Preoperative CT-Based Habitat Radiomics Classifiers Predict Recurrence in Non-Small Cell Lung Cancer
Oya Altinok1,2,3, Wai Lone J Ho4, Lary Robinson5
1Department of Cancer Epidemiology, H. Lee Moffitt Cancer Center & Research Institute, Tampa, FL, USA.
Summary
This study developed a CT-based radiomics classifier to predict non-small cell lung cancer (NSCLC) recurrence. Habitat-based radiomics showed superior performance in identifying high-risk patients for recurrence-free survival.
Area of Science:
- Radiomics and Medical Imaging
- Oncology
- Biomarker Discovery
Background:
- Patient outcomes after non-small cell lung cancer (NSCLC) surgery vary despite similar staging.
- Identifying reliable biomarkers for predicting recurrence is crucial for personalized treatment strategies.
Purpose of the Study:
- To develop and validate a pre-surgical CT-based radiomics classifier for predicting recurrence risk in NSCLC patients.
- To compare the predictive performance of intratumoral radiomics, habitat-based radiomics, and a combined model.
Main Methods:
- A cohort of 293 surgically resected NSCLC patients was divided into training and testing sets.
- Tumor habitats were identified using unsupervised clustering on pre-surgical CT images.
- Radiomic features were extracted from intratumoral and habitat-defined regions to build logistic regression classifiers.
Main Results:
- The combined radiomics classifier achieved the highest AUC (0.82), outperforming intratumoral (0.75) and habitat (0.81) models.
- Habitat-based radiomics significantly stratified patients for recurrence-free survival (HR=5.41), with habitat-derived information being the strongest predictor.
- High-risk patients identified by the combined model showed the largest risk estimate (HR=8.43).
Conclusions:
- Habitat-based radiomics offers superior predictive performance for NSCLC recurrence compared to intratumoral radiomics.
- CT-based radiomics classifiers can effectively stratify patients based on recurrence risk, aiding in treatment decisions.
