← Back to Home

AI Helps Lung Nodule Clinics Reach More Patients, but Benefits for Early Lung Cancer Diagnosis Remain Unproven

After three U.S. clinics introduced an AI platform, the numbers of new patients and patients undergoing invasive procedures increased significantly; however, cancer diagnoses, early-stage cases, and time to diagnosis showed no statistically significant improvement.

By SURL BioNews

After a lung nodule is discovered incidentally on a CT scan, ensuring that the patient receives follow-up and specialist assessment is an important test of the care pathway. A study of three U.S. lung nodule clinics found that more patients entered the clinics and more underwent invasive diagnostic procedures after an AI-assisted platform was introduced. However, the study cannot yet provide a definitive answer on whether increased patient volume translates into earlier detection of lung cancer.

Published on October 14 in PLOS One, the study covered a regional health system, a community hospital, and an academic medical center. The team compared the same months one year apart before and after each institution introduced the platform, including a combined 12.5 months of clinic data across the three institutions for each period. Data before implementation were collected retrospectively, while data after implementation were recorded prospectively.

The platform evaluated was Optellum’s Virtual Nodule Clinic. It identifies patients with lung nodules from radiology reports and combines malignancy risk scoring from CT images with follow-up and missed-appointment reminders. Physicians then decide whether to continue observation or pursue further testing based on risk, guidelines, and clinical judgment. The results therefore reflect changes across the entire care pathway and cannot be attributed solely to the imaging algorithm.

The number of new patients included in the analysis rose from 138 to 232, an increase of approximately 68%; the number undergoing at least one invasive procedure rose from 61 to 100, an increase of approximately 64%. Both differences were statistically significant. The latter measure counts patients, rather than the total number of procedures. A company update from Optellum also cited these figures and noted that patient volume increased at all three institutions, with the largest increase at the institution that previously had no patient tracking system.

Evidence of diagnostic benefits was more limited. The number of confirmed lung cancer cases rose from 37 to 48, while early-stage cases, as defined by the study, rose from 18 to 30. Neither difference was statistically significant. The median time from the initial CT scan to diagnosis changed only from 58 days to 57 days, also without a significant difference. These results are insufficient to demonstrate an improvement in early diagnosis, but they also do not establish that the platform provides no benefit.

Whether the additional testing was appropriate also warrants examination. In an exploratory analysis, the proportion of patients who underwent invasive procedures and ultimately received a benign diagnosis rose from 24.6% to 35.0%, a difference that was not statistically significant. This does not directly establish that the testing was unnecessary, but it reminds health systems to consider the burden of testing on patients when evaluating outcomes.

The study had no concurrent control group and could not consistently capture changes in imaging volume, clinic capacity, or patient mix at each institution. It can therefore identify an association between platform implementation and increased activity, but cannot establish causality. The developer also participated in study design, analysis, and manuscript preparation, and three authors were employees holding company equity or stock options. These financial interests should be considered when interpreting the findings.

Records indexed by Sciety show that a preprint of the same study was available as early as November 2025. Its method of counting procedures, definition of early-stage cancer, and some statistical values differed from those in the final version; the figures in this article follow the published paper. The clearer finding at present is that the clinics reached more patients. Whether this can also reduce delays, improve early diagnosis, and keep the burden of testing under control still requires longer follow-up and studies with control groups.

References

  1. PLOS One
  2. Optellum via LinkedIn
  3. Sciety