Toward clinically usable CAD for lung cancer screening with computed tomography

Area of Science:

• Medical Imaging
• Radiology
• Artificial Intelligence in Healthcare

Background:

• Lung cancer screening trials have established criteria for computer-aided detection (CAD) systems.
• Publicly available annotated computed tomography (CT) datasets enable robust CAD system evaluation.
• There is a need for clinically validated CAD systems for lung nodule detection and assessment.

Purpose of the Study:

• To define clinically appropriate computer-aided lung nodule detection (CAD) requirements and protocols.
• To develop and evaluate a new CAD system for lung nodule detection and measurement.
• To assess the system's performance using a publically available annotated CT image dataset.

Main Methods:

• Development of an automated lung nodule detection and measurement system using intensity thresholding, Euclidean Distance Transformation, and watershed segmentation.
• Evaluation of the system's performance against the Lung Imaging Database Consortium (LIDC) CT reference dataset.
• Testing on thin-section CT scans from 108 LIDC subjects.

Main Results:

• The system achieved 100% sensitivity for nodules ≥4 mm and ≥8 mm, with median false positive rates of 0.
• Concordance correlation coefficients for nodule diameter and volume measurements were 0.91 and 0.90, respectively.
• The CAD system demonstrated high accuracy in detecting and measuring lung nodules.

Conclusions:

• The developed CAD system meets clinical requirements for lung nodule detection and assessment.
• The system exhibits high sensitivity and a low false positive rate, crucial for clinical adoption.
• Automated volume measurements show strong agreement with the reference standard, ensuring clinical usability.