Deep Learning Techniques Can Enhance Diagnosis of Meniere Disease

MONDAY, Jan. 26, 2026 (HealthDay News) -- Deep learning techniques can enhance diagnosis of Meniere disease (MD) and severity grading, according to a study published in the January issue of Medical Physics.Zheng Wang, from the Hunan First Normal University in Changsha, China, and colleagues developed and evaluated a novel deep learning-based framework, the multistage severity assessment system (MSAS), designed for precise segmentation and accurate stratification of MD severity using two-dimensional magnetic resonance imaging (MRI). MSAS was developed using pixel-level manual segmentation on MRI datasets from a development cohort of 189 patients and validated in an independent external set of 70 patients.The researchers found that MSAS demonstrated strong performance, achieving overall accuracy and an area under the curve of 0.971 and 0.995, respectively. Mean average precision scores of 0.887 and 0.877 in the internal and external sets, respectively, were yielded by vestibular region detection. Dice coefficients for segmentation were 0.940 and 0.941 for internal and external datasets, respectively; intersection over union values were consistent at 0.889 across datasets. Slice attention maps and gradient-weighted class activation mapping visualizations achieved enhanced interpretability, effectively assisting clinical decision-making."These results advance the methodological and translational readiness of artificial intelligence-assisted evaluation for Meniere's disease and outline a practical path toward scalable, clinically integrated deployment," the authors write.Abstract/Full Text (subscription or payment may be required).Sign up for our weekly HealthDay newsletter