Assessing AI for Detecting Lung Nodules and Cancer: Pre- and Post-Deployment Study (AI-Lung)

University of Florida

Status: Withdrawn

Conditions: Lung Cancers; Computer-Aided Detection; Early-Stage Lung Cancer; Lung Nodules; Artificial Intelligence in Radiology

Study type: Observational

Funder types: Other

Identifiers

NCT06746324

qXR-LN-UFL-001

Details and patient eligibility

About

The study evaluates the impact of qXR-LN compared to standard radiologist-only interpretations before and after AI deployment. The goal is to compare how well lung nodules and cancers are detected in two time periods: before and after the implementation of the AI tool in routine clinical practice. The study aims to determine whether the AI system can help radiologists identify more actionable lung nodules and diagnose lung cancer earlier, ultimately improving patient outcomes.

No changes will be made to patients' standard care, and all treatment decisions will be based on the clinical judgment of physicians. The study includes patients over 35 years old who undergo chest X-rays for various medical reasons, excluding those with known lung cancer.

Full description

This study evaluates the clinical impact of the FDA-cleared artificial intelligence (AI) tool, qXR-LN, for detecting lung nodules and diagnosing lung cancer using chest X-rays (CXR). The study employs an ambispective observational cohort design with two cohorts: pre-deployment (before AI implementation) and post-deployment (after AI implementation).

The primary objective is to assess differences in lung nodule detection rates and the percentage of lung cancers diagnosed through the nodule pathway between the two cohorts. Secondary objectives include evaluating whether the AI tool aids in detecting more early-stage lung cancers and identifying reasons for patients dropping out of the nodule clinic pathway.

In the post-deployment cohort, qXR-LN integrates seamlessly with the hospital's existing systems to provide real-time AI findings on radiologists' workstations. Radiologists can accept or reject AI suggestions, ensuring that the final decisions remain under human supervision. Data from both cohorts, including patient demographics, nodule detection rates, cancer diagnoses, and treatment outcomes, will be collected and analyzed.

The study excludes patients under 35 years old and those with known lung cancer at the time of imaging. Ethical considerations include obtaining waivers of consent where applicable and ensuring minimal risk to participants. The findings of this study aim to inform clinical practices and enhance the use of AI tools in lung cancer screening and diagnosis.

Sex: All

Ages: 35+ years old

Volunteers: No Healthy Volunteers

Inclusion criteria

• Age ≥35 years at the time of chest X-ray acquisition
• Chest X-ray must be obtained as part of routine care (e.g., ordered for respiratory complaints, screening, or other clinical indications)
• Chest X-ray performed using CR/DR/DX imaging modality
• Examination described as "Chest"
• View: PA or AP
• Patient positioned as Erect or Supine
• Image available in valid DICOM format with proper DICOM prefix values (including "DICM" in the header)

Exclusion criteria

• Patients aged <35 years at the time of chest X-ray
• Patients with known lung cancer at the time of chest X-ray acquisition
• Lateral views or any imaging modality other than CR/DR/DX
• Imaging or anatomy not specified as Chest (e.g., different body parts or modalities)

Trial design

0 participants in 2 patient groups

Pre-Deployment Cohort

Description:

Patients undergoing standard chest X-rays prior to the introduction of the AI-based Computer Aided Detection (CAD) system. This cohort represents the baseline population used for comparison, with no AI intervention applied during their imaging or reporting process.

Post-Deployment Cohort

Description:

Patients undergoing chest X-rays after the AI-based Computer Aided Detection (CAD) tool has been integrated into the clinical workflow. Although not assigned as an "intervention group" per a traditional trial protocol, these patients receive imaging evaluated by the AI tool, and the impact on diagnostic outcomes will be compared to the pre-deployment cohort.