify high-risk patients with extracranial or intracranial complications and clear bone lesions and these patients could be treated aggressively, based on the MR imaging findings alone. Patients without any alarming findings (no complications nor signs of bone destruction, only mild mucosal enhancement, no significant diffusion restriction with intramastoid ADC above 1.2 × 10⁻³ mm²/s) are very unlikely to have coalescent AM and unlikely to benefit from additional CT. Only in the third group with equivocal findings in MRI (no complications or clear bone defects, but having intense enhancement or marked diffusion restriction with ADC 1.2 × 10⁻³ mm²/s or less) additional CT could give valuable information on bone integrity to confirm or rule out coalescent AM.

Avoiding unnecessary CT scans in AM is important, because the majority of patients are children [16]. To date, CT has been the primary recommended modality for imaging acute mastoid infections due to its ability to detect inflammatory bone erosions, especially when modern, sub-millimeter, high-resolution temporal bone protocols are applied [9]. The CT has better availability and lower cost when compared with MRI; therefore, it is likely that CT will remain the first line imaging modality for AM also in the future. As MRI is more sensitive in detecting intracranial infections [10‐15], MRI is preferred over CT in case of suspected intracranial complications. Our recommendation is to perform additional CT scans (to demonstrate bone integrity) after initial MRI, only for the intermediate risk group. Based on ALARA principles, MRI could also be the first line imaging modality for pediatric patients in situations where cost and availability do not limit its use. In young children, however, the benefits of radiation-free imaging have to be weighed against the potential hazards of anesthesia, if needed for MRI but not for CT.

With respect to contrast agents, a personalized approach must be applied when choosing the imaging modality for the patient with all the risks and benefits weighed individually. The MRI may be the preferable method in patients with iodine allergy, whereas CECT, despite the radiation hazard, is likely to possess lower fetal risk than Gd-enhanced MRI when imaging pregnant patients. Gadolinium deposition in brain tissue is a recent concern that has been detected after repeated administration of Gd-based contrast agents also in patients with normal kidney function, although firm evidence of its harmful effects is lacking according to current knowledge [34]. Both in CT and in MRI contrast agent is mainly needed for visualization of life-threatening intracranial complications (such as venous sinus thrombosis, intracranial abscesses or meningitis) and the potential benefits of its use in most cases are likely to outweigh the potential risks.

There are several limitations of this study. Not all patients with AM undergo radiological imaging in our institution. Therefore, our cohort probably represents patients with clinically more severe or treatment-resistant disease than average. The final number of patients enrolled was small, because only few patients had undergone both imaging studies within a maximum 48 h period (required to ensure reliable correlation between CT and MRI). The study did not include very young children (less than 2 years old), although AM incidence is the highest in this group. The main reason for this was unavailability of both CT and MRI scans in this age group. Structured surgical protocols were not available. Due to the retrospective nature of the study, there was also some variability in imaging equipment, parameters and MRI sequences used; however, this did not influence the results of the ADC ROC analysis. The material for this study has been collected over years and the MRI protocol was created in 2010 with technical possibilities available at that time. The ss EPI-DWI technique was chosen because it enabled covering the whole mastoid in a short scanning time and creating ADC maps, which was prioritized over spatial resolution or distortion-free images. By now, the DWI techniques have evolved and in our institution the sequence in AM protocol has been changed to readout-segmented echo-planar DWI (RESOLVE by Siemens Healthcare, Erlangen, Germany).