Related Experiment Video
Updated: May 31, 2025
05:22
Author Spotlight: Demonstrating Systematic Endobronchial Ultrasound to New Endoscopists
Published on: August 11, 2023
1.8K
A New Deep Learning-Based Method for Automated Identification of Thoracic Lymph Node Stations in Endobronchial
Øyvind Ervik1,2, Mia Rødde3, Erlend Fagertun Hofstad3
1Clinic of Medicine, Nord-Trøndelag Hospital Trust, Levanger Hospital, 7601 Levanger, Norway.
Journal of Imaging
|January 24, 2025
Summary
A new deep learning model classifies thoracic lymph nodes using endobronchial ultrasound (EBUS) images, aiding lung cancer staging. This artificial intelligence approach shows promise for real-time clinical application in lymph node station identification.
Area of Science:
- Pulmonology
- Medical Imaging
- Artificial Intelligence
Background:
- Endobronchial ultrasound-guided transbronchial needle aspiration (EBUS-TBNA) is vital for thoracic lymph node sampling.
- Accurate lymph node staging is critical for lung cancer treatment decisions.
Purpose of the Study:
- To develop and evaluate a deep learning model for classifying thoracic lymph node stations using EBUS images.
- To assess the feasibility of real-time application of this AI model in clinical settings.
Main Methods:
- A deep neural network (DNN) was trained on 28,134 EBUS images from 56 patients.
- Lymph node stations were labeled in real-time by bronchoscopists using the Mountain Dressler nomenclature.
- The DNN model's performance was evaluated against intraoperative labels.
Main Results:
- The DNN model achieved an overall classification accuracy of 59.5%.
- Station 4L showed the highest performance metrics (precision, sensitivity, F1 score ~77.6%).
- The model demonstrated rapid processing (0.65s/10 images), indicating real-time potential.
Conclusions:
- A DNN model can classify thoracic lymph node stations from EBUS images.
- The developed method shows promising performance and potential for clinical integration in lung cancer staging.
