AIM in Méniè re’s Disease 122

Ménière’s disease (MD) is difficult to diagnose objectively and evaluate the treatment outcomes. Although pure tone audiometry is the only objective test included in the diagnostic criteria, inner ear MRI technique, which was recently developed to visualize endolymphatic hydrops (EH), is useful for the diagnosis of MD. However, analyzing methods are reported to be diverse, and sometimes, they are timeconsuming and complicated. In recent years, the rapidly developing field of artificial intelligence (AI) showed outstanding performance in image recognition. In particular, convolutional neural network (CNN) based on deep learning plays a remarkable role in today’s medical field, where imaging analysis is critical. We developed a CNN-based deep learning model called INHEARIT (INner ear Hydrops Estimation via ARtificial InTelligence) for automatic calculation of EH ratio in a segmented region of the cochlea and vestibule. The model can generate results that are highly consistent with those generated by manual calculation more quickly. This automated analysis of inner ear MRI using deep learning would be useful for diagnosis and follow-up of MD. It is also expected to be widely used in differential diagnosis of various EH-related diseases.