AI/ML-Based Software Improves Diagnostic Accuracy in Lung Cancer

The diagnostic accuracy of radiologists aided by the eyonis LCS AI software was greatly improved compared with when radiologists were unaided.

Eyonis™ Lung Cancer Screening (LCS), a proprietary artificial intelligence (AI)/machine learning–based computer-aided detection (CADe)/diagnosis (CADx) software as a medical device, has met the primary end point of the multi-case, multi-reader, retrospective RELIVE trial (NCT06751576), according to a press release from the developer, Median Technologies.¹

Released topline data from the trial demonstrated that eyonis LCS in conjunction with a radiologist achieved clinically meaningful and statistically significant improvement over radiologist assessment alone with respect to diagnostic accuracy (P = .027).

Filings for FDA 510(k) approval and EU CE marketing submission are anticipated to be complete in the second quarter of 2025. FDA clearance is expected as early the third quarter of 2025, and a commercial launch is anticipated following the decision.

The eyonis LCS software was designed to enhance the diagnostic accuracy of radiologists when analyzing low-dose computed tomography (LDCT) scans for lung cancer screening by aiding in the detection, localization, and characterization of lung nodules. The software accomplishes this by reducing false positives and driving clinical management to avoid unnecessary follow-up procedures.

RELIVE was a controlled, blinded, and randomized retrospective study that evaluated the efficacy and safety of eyonis LCS.

“We believe timely screening of the high-risk populations using eyonis LCS can enable doctors to save more lives while further reducing medical costs,” Fredrik Brag, founder and chief executive officer of Median Technologies, stated in the press release.¹ “Furthermore, using eyonis LCS can not only save lives but also prevent healthy patients from undergoing unnecessary medical procedures. This will avoid unnecessary distress for patients and afford payers tremendous cost savings on unnecessary procedures in addition to obviating the even greater costs of palliative care required for late-stage lung cancer management.”

The trial included a cohort of 480 patients who were at a high risk of developing lung cancer by using retrospectively collected imaging and clinical data from patients from 5 major cancer centers in the US and EU. Patients were separated into 2 groups: the control arm, where LDCT scan image readings were performed by radiologists without the assistance of eyonis LCS; and the test arm, where LDCT images were performed by radiologists with the assistance of eyonis LCS through end-to-end processing of chest LDCT images that yielded a malignancy score for each pulmonary node.²

Eligible patients were between 50 and 80 years of age with current or past smoking of 20 pack years or more. Additionally, patients were screened and surveilled for lung cancer according to United States Preventive Services Task Force 2021 Criteria and received LDCT because of inclusion of the high-risk category for lung cancer.

Those with prior lung resection, a pacemaker or other medical devices in the thorax that obscure CT acquisition, and images used during the AI model’s development were excluded from participating. Patients with hilar and/or mediastinal cancer, only ground glass cancer, solid or part-solid nodules greater than 30 mm, imaging of missing slices or slice thickness of more than 3 mm, or partial cover of the lung were also excluded.

Secondary trial end points include sensitivity at max Youden, specificity at max Youden, recall rates for patients who do not have cancer, recall rates for patients with cancer, and time analysis, among others.

Further RELIVE data, secondary end points included, are still being evaluated and will be shared in the coming weeks, with more complete RELIVE data aimed to be presented in future developer communications as well as medical and scientific conferences.

Eyonis LCS was also evaluated in the multicenter, retrospective REALITY trial (NCT06576232), which evaluated the software through data of 1147 patients to characterize patients with vs without cancer.

References

1. eyonis™ Lung Cancer Screening (LCS) meets primary endpoint in RELIVE clinical trial, the final pivotal study required for regulatory submissions. News release. Median Technologies. February 3, 2025. Accessed February 4, 2025. https://tinyurl.com/eknt8vz3
2. Retrospective clinical trial comparing radiologists' diagnosis accuracy in lung cancer screening population with and without the help of an AI/​ML tech-based SaMD (RELIVE). ClinicalTrials.gov. Updated December 30, 2024. Accessed February 4, 2025. https://tinyurl.com/4jn4sapp