A pairwise radiomics algorithm-lesion pair relation estimation model for distinguishing multiple primary lung cancer...

Area of Science:

• Radiology
• Oncology
• Medical Imaging Analysis

Background:

• Accurate differentiation between multiple primary lung cancer (MPLC) and intrapulmonary metastasis (IPM) is essential for appropriate treatment and prognosis.
• Current methods often lack non-invasive pre-operative differentiation capabilities.

Purpose of the Study:

• To develop and validate a non-invasive computed tomography (CT) radiomic model for pre-operative differentiation of MPLC from IPM.
• To assess the diagnostic performance of the developed model against experienced clinicians.

Main Methods:

• Retrospective analysis of 168 patients with multiple lung cancers (307 lesion pairs).
• Extraction of radiomic features from CT scans to calculate lesion pair deviations.
• Development of the Lesion Pair Relation Estimation (PRE) model using selected radiomic features and a random forest classifier.
• Validation in internal and independent external cohorts, comparing model performance with two clinicians.

Main Results:

• The PRE model, utilizing seven radiomic features, demonstrated high diagnostic performance.
• Mean Area Under the ROC Curve (AUC) for training, internal validation, and external validation were 0.983, 0.844, and 0.793, respectively.
• The PRE model significantly outperformed two experienced clinicians (AUCs 0.619 and 0.580).

Conclusions:

• The CT radiomic feature-based PRE model shows potential as an accurate non-invasive tool for differentiating MPLC from IPM.
• This model can assist clinicians in pre-operative decision-making for lung cancer patients.