Using Artificial Intelligence to Identify Lymph Nodes and Blood Vessels in Endobronchial Ultrasound Images for Lung Cancer Evaluation

What this Trial Is About

Researchers are exploring the use of a Deep Neural Network DNN to assist in evaluating mediastinal and hilar lymph nodes during Endobronchial Ultrasound EBUS procedures. The study aims to determine how well the DNN can identify lymph nodes and blood vessels in patients with undiagnosed enlarged lymph nodes, using ultrasound images. This study is conducted across multiple centers and focuses on lung cancer-related imaging. The study involves training the DNN with annotated ultrasound images to recognize and segment lymph nodes and blood vessels. Initially, the DNNs performance will be assessed using postoperative processed images and static EBUS images. In a later phase, the DNN will be applied in real time during EBUS procedures. The intervention includes a machine learning algorithm running on EBUS images to label mediastinal lymph nodes and their levels during the procedure. Participants referred to thoracic departments with enlarged lymph nodes will be involved. Researchers will evaluate the DNNs capability over 8 months, along with measures of precision, sensitivity, specificity, similarity coefficient, run-time, and monitor adverse events shortly after procedures. The study includes various assessments of the DNNs accuracy and safety, with data collected from ultrasound imaging during and after the EBUS procedure.

Augmented Endobronchial Ultrasound (EBUS-TBNA) With Artificial Intelligence