The Value of Topological Radiomics Analysis in Predicting Malignant Risk of Pulmonary Ground-Glass Nodules: A...

Area of Science:

• Medical Imaging
• Radiomics
• Computational Pathology

Background:

• Accurate differentiation of malignant ground-glass nodules (GGNs) in lung CT is vital for lung adenocarcinoma treatment.
• Current imaging methods struggle with early-stage benign vs. malignant GGN distinction.
• This study introduces topological radiomics for enhanced GGN malignancy prediction.

Purpose of the Study:

• To predict the malignancy risk of GGNs using topological data analysis and texture analysis.
• To evaluate the efficacy of topological radiomics in distinguishing benign from malignant GGNs.
• To develop an integrated predictive model combining clinical, radiomics, and topological features.

Main Methods:

• Retrospective analysis of 3223 patients across two centers (Jan 2018 - June 2023).
• Development of topological radiomics features based on homology for GGNs.
• Integration of machine learning and deep learning algorithms with clinical, radiomics, and topological features.

Main Results:

• Topological radiomics significantly improved GGN differentiation, achieving AUCs of 0.85 and 0.862 in validation sets.
• The topological radiomics model showed higher sensitivity (80.7%-82.3%) than clinical models alone.
• The comprehensive model integrating all features reached the highest AUC of 0.879.

Conclusions:

• Topological radiomics shows strong potential for improving the prediction of GGN malignancy.
• Integrating topological features enhances diagnostic accuracy for GGNs.
• The comprehensive model offers a more reliable basis for GGN treatment strategies.