Accuracy of 2D BLADE Turbo Gradient- and Spin-Echo Diffusion Weighted Imaging for the Diagnosis of Primary Middle Ear Cholesteatoma

A 3D and Explainable Artificial Intelligence Model for Evaluation of Chronic Otitis Media Based on Temporal Bone Computed Tomography: Model Development, Validation, and Clinical Application

BACKGROUND

Temporal bone computed tomography (CT) helps diagnose chronic otitis media (COM). However, its interpretation requires training and expertise. Artificial intelligence (AI) can help clinicians evaluate COM through CT scans, but existing models lack transparency and may not fully leverage multidimensional diagnostic information.

OBJECTIVE

We aimed to develop an explainable AI system based on 3D convolutional neural networks (CNNs) for automatic CT-based evaluation of COM.

METHODS

Temporal bone CT scans were retrospectively obtained from patients operated for COM between December 2015 and July 2021 at 2 independent institutes. A region of interest encompassing the middle ear was automatically segmented, and 3D CNNs were subsequently trained to identify pathological ears and cholesteatoma. An ablation study was performed to refine model architecture. Benchmark tests were conducted against a baseline 2D model and 7 clinical experts. Model performance was measured through cross-validation and external validation. Heat maps, generated using Gradient-Weighted Class Activation Mapping, were used to highlight critical decision-making regions. Finally, the AI system was assessed with a prospective cohort to aid clinicians in preoperative COM assessment.

RESULTS

Internal and external data sets contained 1661 and 108 patients (3153 and 211 eligible ears), respectively. The 3D model exhibited decent performance with mean areas under the receiver operating characteristic curves of 0.96 (SD 0.01) and 0.93 (SD 0.01), and mean accuracies of 0.878 (SD 0.017) and 0.843 (SD 0.015), respectively, for detecting pathological ears on the 2 data sets. Similar outcomes were observed for cholesteatoma identification (mean area under the receiver operating characteristic curve 0.85, SD 0.03 and 0.83, SD 0.05; mean accuracies 0.783, SD 0.04 and 0.813, SD 0.033, respectively). The proposed 3D model achieved a commendable balance between performance and network size relative to alternative models. It significantly outperformed the 2D approach in detecting COM (P≤.05) and exhibited a substantial gain in identifying cholesteatoma (P<.001). The model also demonstrated superior diagnostic capabilities over resident fellows and the attending otologist (P<.05), rivaling all senior clinicians in both tasks. The generated heat maps properly highlighted the middle ear and mastoid regions, aligning with human knowledge in interpreting temporal bone CT. The resulting AI system achieved an accuracy of 81.8% in generating preoperative diagnoses for 121 patients and contributed to clinical decision-making in 90.1% cases.

CONCLUSIONS

We present a 3D CNN model trained to detect pathological changes and identify cholesteatoma via temporal bone CT scans. In both tasks, this model significantly outperforms the baseline 2D approach, achieving levels comparable with or surpassing those of human experts. The model also exhibits decent generalizability and enhanced comprehensibility. This AI system facilitates automatic COM assessment and shows promising viability in real-world clinical settings. These findings underscore AI's potential as a valuable aid for clinicians in COM evaluation.

TRIAL REGISTRATION

Chinese Clinical Trial Registry ChiCTR2000036300; https://www.chictr.org.cn/showprojEN.html?proj=58685.

Evaluation of Three Imaging Methods to Quantify Key Events in Pelvic Bone Metastasis

BACKGROUND

The purpose of this study is to compare turbo spin echo diffusion-weighted images in radial trajectory (BLADE DWI) with multi-shot echoplanar imaging (RESOLVE DWI) for imaging the metastatic lesion in the pelvic bone to find a correlation between ADC values and standardized uptake values (SUVs) of FDG uptake in PET/CT. The study also seeks to compare the values of metastatic lesions with those of benign bone lesions, specifically red marrow hyperplasia.

METHODS

The retrospective IRB-approved study included patients with bone metastasis and red marrow hyperplasia in the pelvic bone who underwent 3.0 T MRI with BLADE/RESOLVE DWI sequences and F-18 FDG PET/CT within one month. BVC (best value comparator) was used in determining the nature of bone lesions. Apparent diffusion coefficient (ADC) and standardized uptake value (SUV) were measured by a radiologist and a nuclear medicine physician. MRI image quality was graded with a Likert scale regarding the visualization of the sacroiliac joint, sacral neural foramen, hamstring tendon at ischial tuberosity, and tumor border. Signal-to-noise ratio (SNR) and imaging time were compared between the two DWIs. Mean, peak, and maximum SUVs between metastatic and benign red marrow lesions were compared. SUVs and ADC values were compared. AUROC analyses and cut-off values were obtained for each parameter. Mann-Whitney U, Spearman's rho, and Kolmogorov-Smirnov tests were applied using SPSS.

RESULTS

The final study group included 58 bone lesions (19 patients (male: female = 6:13, age 52.5 ± 9.6, forty-four (75.9%) bone metastasis, fourteen (24.1%) benign red marrow hyperplasia). ADCs from BLADE and RESOLVE were significantly higher in bone metastasis than red marrow hyperplasia. BLADE showed higher ADC values, higher anatomical scores, and higher SNR than RESOLVE DWI (p < 0.05). Imaging times were longer for BLADE than RESOLVE (6 min 3 s vs. 3 min 47 s, p < 0.05). There was a poor correlation between ADC values and SUVs (correlation coefficient from 0.04 to 0.31). The AUROC values of BLADE and RESOLVE MRI ranged from 0.892~0.995. Those of PET ranged from 0.877~0.895. The cut-off ADC values between the bone metastasis and red marrow hyperplasia were 355.0, 686.5, 531.0 for BLADE min, max, and average, respectively, and 112.5, 737.0, 273.0 for RESOLVE min, max, and average, respectively. The cut-off SUV values were 1.84, 5.01, and 3.81 for mean, peak, and max values, respectively (p < 0.05).

CONCLUSIONS

Compared with RESOLVE DWI, BLADE DWI showed improved image quality of pelvic bone MRI in the aspect of anatomical depiction and SNR, higher ADC values, albeit longer imaging time. BLADE and RESOLVE could differentiate bone metastasis and red marrow hyperplasia with quantifiable cut-off values. Further study is necessary to evaluate the discrepancy between the quantifiers between PET and MRI.

Performance of 2D BLADE turbo gradient- and spin-echo diffusion-weighted imaging in the quantitative diagnosis of recurrent temporal bone cholesteatoma

BACKGROUND

Diffusion-weighted imaging (DWI) has become an important tool for the detection of cholesteatoma. The purpose of this study was to explore the value of 2D BLADE turbo gradient- and spin-echo imaging (TGSE BLADE) DWI in the quantitative diagnosis of recurrent temporal bone cholesteatoma (CS).

METHODS

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 magnetic resonance imaging (MRI). Two radiologist assessed images independently. Quantitative parameters such as signal intensity ratio (SIR) calculated using, as a reference, the inferior temporal cortex (SIRT) and the background noise (SIRN), apparent diffusion coefficient (ADC) value, and ADC ratio (with pons as reference) measured on TGSE BLADE sequences were assessed. Using receiver operating characteristic (ROC) curve analysis, the optimal threshold and diagnostic performance for diagnosing recurrent CS were determined. Pair-wise comparison of the ROC curves was performed using the area under the ROC curve (AUC).

RESULTS

Finally, 44 patients were included in this study, including 25 CS and 19 non-cholesteatoma (NCS). Mean SIRT and mean SIRN on TGSE BLADE DWI were significantly higher for CS than NCS lesions (p < 0.001). Meanwhile, mean ADC values and mean ADC ratios on ADC maps were significantly lower in the CS group than in the NCS group (p < 0.001). According to ROC analysis, the diagnostic efficacy of quantitative parameters such as SIRT (AUC = 0.967), SIRN (AUC = 0.979), ADC value (AUC = 1.0), and ADC ratio (AUC = 0.983) was significantly better than that of qualitative DWI (AUC = 0.867; p = 0.007, 0.009, 0.011 and 0.037, respectively).

CONCLUSIONS

Residual/recurrent temporal bone CS can be accurately detected using quantitative evaluation of TGSE BLADE DWI.