Distinguishing multiple primary lung cancers from intrapulmonary metastasis using CT-based radiomics

Area of Science:

• Radiology
• Oncology
• Medical Imaging Analysis

Background:

• Distinguishing multiple primary lung cancers (MPLCs) from intrapulmonary metastases (IPMs) is crucial for effective treatment planning.
• Accurate differentiation impacts patient management and prognosis in lung cancer cases.

Purpose of the Study:

• To develop and evaluate CT-based radiomics models for efficient discrimination between MPLCs and IPMs.
• To compare the performance of radiomics models against clinical-CT models.

Main Methods:

• Retrospective analysis of 127 patients with 254 pathologically confirmed lung tumors.
• Extraction of radiomics features from CT scans, including refined radiomics with relative differences.
• Development of radiomics model (RM), refined radiomics model (RRM), and fusion models (FM1, FM2) using L1-norm regularization and ANOVA for feature selection.
• Performance assessment using Area Under the Receiver Operating Characteristic Curve (AUC-ROC).

Main Results:

• The refined radiomics model (RRM) achieved a higher AUC (0.870) than the standard radiomics model (RM, 0.857) and the clinical-CT model (CCM, 0.782).
• Fusion models (FM1, FM2) further improved prediction performance, with AUCs of 0.885 and 0.889, respectively.
• No significant performance difference was observed among the fusion models and the RRM.

Conclusions:

• CT-based radiomics models demonstrate strong performance in differentiating MPLCs from IPMs.
• These models hold potential for early diagnosis and guiding treatment strategies for lung cancer patients.
• The refined and fusion radiomics approaches offer enhanced diagnostic accuracy compared to conventional methods.