Multinodular disease: anatomic localization at thin-section CT--multireader evaluation of a simple algorithm.

Area of Science:

• Radiology
• Pulmonary Medicine
• Medical Imaging Analysis

Background:

• Accurate anatomic localization of small pulmonary nodules on thin-section computed tomography (CT) is crucial for diagnosis and management.
• Interobserver variability in interpreting CT images can impact diagnostic accuracy.
• Existing algorithms may require evaluation for reproducibility and precision.

Purpose of the Study:

• To assess the interobserver variability and accuracy of a specific algorithm for localizing small lung nodules on thin-section CT.
• To determine the reliability of the algorithm in categorizing nodule locations.

Main Methods:

• Four experienced radiologists independently applied an algorithm to thin-section CT images of 58 patients.
• Nodules were categorized into four predefined anatomic locations: perilymphatic, random, associated with small airways disease, or centrilobular.
• Algorithm accuracy was validated against literature-based expectations, and interobserver variability was quantified.

Main Results:

• Complete observer agreement on nodule localization was achieved in 79% of cases, with 94% accuracy in nodule identification.
• Triple concordance occurred in an additional 17% of cases, indicating high overall agreement.
• The primary source of disagreement involved distinguishing between perilymphatic and small airways disease-associated nodules.

Conclusions:

• The evaluated algorithm is reproducible and accurate for the anatomic localization of small lung nodules on thin-section CT.
• The algorithm facilitates consistent nodule localization, potentially improving diagnostic reliability.
• Further refinement may be needed to address specific areas of observer disagreement, such as differentiating certain nodule types.