Related Experiment Video
Updated: Jul 8, 2026
03:49
Enhanced Cochlear Coverage and Hearing Preservation in High-Frequency Hearing Loss via Electric Acoustic Stimulation with Longer Electrode
Published on: October 11, 2024
Deep Learning-Assisted Prediction of Hearing Outcomes After Anatomically Successful Type I Tympanoplasty
Te-Yi Liu1,2, Hsiang-Chih Chang3, Pa-Chun Wang4,5
1Department of Otolaryngology, Hsinchu Cathay General Hospital, Hsinchu, Taiwan.
Summary
A new deep learning model uses tympanic membrane images to predict hearing outcomes after surgery. This tool shows modest accuracy in predicting air-bone gap closure and residual air-bone gap, aiding patient counseling.
Area of Science:
- Otolaryngology
- Medical Imaging
- Artificial Intelligence
Background:
- Type I tympanoplasty is a common procedure for tympanic membrane (TM) perforations, aiming to restore hearing.
- Predicting postoperative hearing outcomes remains challenging due to limited reliable tools.
Purpose of the Study:
- To develop and validate a deep learning model that integrates automated TM image features and clinical data.
- The model aims to predict postoperative air-bone gap (ABG) closure and residual ABG in patients undergoing type I tympanoplasty.
Main Methods:
- A diagnostic and prognostic model development study was conducted using otoendoscopic images and clinical data from 121 patients.
- A Mask Region-Based Convolutional Neural Network (Mask R-CNN) was employed for TM and perforation segmentation.
- Prognostic models were built using automated image features combined with demographic, clinical, and audiometric data.
Main Results:
- The Mask R-CNN model achieved high accuracy in TM (0.884 CPA) and perforation (0.901 CPA) detection.
- The prognostic models demonstrated moderate predictive ability, with R-squared values of 0.418 for ABG closure and 0.363 for residual ABG.
- Prediction accuracy reached 97% within 10 dB and 74% within 5 dB, significantly outperforming baseline predictions.
Conclusions:
- Deep learning-assisted analysis of TM images offers modest predictive capabilities for hearing outcomes following type I tympanoplasty.
- This image-based approach shows potential as a supplementary tool for preoperative patient counseling in otologic practice.
