Abstract:
Objectives To evaluate the effectiveness of the radiomic features of positioning computed tomography (CT) in prognostication, including overall survival (OS) and progression-free survival (PFS), among patients with limited-stage small cell lung cancer (LS-SCLC) and to improve individualized treatment for them.
Methods A total of 193 patients with LS-SCLC, who were admitted to the Tianjin Medical University Cancer Institute & Hospital, were enrolled in this retrospective study, conducted from September 2013 to December 2019. The patients were randomly assigned into the training and testing groups in a ratio of 7:3. The gross tumor volume (GTV) was segmented by experienced radiologists to extract features as regions of interest. The random forest classification was used to further analyze possible prognostic factors, selected via t-test and least absolute shrinkage and selection operator (LASSO). The performance of the models was evaluated considering the area under the receiver operating characteristic curve (AUC).
Results The median OS of the whole cohort was 29.77 months, and the median PFS was 19.03 months. A total of 1,037 radiomic features were extracted from the CT location images, including the first-order, shape, and texture features. All patients were selected considering different standards to develop models, including cohorts of OS ≤ 12 months or OS ≥ 36 months, OS ≤ 12 months or OS ≥ 60 months, and PFS ≤ 6 months or PFS ≥ 24 months. The mean AUCs of all models were 0.73, 0.79, and 0.70, respectively. The most important features were original_ngtdm_Strength, wavelet-HHL_ngtdm_Busyness, wavelet-LLH_glcm_ClusterShade, and wavelet-LLH_glcm_Correlation.
Conclusions The radiomic model, based on positioning CT, was a viable prognostication tool for LS-SCLC. A combined model that considers clinical factors and radiomic features may help obtain more ideal results.