Background
Cholesteatoma (CS) is a disease characterized by abnormal growth of keratinized squamous epithelium, which is easy to invade the surrounding structure [
1] and needs surgical resection. According to a meta-analysis [
2], the postoperative recurrence rate of CS is high, ranging from 9 to 70% depending on the surgical approach. Hence, second look surgery was routinely performed in the past [
3]. However, in addition to the general surgical injury, the operation of middle ear CS will also lead to complications such as wax deposition, otorrhea, maze fistula and so on [
4].
BLADE, also called PROPELLER, is a turbo spin-echo (TSE)-based diffusion-weighted imaging (DWI) technique. Since the advent of propeller TSE DWI [
5], it has been shown to perform well in detecting cholesteatoma [
6‐
8]. 2D BLADE turbo gradient echo and spin echo imaging (TGSE BLADE) is the latest modified diffusion sequence using BLADE technology, and studies have shown that TGSE BLADE is superior to readout-segmented echo-planar imaging (RESOLVE) DWI in detecting cholesteatoma [
9,
10]. However, the diagnostic efficacy of TGSE BLADE sequence for recurrent cholesteatoma, as well as its quantitative parameters remain to be further studied.
Signal intensity (SI) on DWI and apparent diffusion coefficient (ADC) on ADC maps are quantitative indicators for detecting CS [
11‐
13]. However, the measurement of SI and ADC values is at least partially influenced by the acquisition technique, the imaging system, and the post-processing platform used [
13‐
15]. Using the SI of subtemporal cortex (SIRT) and background noise (SIRN), and the ADC of the pons (ADC ratio) as references, the calculated quantitative index is more valuable for research [
13,
16,
17].
Therefore, the purpose of this study was to explore the value of TGSE BLADE DWI in the quantitative diagnosis of recurrent temporal bone CS.
Methods
Patients selection
This study was approved by the Internal Review Board and written consent was obtained from each of the patients. From March 2018 to October 2021, 67 patients with suspected recurrence of temporal bone CS after assessment by clinical otorhinolaryngologists who had undergone previous ear surgery for CS were prospectively evaluated by a dedicated protocol, including T1 weighted image (T1WI), T2 weighted fat-suppression image (T2WI-FS) and TGSE BLADE DWI. Pregnant women and the patients with cardiac pacemaker/metallic implant or claustrophobia were excluded from the study. Moreover, 23 patients in whom recurrent CS was detected by MRI, were excluded since they did not accept surgical treatment or had co-morbidities that made the surgery impossible. Finally, there were 44 patients (24 males and 20 females, age range 13–57 years, mean age 34.2 years) included in this prospective study. They all underwent a second surgery, and the surgical method was wall-down surgery.
MRI protocol
All patients underwent axial T1WI, T2WI-FS, TGSE BLADE DWI (a prototype TGSE BLADE DWI sequence) at 3 Tesla scanner (MAGNETOM Prisma, Siemens Healthcare, Erlangen, Germany) with a 64-channel brain coil. The parameters for TGSE BLADE DWI were as follows: TR/TE = 4000/62 ms; slice thickness/gap = 2/0.2 mm; slices = 21; bandwidth = 520 Hz/Px; FOV = 280 × 280 mm
2; matrix = 192 × 192; voxel size = 1.5 × 1.5 × 2.0 mm
3; NEX = 1; diffusion mode = 4 scan trace; b = 0, 1000 s/mm
2; turbo factor = 13; EPI factor = 3; and data acquisition time = 3 min 46 s. The readout was performed by Alsop method in this study [
18].
Image assessment
Qualitative analysis
For each patient, two observers (9 years and 12 years working experience in head and neck imaging, respectively) assessed images independently. The cases were randomly presented to the reviewers who were blinded to the clinical data. Combined evaluation of TGSE BLADE DWI images and T1WI/T2WI images was performed and the presence of CS was scored: diagnostic confidence scores (1 = very low, 2 = low, 3 = medium, 4 = high, and 5 = very high).
The diagnosis is based on the relative hyperintensity on DWI compared to brain tissue, unless the lesion is hyperintense compared with cerebral white matter on T1WI images, which is highly suggestive of the possibility of cholesterol granuloma [
13,
19]. When both readers assigned identical scores to the individually assessed images, the data were collected; when their initial scoring differed, the final score was recorded by consensus. Scores 4 and 5 were considered positive for the prediction of CS.
Quantitative analysis
The two radiologists manually delineated regions of interest (ROI) of the lesions in DWI and ADC maps, respectively, on a PACS workstation. SI measurements were obtained using an ROI (2 mm2) placed within the brightest part of the signal abnormality on the DWI image. The ROI was then copied onto the ADC map to obtain the ADC value. Using the same ROI size, they also obtained the inferior temporal and the background noise to calculate SIRT and SIRN. The background noise ROI was placed far from the skin profile. We calculated using the following formula: SIRT = SI/SI temporal cortex, SIRN = SI/SI noise. Moreover, ADC of the pons at the level of the internal auditory canal was measured with the same ROI size. The ADC ratio was calculated using, as a reference, the pons with the following formula: ADC ratio = ADC value of the lesion/ADC value of the pons.
The SIRT, SIRN, ADC values, and ADC ratios recorded by the two observers were averaged to obtain the final value and defined as the mean SIRT, mean SIRN, mean ADC value, and mean ADC ratio. With T1WI and T2WI-FS guidance, one reader measured the maximum size of the signal abnormality for each patient on TGSE BLADE DWI images.
Statistical analysis
Agreements between the presence of CS, SI and ADC measurements were evaluated by using Cohen’s kappa and intra-class correlation coefficient (ICC). The histologic findings served as the gold standard for the presence of a CS. Differences in mean SIRT, mean SIRN, mean ADC value, and mean ADC ratio between CS and non-cholesteatoma (NCS) lesions were evaluated using the independent t-test. Using receiver operating characteristic (ROC) curve analysis, the area under the ROC curve (AUC) and 95% confidence interval (95% CI) were calculated: the optimal threshold of diagnosis was determined, and the sensitivity, specificity, and accuracy were also computed to evaluate the diagnostic efficacy of the parameters. Pair-wise comparisons of ROC curves were performed using AUC. The threshold for significance was set at P < 0.05. In addition, SPSS software, version 20, was used to perform statistical analysis, and MedCalc 18.2.1 was used to draw ROC curves.
Discussion
The results of this study indicate that the quantitative parameters of TGSE BLADE DWI are helpful in distinguishing residual/recurrent temporal bone CS from NCS lesions.
TGSE BLADE is a multi-shot technique that solves the phase instability inherent in multi-shot DWI by acquiring k-space centers each time and using these data for phase correction. For each shot, a rotating rectangular part of k-space was acquired, after which the reading within the rectangle was performed using the TGSE method (a method of combining TSE and gradient-echo). In contrast to prior BLADE-based methods, TGSE BLADE uses gradient echoes to reduce the acquisition time and minimizes the susceptibility artifacts by separating the gradient echoes and spin echoes between diffusion preparation and data acquisition [
9]. Unlike the single diffusion encoding direction of half-Fourier acquisition single-shot turbo spin-echo (HASTE), TGSE BLADE provides multidirectionally dispersed diffusion patterns, such as 3-scan-trace, 4-scan-trace, etc., to reduce the direction dependency of signal intensities [
21].
The diagnostic efficacy of TGSE BLADE DWI in residual/recurrent temporal bone cholesteatoma has not been previously reported, and it belongs to the non-echo planar imaging (non-EPI) techniques. In this study, the diagnostic sensitivity, specificity, and accuracy of qualitative DWI were 84%, 89.5%, and 86.4%, respectively. Pooled sensitivity of non-EPI DWI for the detection of residual/recurrent CS in a recent meta-analysis was 80–82% with a specificity of 90–100% [
22], which is generally consistent with the results of this study. In addition, it has been shown that the RESOLVE sequence has a sensitivity of 68–76% and a specificity of 60% in primary cholesteatoma [
23]; the sensitivity and specificity of TGSE BLADE DWI in this study were better than those of this sequence, which is consistent with the results of previous studies [
10].
For cholesteatoma imaging, it has been proposed to use the pons, infratemporal cortex and background noise as a reference [
12,
13,
24]. In this study, quantitative parameters such as SIRT, SIRN, ADC value, and ADC ratio have excellent ability to differentiate CS from NCS. These findings are consistently supported by previously reported results [
12,
25,
26]. Recently, Cavaliere et al. [
27] reported that in 109 patients, the ADC value of non-EPI yielded 97% sensitivity and 100% specificity, yet the size of the CS was not mentioned. And it has been shown [
13] that the diagnostic sensitivity, specificity and accuracy of SIRT are 100%. These conclusions, like the present results, reflect the superiority of using quantitative parameters for the diagnosis of CS. In this study, the parameter has good sensitivity and specificity for distinguishing CS from NCS lesions when using ADC of 1.27 × 10
−3 mm
2/s as the cut-off value, which is basically consistent with the ADC cut-off value of 1.24 × 10
−3 mm
2/s calculated by Özgen et al. [
13] and the cut-off value of 1.3 × 10
−3 mm
2/s calculated by Lingam et al. [
11].
The AUCs calculation of ROC curves further showed that quantitative parameters such as SIRT, SIRN, ADC value and ADC ratio had dramatically better diagnostic efficacy compared with qualitative DWI (p = 0.007, 0.009, 0.011 and 0.037, respectively). As in previous studies [
11], ADC value was similarly found to be effective in improving the already high specificity of the qualitative analysis. Few studies have evaluated SI values on DWI due to magnetic field inhomogeneities on MRI and intensity variations of scanner-related artifacts, but the calculation of signal intensity ratio (SIR) may improve the reproducibility of numerical measurements to some extent (measurements were highly consistent between the two radiologists). However, specific thresholds for different centers remain to be further assessed.
The diagnostic superiority of TGSE BLADE DWI for small CS [
9] was also reflected in the detection of residual/recurrent temporal bone CS, and the smallest diameter CS in this study was 2.4 mm, which was clearly visualized on DWI images (Fig.
1). This is consistent with previous findings [
10,
28].
The present study has some limitations. First, only patients with suspected CS recurrence in the temporal bone were included in this study, and not all patients underwent a second surgical exploration after primary ear surgery, so there may be a selection bias. Secondly, the number of cases included in this study was small, which hinders us from concluding that the detection of CS using DWI quantitative parameters may be the first choice after surgery. Finally, the values of the quantitative parameters of the lesion were correlated with the ROI delineated by the observers, so measurement bias was inevitable.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.