Diffusion-weighted magnetic resonance imaging with echo-planar and non-echo-planar (PROPELLER) techniques in the clinical evaluation of cholesteatoma

A cholesteatoma is a benign, pseudotumoral lesion formed by accumulation of keratinized squamous epithelium within the cavities of the middle ear. It can behave aggressively because of ability to destroy bone and affect the facial nerve and inner ear, and can even cause intracranial complications. As such, the treatment of choice is surgery. It can be congenital or acquired, with the latter being more common, especially in chronic otitis media with marginal perforations of the tympanic membrane. Although the diagnosis is essentially clinical (by direct examination), this is usually accompanied by imaging techniques, especially in patients with suspected recurrence, as an alternative to second-look surgery. In light of problems posed by computed tomography (CT) [1,2], magnetic resonance imaging (MRI) has been considered to be a viable option given its greater ability to distinguish between different tissues [3]. Studies in this regard have been conducted with intravenous contrast followed by T1-weighted image acquisition 40 to 60 minutes after contrast administration [4,5]. Diagnosis is based on an absence of contrast uptake. However, interpretation of the results requires expertise [6]. Diffusion-weighted imaging (DWI) based on restricted water movement [7], has been proposed as an alternative [3,8-17], because of the high keratin content. Thus, whereas, echoplanar imaging (EPI-DWI) has the disadvantages of lower spatial resolution and the presence of artefacts due to the proximity of bone [8,18], the introduction of non-echo-planar Abstract


Introduction
A cholesteatoma is a benign, pseudotumoral lesion formed by accumulation of keratinized squamous epithelium within the cavities of the middle ear. It can behave aggressively because of ability to destroy bone and affect the facial nerve and inner ear, and can even cause intracranial complications. As such, the treatment of choice is surgery. It can be congenital or acquired, with the latter being more common, especially in chronic otitis media with marginal perforations of the tympanic membrane. Although the diagnosis is essentially clinical (by direct examination), this is usually accompanied by imaging techniques, especially in patients with suspected recurrence, as an alternative to second-look surgery.
The aim of this study was to assess the utility of EPI-DWI and non-EPI (periodically rotated overlapping parallel lines with enhanced reconstruction-PROPELLER-DWI) diffusionweighted MRI, for the diagnosis of both primary and recurrent cholesteatoma, and to correlate the results with clinical and surgical fi ndings. We want to determine which DWI-technique should be used, if it would be advisable to use them to monitor asymptomatic patients and also establish if there are appropriate to diagnose primary cholesteatoma.

Materials and methods
This was a prospective study of 30 consecutive patients, six of whom had bilateral involvement. Three patients with unilateral involvement did not attend their appointment after the MRI and were excluded. A total of 33 ears were therefore examined.
A total of 21 of the ears examined had a history of previous cholesteatoma surgery. Recurrence was suspected in 8 of these, and a further 8 presented no clinical manifestations or examination fi ndings (one of these ears belonged to a patient with clinical manifestations in the other ear). One ear could not be assessed by examination because of postsurgical stenosis of the external auditory canal. The diagnosis was uncertain for 2 ears, and in a further 2 a previous CT scan suggested recurrence even though there was no clinical evidence of this.
There was no history of surgery in 12 cases: 10 of these had suggestive clinical fi ndings, and, in the remaining 2, a previous CT scan was suggestive but, as with the other group, there was no evidence on examination. Ten patients who underwent ear surgery for other reasons were included as negative controls prior to surgery. None of them showed signs of cholesteatoma during the intervention nor showed a positive result by DWI.  Medical records were reviewed, and imaging data were correlated with surgical fi ndings in patients who underwent surgery following radiological examination. Descriptive statistics were calculated for the variables patient age (mean and standard deviation) and sex (total numbers and percentages).
Sensitivity and specifi city results and positive and negative predictive values were obtained for the PROPELLER-DWI and EPI-DWI techniques in comparison with surgical fi ndings, and 95% confi dence intervals (95%CI) were calculated. Statistical analysis was performed following the recommendations published by Rodríguez Artalejo et al. [19].

Results
The mean patient age was 48.66 years (range of 17 to 75 years and standard deviation: 14.45), with 60% of patients being male and 40% female.
The PROPELLER-DWI technique yielded fi ndings consistent with cholesteatoma in 19 cases. EPI-DWI was also positive in four of these cases, although this technique underestimated lesion size and artefacts due to bone proximity interfered with the images obtained ( Figure 1). Moreover, on another 6 occasions this technique produced fi ndings that were not suffi ciently specifi c to allow a diagnosis to be made ( Figure 2).
All other cases were negative (Figure 3), as were the negative controls.
Due to their clinical relevance and management, the following two cases, both of which were diagnosed using PROPELLER-DWI, are highlighted: In one case in which CT showed erosion of the tegmentum tympani, MRI proved conclusive for assessing temporal lobe herniation, which is essential when planning surgery.
One case of atypical location, with occupation of the epitympanum and no perforation of the tympanic membrane, was found in an adult patient.  [6]. In our series, however, using a 1.5 T scanner and the same matrix, we were   able to diagnose 3 mm lesions (range: 3-22 mm), as was also the case for other authors using a 256x256 matrix [3,6]. In any case, from a clinical point of view, it is widely accepted that under-diagnosed lesions of less than 2-3 mm are not clinically relevant and may be eligible for periodic monitoring to detect any growth [20].
The utility of non-EPI-DWI techniques is supported by numerous studies, most of which used the half-Fourier acquisition single-shot turbo spin-echo (HASTE) technique [10,11,15]. In contrast, there are few published studies based on PROPELLER-DWI, which is the technique used at our institution. This is a multi-shot fast spin echo sequence with radial data acquisition in the k-space. It provides fast imaging and good resolution, with less distortion and fewer artefacts in scans of the skull base when compared with EPI-DWI sequences [8]. to 29% for sensitivity and 66% to 33% for NPV [8,9,13,26].
False positives are usually due to recent surgery with residual haemorrhage, cholesterol granulomas [1] and abscesses [1,3], as well as reconstruction with bone powder [18], although they tend to be easy to identify based on clinical fi ndings [6].
There were no false positives in our series. Like other authors [3,11,14,16,27,28], we obtained some false negatives: 16 by EPI-DWI and only one by PROPELLER-DWI. The latter was a false negative by both techniques and its size on surgery was 5 mm.
The presence of retraction pockets and mural cholesteatomas with little keratin content limits the utility of diffusion-    weighted MRI as they produce no hyperintensity [3]. Their size also has an infl uence as they below the technique's limit of detection when smaller than 3 mm by non-EPI-DWI [1].
In contrast, with EPI-DWI techniques, the sensivity reaches 100% only when the size of the lesion is equal or greater than 5 mm. The lowest size of cholesteatoma detected seems to range from 4 to 5 mm [29,30,31]. Venail et al [29], only reported the diagnosis of 1 lesion of 3 mm. And in the study by Aikele et al [30], the three missed lesions were smaller than 5 mm. These data confi rm the superiority of non-EPI-DWI techniques face to EPI-DWI. Motion artefacts are another limitation described in studies using the non-echo-planar diffusion-weighted HASTE technique [11,12]. We believe the high number of false  [14,33], and the examination should be combined with basic T1-and T2-weighted sequences in order to avoid errors, for example from interposed fat from a previous intervention [34].  [15], obtained 100% sensitivity and specifi city in both groups, as was also the case in a recent meta-analysis [33], which showed comparable results for primary and postsurgical groups. As our case series is small, we cannot draw any conclusions regarding this aspect.
We emphasise that non-EPI-DWI proved positive in two patients with no clinical signs, and these fi ndings were confi rmed during surgery. We have not found any similar references in the literature. These results suggest that some cases might be inaccessible to the clinician by otoscopy as a result of their location or small size. Alternatively, examination may not be feasible, for example because of a narrow external auditory canal, as was the case with one of our patients. As a result, we suggest monitoring these patients periodically by MRI even if they are asymptomatic.
CT has traditionally been regarded as the radiological technique of choice to accompany a diagnosis of primary and recurrent cholesteatoma. However, this technique has low specifi city for distinguishing cholesteatoma from other tissues (fi brosis, granulation tissue, infl ammation), especially in patients who have already undergone surgery and in whom recurrence is suspected, as well as in cases of atypical location or dubious diagnosis [2,35], leading to unnecessary surgery.
Our fi ndings support these fi ndings as none of the patients with a suggestive CT scan yielded a positive result by DWI, and in the case in which surgery was performed, no cholesteatoma was seen during the intervention. CT should be omitted in cases that prove negative by PROPELLER-DWI, thereby reducing the radiation dose in such patients, who are usually monitored by this technique.
There are some limitations in this study. The number of cases is low and, as we have described before, the number of false-negatives may be underestimated as 11 patients were not operated upon. Moreover, the images interpretation was done by a single radiologist, therefore we were unable to establish interobserver agreement.
In conclusion, non-EPI-DWI by PROPELLER-DWI is a reliable technique for diagnosing both recurrent and primary cholesteatoma, thus greatly assisting decision-making by the otorhinolaryngologist. With EPI-DWI, in contrast, interpretation is hindered by the presence of artefacts. As we found positive results in asymptomatic patients, we recommend regular monitoring by PROPELLER-DWI, even in the absence of clinical fi ndings.