Lung Biopsy Still in the Procedure Room, AI Gets the First Look: 1,000-Patient Study Completes Enrollment

The challenge of lung cancer biopsy lies not only in whether tissue can be obtained, but also in whether physicians can determine on the spot if the sample is sufficient. Invenio Imaging announced that the ON-SITE study evaluating AI-assisted interpretation of lung biopsies has completed enrollment. If the system is validated for use, it could provide preliminary feedback while the patient is still undergoing bronchoscopy, reducing the risk of repeat sampling due to an inadequate specimen.

This prospective, multicenter study enrolled 1,006 participants and obtained more than 3,250 fresh bronchoscopic biopsy specimens across seven medical centers. ClinicalTrials.gov lists it as NCT07045103, with trial code INV-01-2022. Its objectives include using fresh specimens to train, refine, and perform pivotal validation of a deep-learning algorithm. The registry page still lists an estimated enrollment of 900 participants, which appears to be older information predating the latest announcement that enrollment has been completed.

The product under evaluation, NIO Lung Cancer Reveal, analyzes images of fresh, unprocessed tissue acquired by the NIO system to identify morphology suspicious for cancer. Its workflow eliminates staining and sectioning, with the goal of rapidly providing procedural guidance in the bronchoscopy suite instead of waiting until the routine pathology report is completed to learn the sampling result.

The study covers four common sampling scenarios: forceps biopsy, transbronchial needle aspiration of peripheral lung lesions, EBUS-guided lymph-node needle aspiration, and cryobiopsy. Montefiore Einstein, one of the participating centers, said the tissue remains preserved after imaging and can still be used for standard pathology and molecular testing. This is crucial to real-world implementation because a rapid AI readout cannot replace the tests required for subsequent definitive diagnosis and tumor classification.

The U.S. FDA granted the module Breakthrough Device designation in 2024. The stated use at the time was to assist in evaluating bronchoscopic forceps lung biopsies. This designation can accelerate interactions with regulators, but it does not mean the product has received marketing authorization. The ON-SITE study’s inclusion of additional sampling methods also should not be interpreted directly as evidence that the system has the same level of performance in every setting.

The publicly available information currently establishes that the study has completed enrollment, but it does not yet provide sensitivity, specificity, or consistency across centers, nor does it indicate whether AI feedback can actually reduce repeat sampling or improve diagnostic timelines. The next key questions will be the prespecified validation results, performance stratified by sampling method, and how clinical personnel should use or disregard the system’s recommendations when it produces an incorrect assessment.