ARCADES: Assessment of a Radiomics-based Computer-Aided Diagnosis Tool for Pulmonary nodulES

Lung cancer is the deadliest malignancy in the U.S. and worldwide due to its high incidence and late stage at diagnosis. Early-stage lung cancer can be detected early as pulmonary nodules (PNs) on chest imaging, but less than 10% of all PNs are lung cancer, and the only way to definitively confirm diagnosis is with an invasive diagnostic procedure associated with considerable risks to patients. Therefore, accurate PN cancer risk assessment and shared decision-making with patients is crucial to guiding PN management decision-making. However, there exists substantial clinical equipoise in how best to perform PN cancer risk assessment.

A bevy of clinical risk prediction models and commercial biomarkers have been developed to aid clinicians in PN cancer risk assessment over the past 3 decades. While several studies have externally validated the diagnostic accuracy of these risk assessment tools and suggested clinical utility in retrospective cohorts, there exist no high-quality prospective data demonstrating the clinical impact of these tools on PN management decision-making among practicing clinicians. An NCI-funded project led by PC3I Faculty Roger Kim, MD, MSCE, DAABIP, ATSF seeks to assess the effect of a radiomics-based computer-aided diagnosis (CAD) tool on PN management, and to identify influential factors in clinicians’ pulmonary nodule risk assessment and how risk assessment tools are used and perceived.

• Artificial Intelligence (AI) Radiomics-based Computer-Aided Diagnostic (CAD) Tool:
  In a retrospective cohort study (2020-2022) of patients who underwent bronchoscopic lung biopsy for pulmonary nodules, the research team evaluated the theoretical impact of using a commercially available AI radiomics-based diagnosis tool alongside available clinical information to predicting the risk of PN malignancy. Findings from this study showed that adding supplemental radiomics-based data significantly improved cancer risk prediction and a clinically meaningful reclassification in predicted malignancy risk—the first study to do so.
• Clinical Utility of Advanced Practice Provider (APP) Use of AI Radiomics-Based Tool:
  A retrospective, multi-case study (2024) assessed the effect of APP use of the AI tool on diagnostic accuracy and decision-making for pulmonary nodules. Using CT imaging data from 300 chest scans, Dr. Kim and team found that the AI-tool’s assistance led to an increase of APPs’ average diagnostic accuracy by 9 percentage points and improved procedure recommendation for malignant PNs.
• Clinicians’ Cancer Risk Assessment Among Patients with Pulmonary Nodules:
  A national survey of clinicians in the U.S. and Canada (2024-2025) found that substantial heterogeneity exists in PN cancer risk assessment and management clinical practice, as clinicians generally agree that PN evaluation is challenging (71.4%) and requires improvement (85.9%), and only 50.5% report having the clinical resources necessary to provide optimal PN care.
• Assessment of a Radiomics-based Computer-Aided Diagnosis Tool for Pulmonary nodulES (ARCADES) pragmatic randomized clinical trial:
  This is an ongoing pragmatic randomized clinical trial that is the first study to prospectively study the clinical impact of a commercial AI radiomics-based CAD tool on clinical decision-making for PN management (NCT05968898).

Findings from this work demonstrate the promise of radiomics-based tools for improving the PN diagnostic process, but they also highlight a crucial care gap and a lack of standardization in PN evaluation and adequate resources for clinicians. Future prospective studies are recommended to assess the tool’s effectiveness in routine clinical practice and the effect on clinical decision-making. Additionally, these findings provide opportunities for clinicians, researchers, industry partners, and health system leaders to develop, test, and implement strategies to improve current pulmonary nodule evaluation practices.

National Cancer Institute (NCI) K08 Mentored Clinical Scientist Career Development Award (K08CA279881)

ACS IRG-22-150-41-IRG American Cancer Society Institutional Research Pilot Grant (ACS IRG-22-150-41-IRG)

Optellum, Ltd.