Preoperative Prediction of Spread Through Air Spaces in Lung Cancer Using ¹⁸F-FDG PET-Based Radiomics and Peritumoral...

Area of Science:

• Oncology
• Radiology
• Medical Imaging

Background:

• Spread Through Air Spaces (STAS) is an aggressive lung cancer invasion pattern linked to poor outcomes.
• Current STAS identification relies on postoperative pathology, posing a challenge for preoperative prediction.
• Noninvasive methods for predicting STAS are crucial for personalized treatment strategies.

Purpose of the Study:

• To assess the feasibility of predicting STAS using radiomic and clinicoradiomic models derived from 18F-FDG PET/CT.
• To develop and validate models for noninvasive STAS prediction in lung cancer patients.

Main Methods:

• Retrospective analysis of lung cancer patients with preoperative 18F-FDG PET/CT scans.
• Extraction of radiomic features from intratumoral and peritumoral regions.
• Development of clinical, radiomic-only, and combined clinicoradiomic models using LASSO and logistic regression, with performance evaluated via cross-validation and ROC analysis.

Main Results:

• Radiomic features related to intratumoral metabolism and peritumoral heterogeneity significantly correlated with STAS.
• The combined clinicoradiomic model achieved superior discrimination (AUC ≈ 0.75) compared to other models.
• Lower eosinophil count, SUVmin_tumor, and intratumoral SUV skewness were independent predictors of STAS.

Conclusions:

• Preoperative STAS prediction in lung cancer is achievable using 18F-FDG PET/CT-based radiomics and clinical data.
• This noninvasive approach offers biological insights beyond anatomical assessment.
• Further validation in prospective, multicenter studies is warranted.