AI used to help improve lung cancer biopsies

TORONTO and HAMILTON, Ont. – For patients awaiting lung cancer staging results, delays carry real consequences. Staging, in other words, determining whether cancer has spread beyond the lung to nearby lymph nodes, shapes the entire treatment pathway: surgery, chemotherapy, radiation, or palliative care.

The standard procedure for that determination is endobronchial ultrasound-guided transbronchial needle aspiration, or EBUS-TBNA, a minimally invasive technique that uses a flexible scope and ultrasound imaging to sample lymph nodes in the chest without open surgery. It is well-established, widely performed, and highly effective when performed by appropriately trained clinicians – though its diagnostic performance can vary.

Published research has associated variability in EBUS-TBNA diagnostic outcomes with operator experience and procedural volume, among other factors. When results are inconclusive, patients return for repeat procedures, extending uncertainty in a disease where the speed and accuracy of cancer staging matters.

Addressing that variability is a key research question at the centre of a new McMaster University led clinical trial now underway, with University Health Network’s Toronto General Hospital as a key site. The study aims to evaluate the impact of an AI-assisted intervention developed by Node AI on EBUS-TBNA performance.

The case for AI assistance
Node AI, a Hamilton-based medical AI company co-founded by thoracic surgeon Dr. Wael Hanna, AI scientist Dr. Anthony Gatti, and CEO Mackensey Bacon, has developed a real-time decision-support platform designed to assist clinicians performing EBUS-TBNA.

The platform is designed to integrate into existing EBUS clinical workflows via a cloud-based interface, requiring no additional hardware and with the intention of supporting all major bronchoscope manufacturers, the specialized scopes used to navigate and image the airways during the procedure.

During a biopsy, the AI is designed to analyze ultrasound video in real-time, with the aim of supporting clinicians in detecting lymph node anatomy and informing targeting decisions. The algorithm has been developed using clinical research data and an EBUS video dataset built over more than seven years.

“We know that outcomes can vary depending on the experience of the clinician performing the procedure. The question we are trying to answer through this trial is whether AI assistance can help support more consistent performance and whether that translates into better results for patients. That is a hypothesis worth testing rigorously,” said Dr. Waël Hanna (pictured), president, CMO and co-founder, NodeAI.

The Clinical Trial
The trial aims to evaluate whether integrating this AI-based guidance into the procedure can help reduce variability by supporting more consistent anatomical pattern recognition, with the goal of improving diagnostic performance across a broader range of operators and clinical settings.

The trial, which has received Research Ethics Board (REB) approval and will be conducted with informed patient consent, will enroll 100 participants over three months. The primary endpoint is the platform’s ability to process EBUS imaging and return real-time predictions on more than 90 percent of captured images, a threshold that, if met, would establish the technical feasibility of real-time AI integration during the procedure.

Secondary endpoints will assess whether AI guidance is associated with improvements in diagnostic yield and reductions in variability between operators.

UHN’s Toronto General Hospital was selected as a study site because of its longstanding expertise in EBUS-TBNA and its role in advancing the procedure through clinical research and training.

As one of the world’s most experienced centres in EBUS-TBNA, it provides an appropriate environment in which to rigorously evaluate whether AI-assisted guidance can be integrated into established clinical workflows and whether it warrants further study.

Dr. Kazuhiro Yasufuku, who played a central role in developing EBUS-TBNA at the institution, is now site lead in the UHN study at its Toronto General Hospital.

A question worth asking
Lung cancer is the leading cause of cancer-related death in Canada.

Accurate, timely staging is critical to ensuring patients receive the most appropriate treatment as quickly as possible. Research on EBUS-TBNA outcomes suggests that access to consistent, high-quality results may not be uniform across all critical settings, a gap that, if addressable, could have meaningful implications for patients regardless of where they are treated.

NodeAI is at an early stage of answering whether AI has a role to play in narrowing that gap. The current trial is designed to establish whether the technology performs as intended in a controlled clinical environment. If the data supports that, broader validation across a wider range of institutional settings would follow, where the question of equitable access to consistent staging outcomes would be tested in practice. As with any clinical research, the results will determine what comes next. That is precisely the point.

Conflict of Interest Disclosure
Dr. Kazuhiro Yasufuku also serves as an advisor to NodeAI.