Department of Radiation Oncology, Brigham and Women's Hospital, Dana-Farber Cancer Institute, Harvard Medical School, Boston, Massachusetts.
Radiology and Nuclear Medicine, GROW, Maastricht University Medical Centre, Maastricht, the Netherlands.
Clin Cancer Res. 2019 Jun 1;25(11):3266-3275. doi: 10.1158/1078-0432.CCR-18-2495. Epub 2019 Apr 22.
Tumors are continuously evolving biological systems, and medical imaging is uniquely positioned to monitor changes throughout treatment. Although qualitatively tracking lesions over space and time may be trivial, the development of clinically relevant, automated radiomics methods that incorporate serial imaging data is far more challenging. In this study, we evaluated deep learning networks for predicting clinical outcomes through analyzing time series CT images of patients with locally advanced non-small cell lung cancer (NSCLC). Dataset A consists of 179 patients with stage III NSCLC treated with definitive chemoradiation, with pretreatment and posttreatment CT images at 1, 3, and 6 months follow-up (581 scans). Models were developed using transfer learning of convolutional neural networks (CNN) with recurrent neural networks (RNN), using single seed-point tumor localization. Pathologic response validation was performed on dataset B, comprising 89 patients with NSCLC treated with chemoradiation and surgery (178 scans).
Deep learning models using time series scans were significantly predictive of survival and cancer-specific outcomes (progression, distant metastases, and local-regional recurrence). Model performance was enhanced with each additional follow-up scan into the CNN model (e.g., 2-year overall survival: AUC = 0.74, < 0.05). The models stratified patients into low and high mortality risk groups, which were significantly associated with overall survival [HR = 6.16; 95% confidence interval (CI), 2.17-17.44; < 0.001]. The model also significantly predicted pathologic response in dataset B ( = 0.016).
We demonstrate that deep learning can integrate imaging scans at multiple timepoints to improve clinical outcome predictions. AI-based noninvasive radiomics biomarkers can have a significant impact in the clinic given their low cost and minimal requirements for human input.
肿瘤是不断进化的生物系统，医学影像学具有独特的优势，可以在整个治疗过程中监测变化。尽管定性地跟踪病变在空间和时间上的变化可能微不足道，但开发结合了连续成像数据的临床相关、自动化的放射组学方法则更具挑战性。在这项研究中，我们通过分析局部晚期非小细胞肺癌（NSCLC）患者的时间序列 CT 图像，评估了深度学习网络通过分析时间序列 CT 图像来预测临床结局的能力。数据集 A 包括 179 例接受根治性放化疗的 III 期 NSCLC 患者，具有治疗前和治疗后 1、3 和 6 个月随访的 CT 图像（581 次扫描）。使用具有递归神经网络（RNN）的卷积神经网络（CNN）的迁移学习开发了模型，使用单个种子点肿瘤定位。对包括接受放化疗和手术治疗的 89 例 NSCLC 患者的数据集 B 进行了病理反应验证（178 次扫描）。
使用时间序列扫描的深度学习模型对生存和癌症特异性结局（进展、远处转移和局部区域复发）具有显著的预测能力。随着向 CNN 模型中添加更多的随访扫描，模型性能得到了提高（例如，2 年总生存率：AUC=0.74， < 0.05）。这些模型将患者分为低死亡率和高死亡率风险组，与总生存率显著相关[HR=6.16；95%置信区间（CI），2.17-17.44； < 0.001]。该模型还显著预测了数据集 B 中的病理反应（ = 0.016）。
我们证明了深度学习可以整合多个时间点的成像扫描，以提高临床结局预测的准确性。基于人工智能的非侵入性放射组学生物标志物具有成本低、对人力输入要求低的特点，在临床上可能会产生重大影响。
