Deep Learning in Automated Region Proposal and Diagnosis of Chronic Otitis Media Based on Computed Tomography.

OBJECTIVES: The purpose of this study was to develop a deep-learning framework for the diagnosis of chronic otitis media (COM) based on temporal bone computed tomography (CT) scans. DESIGN: A total of 562 COM patients with 672 temporal bone CT scans of both ears were included. The final dataset consisted of 1147 ears, and each of them was assigned with a ground truth label from one of the 3 conditions: normal, chronic suppurative otitis media, and cholesteatoma. A random selection of 85% dataset (n = 975) was used for training and validation. The framework contained two deep-learning networks with distinct functions: a region proposal network for extracting regions of interest from 2-dimensional CT slices; and a classification network for diagnosis of COM based on the extracted regions. The performance of this framework was evaluated on the remaining 15% dataset (n = 172) and compared with that of 6 clinical experts who read the same CT images only. The panel included 2 otologists, 3 otolaryngologists, and 1 radiologist. RESULTS: The area under the receiver operating characteristic curve of the artificial intelligence model in classifying COM versus normal was 0.92, with sensitivity (83.3%) and specificity (91.4%) exceeding the averages of clinical experts (81.1% and 88.8%, respectively). In a 3-class classification task, this network had higher overall accuracy (76.7% versus 73.8%), higher recall rates in identifying chronic suppurative otitis media (75% versus 70%) and cholesteatoma (76% versus 53%) cases, and superior consistency in duplicated cases (100% versus 81%) compared with clinical experts. CONCLUSIONS: This article presented a deep-learning framework that automatically extracted the region of interest from two-dimensional temporal bone CT slices and made diagnosis of COM. The performance of this model was comparable and, in some cases, superior to that of clinical experts. These results implied a promising prospect for clinical application of artificial intelligence in the diagnosis of COM based on CT images.

[Morphological changes associated with low-tone hearing loss in guinea pig models of early endolymphatic hydrops].

The present study was to explore the functional and morphological changes in cochleas of guinea pig models of early endolymphatic hydrops. Thirty albino guinea pigs were randomly divided into three groups: control, 4-week model and 8-week model groups. For each group, n = 10. Model groups were operated on the right ears to result in endolymphatic hydrops with the method of slight destruction of endolymphatic sac and duct from extradural posterior cranial fossa approach, and the animals in control group were sham operated. Electrocochleogram recorded by trans-tympanic approach and auditory brainstem response (ABR) were tested in preoperative model groups, control group, 4-week model group and 8-week model group to assess the hearing changes. Histologic morphometry was used to quantify hydrops by testing scala media area (SMA) ratio. Scanning electron microscope was used to assess the changes of cochlea hair cells. The results showed that the summating potential/compound action potential (SP/AP) ratio of electrocochleogram in 4-week model group (0.33 ± 0.14) and 8-week model group (0.43 ± 0.14) increased significantly, compared with that in control group (0.07 ± 0.06). The maximum SMA ratio in 4-week model group (2.64 ± 0.10) and 8-week model group (3.54 ± 0.13) increased significantly, compared with that in control group (1.06 ± 0.08). The results of maximum SMA ratio correlated with SP/AP ratio of electrocochleogram (r = 0.86). The results of hearing threshold of ABR revealed that the operated ears of model groups were higher than the preoperative results at frequencies of 2 kHz and 4 kHz. And the damage of cochlea hair cells in operated ears occurred in apical and subapical turns. These results suggest the increased SP/AP ratio of electrocochleogram can indicate early endolymphatic hydrops. There is low-tone hearing loss in guinea pig models of early endolymphatic hydrops, and it may be associated with the abnormalities of the stereocilia among the outer hair cells in operated ears which occurs in apical and subapical turns.

[Analysis of correlation between congenital auricular deformities and middle ear malformations].

OBJECTIVE: To explore the potential value of knowing the relationship between congenital auricular deformities and middle ear malformations. METHODS: A total of 86 patients with congenital auricular deformities and middle ear malformations, including 51 males and 35 females, were admitted from January 2008 to December 2009 to the Eye Ear Nose and Throat Hospital of Fudan University. Fifty-eight patients had unilateral deformities (R:L = 34:24), while 28 were bilateral. One hundred and fourteen ears with congenital auricular deformities were included. High-resolution CT (HRCT) data was obtained from each patient. The auricular deformities were classified into three grades using the Marx H classification system. The modified Jahrsdoerfer grading system was used to score the malformations using HRCT data. The correlation between the grades of auricular deformities and scores of middle ear malformations was analyzed using Spearman rank correlation analysis. RESULTS: The Marx H grades of congenital auricular deformities were 12 patients with grade I, 25 patients with grade II and 77 patients with grade III, while their corresponding Jahrsdoerfer scores were 7.8 ± 2.4, 6.8 ± 2.6 and 6.0 ± 2.8, respectively. The statistical analysis suggested a trend of negative correlation between the Marx H grades of auricular deformities and the Jahrsdoerfer scores of middle ear malformations (r = -0.2386, P = 0.0106). CONCLUSION: There was a trend to a negative correlation between congenital auricular deformities and middle ear malformations.