A Fully Automated Analytic System for Measuring Endolymphatic Hydrops Ratios in Patients With Ménière Disease via...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Otolaryngology

Background:

• Endolymphatic hydrops (EH) analysis via inner ear MRI is crucial for Ménière disease diagnosis.
• Previous automated EH analysis methods had limitations.
• Artificial intelligence (AI) integration in medical imaging is rapidly advancing.

Purpose of the Study:

• To develop a fully automated system for measuring EH ratios using MRI.
• To enhance the accuracy and clinical usability of EH analysis for Ménière disease.
• To improve upon previous AI-based EH analysis systems.

Main Methods:

• Developed novel 3into3Inception and 3intoUNet networks for inner ear segmentation.
• Integrated networks into the INHEARIT-v2 system for automated EH analysis.
• Implemented an automatic representative image selection function from MRI stacks.

Main Results:

• Achieved high segmentation performance with 3intoUNet (0.811 IoU).
• INHEARIT-v2 automatically selected representative MRI slices comparable to expert selections.
• Demonstrated high correlation (ICC 0.941) between INHEARIT-v2 and expert EH ratio measurements.
• Automated analysis of EH ratios completed in approximately 3.5 seconds per patient.

Conclusions:

• INHEARIT-v2 provides a highly accurate and automated solution for EH ratio measurement from MRI.
• The system shows excellent agreement with expert assessments, improving objectivity.
• This AI tool can significantly aid clinicians by reducing workload and enhancing diagnostic precision.