Theoretical Clinical Utility of Advanced Practice Provider Use of an Artificial Intelligence Radiomics-based Tool for Pulmonary Nodule Evaluation and Management.
Authors
Affiliations (6)
Affiliations (6)
- Division of Pulmonary, Allergy, and Critical Care, Department of Medicine, University of Pennsylvania, Philadelphia, PA.
- Penn Center for Cancer Care Innovation at the Abramson Cancer Center, University of Pennsylvania, Philadelphia, PA.
- Optellum, Ltd., Oxford, United Kingdom.
- Division of Pulmonary, Critical Care, and Sleep Medicine, University of Connecticut School of Medicine, Farmington, CT.
- Division of Pulmonary, Critical Care, and Sleep Medicine, Department of Medicine, New York University, New York, NY.
- Veterans Affairs New York Harbor Healthcare System, New York, NY.
Abstract
As integral members of pulmonary nodule (PN) programs, advanced practice providers (APPs) routinely evaluate PN cancer risk and make management recommendations. It is unknown whether an artificial intelligence (AI) tool impacts APP PN assessment and decision-making. What is the theoretical effect of APP use of a commercially available AI radiomics-based computer-aided diagnosis tool on PN diagnostic accuracy and management decision-making? In this retrospective multi-reader multi-case study performed from May 2024 to June 2024, 6 APP "readers" (4 in pulmonology, 2 in thoracic surgery) independently evaluated 300 chest CT scan "cases", each with an indeterminate PN 5-30 mm in maximal diameter (50% cancer prevalence). Using solely CT imaging data, APPs provided an estimate of cancer risk and management recommendation for each case without and then with AI tool assistance. The effect of the AI tool on readers' diagnostic performance and management decisions was assessed using descriptive statistics, area under the receiver operating characteristic curve (AUC), and reclassification plots and tables. With AI tool assistance, APP readers' average PN diagnostic accuracy increased by 9 percentage points (AUC: 0.79 vs 0.88; <i>P</i><0.001). A higher proportion of malignant PNs were classified as high (>65%) risk (63% vs 46%; <i>P</i><0.001) and recommended for a lung biopsy or surgical resection (72% vs 55%; <i>P</i><0.001) with AI tool assistance. While benign PNs were more often classified as low (<5%) risk (39% vs 34%; <i>P</i><0.001) with AI tool assistance, there was no statistically significant difference in the proportion recommended for an invasive procedure (18% vs 17%; <i>P</i>=0.3). APP use of a commercially available AI radiomics-based tool for PN evaluation was associated with increased diagnostic accuracy and invasive diagnostic procedure recommendation for malignant PNs. Future prospective, randomized clinical trials are required to assess its use in routine clinical practice.