Salazar Pascal, Cheung Patrick, Ganeshan Balaji, Oikonomou Anastasia
Canon Medical Informatics, Minnetonka, MN, United States of America.
Department of Radiation Oncology, Sunnybrook Health Sciences Centre, University of Toronto, Toronto, Ontario, Canada.
PLoS One. 2024 Dec 31;19(12):e0311910. doi: 10.1371/journal.pone.0311910. eCollection 2024.
This retrospective study explores two radiomics methods combined with other clinical variables for predicting recurrence free survival (RFS) and overall survival (OS) in patients with pulmonary metastases treated with stereotactic body radiotherapy (SBRT).
111 patients with 163 metastases treated with SBRT were included with a median follow-up time of 927 days. First-order radiomic features were extracted using two methods: 2D CT texture analysis (CTTA) using TexRAD software, and a data-driven technique: functional principal components analysis (FPCA) using segmented tumoral and peri-tumoural 3D regions.
Using both Kaplan-Meier analysis with its log-rank tests and multivariate Cox regression analysis, the best radiomic features of both methods were selected: CTTA-based "entropy" and the FPCA-based first mode of variation of tumoural CT density histogram: "F1." Predictive models combining radiomic variables and age showed a C-index of 0.62 95% with a CI of (0.57-0.67). "Clinical indication for SBRT" and "lung primary cancer origin" were strongly associated with RFS and improved the RFS C-index: 0.67 (0.62-0.72) when combined with the best radiomic features. The best multivariate Cox model for predicting OS combined CTTA-based features-skewness and kurtosis-with size and "lung primary cancer origin" with a C-index of 0.67 (0.61-0.74).
In conclusion, concise predictive models including CT density-radiomics of metastases, age, clinical indication, and lung primary cancer origin can help identify those patients with probable earlier recurrence or death prior to SBRT treatment so that more aggressive treatment can be applied.
本回顾性研究探讨了两种放射组学方法结合其他临床变量，用于预测接受立体定向体部放疗（SBRT）的肺转移瘤患者的无复发生存期（RFS）和总生存期（OS）。
纳入111例接受SBRT治疗的163个转移瘤患者，中位随访时间为927天。使用两种方法提取一阶放射组学特征：使用TexRAD软件的二维CT纹理分析（CTTA），以及一种数据驱动技术：使用分割的肿瘤和肿瘤周围三维区域的功能主成分分析（FPCA）。
使用Kaplan-Meier分析及其对数秩检验和多变量Cox回归分析，选择了两种方法的最佳放射组学特征：基于CTTA的“熵”和基于FPCA的肿瘤CT密度直方图的第一变异模式：“F1”。结合放射组学变量和年龄的预测模型显示C指数为0.62，95%置信区间为（0.57-0.67）。“SBRT的临床指征”和“肺原发癌起源”与RFS密切相关，并改善了RFS的C指数：与最佳放射组学特征结合时为0.67（0.62-0.72）。预测OS的最佳多变量Cox模型结合了基于CTTA的特征——偏度和峰度——以及大小和“肺原发癌起源”，C指数为0.67（0.61-0.74）。
总之，包括转移瘤的CT密度放射组学、年龄、临床指征和肺原发癌起源的简明预测模型，有助于在SBRT治疗前识别那些可能较早复发或死亡的患者，从而可以采用更积极的治疗方法。
