Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
May 2, 2026

Deep learning for discriminating cochlear malformations on temporal bone CT.

View Full Paper
Ask AI
Bookmark
Share

Authors

ZLZhenhua LiLZLangtao ZhouXBXiang Bin

Discussion

Loading...

Member takes

Overview

Randomized trial assesses deep learning for diagnosing cochlear malformations in temporal bone CT images, indicating diagnostic improvement.

Key Points

  • To evaluate the effectiveness of deep learning models in diagnosing cochlear malformations using temporal bone CT images.
  • Utilized 373 temporal bone CT scans (187 normal, 186 malformed) for analysis.
  • Employed Swin UNETR for automatic cochlea segmentation and classified using ResNet50, EfficientNet-B0, and DenseNet121.
  • Analyzed a test set of 149 sides with deep learning models and otologist assessments.
  • Average Dice coefficient for Swin UNETR was 0.93.
  • AUCs for ResNet50, EfficientNet-B0, and DenseNet121 were 0.93, 0.89, and 0.93, respectively, with DenseNet121 performing best.
  • ResNet50 and DenseNet121 outperformed otologists in diagnosis.

Cite This Study

Li et al. (2026) studied this question.

synapsesocial.com/papers/69f5949771405d493afff5c4https://doi.org/10.1016/j.bjorl.2026.101811
View Full Paper
Ask AI
Bookmark
Share