Department of Otorhinolaryngology, Faculty of Medicine, Dicle University, Diyarbakir, Turkey.
Eur Rev Med Pharmacol Sci. 2023 Jan;27(1):215-223. doi: 10.26355/eurrev_202301_30874.
Cholesteatoma (CHO) developing secondary to chronic otitis media (COM) can spread rapidly and cause important health problems such as hearing loss. Therefore, the presence of CHO should be diagnosed promptly with high accuracy and then treated surgically. The aim of this study was to investigate the effectiveness of artificial intelligence applications (AIA) in documenting the presence of CHO based on computed tomography (CT) images.
The study was performed on CT images of 100 CHO, 100 non-cholesteatoma (N-CHO) COM, and 100 control patients. Two AIA models including ResNet50 and MobileNetV2 were used for the classification of the images.
Overall accuracy rate was 93.33% for the ResNet50 model and 86.67% for the MobilNetV2 model. Moreover, the diagnostic accuracy rates of these two models were 100% and 95% in the CHO group, 90% and 85% in the N-CHO group, and 90% and 80% in the control group, respectively.
These results indicate that the use of AIA in the diagnosis of CHO will improve the diagnostic accuracy rates and will also help physicians in terms of reducing their workload and facilitating the selection of the correct treatment strategy.
继发于慢性中耳炎（COM）的胆脂瘤（CHO）发展迅速，可导致听力损失等重要健康问题。因此，应迅速且高精度地诊断出 CHO 的存在，并进行手术治疗。本研究旨在探讨人工智能应用（AIA）在基于计算机断层扫描（CT）图像诊断 CHO 方面的有效性。
该研究对 100 例 CHO、100 例非胆脂瘤性（N-CHO）COM 和 100 例对照患者的 CT 图像进行了研究。使用 ResNet50 和 MobileNetV2 两种 AIA 模型对图像进行分类。
ResNet50 模型的总体准确率为 93.33%，MobileNetV2 模型的准确率为 86.67%。此外，这两种模型在 CHO 组中的诊断准确率为 100%和 95%，在 N-CHO 组中的诊断准确率为 90%和 85%，在对照组中的诊断准确率为 90%和 80%。
这些结果表明，AIA 用于诊断 CHO 可提高诊断准确率，并有助于医生减轻工作量，选择正确的治疗策略。
