Purpose: 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).Experimental Design: 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).Results: 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 ¼ 0.74, P < 0.05). The models stratified patients into low and high mortality risk groups, which were significantly associated with overall survival [HR ¼ 6.16; 95% confidence interval (CI), 2.17-17.44; P < 0.001]. The model also significantly predicted pathologic response in dataset B (P ¼ 0.016).Conclusions: 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.

Historically, most patients with low-risk prostate cancer (clinical category T1c-T2a, prostate-specific antigen level <10 ng/mL, and Gleason 6 disease) were treated with radical prostatectomy, while radiotherapy-based treatment was the favored approach for high-risk localized prostate cancer. 1 However, conservative management of low-risk prostate cancer with active surveillance or watchful waiting (AS/WW) offers an alternative to radical prostatectomy or radiotherapy, 2 and national guidelines began advocating its use in 2010. 3,4 Nevertheless, current AS/WW rates across the United States are not well established, and it is unclear if increasing acceptance of AS/WW for low-risk prostate cancer might be associated with changes in management patterns in higher-risk prostate cancer. Therefore, we examined US trends in management patterns for localized prostate cancer across risk groups.Methods | The custom Surveillance, Epidemiology, and End Results (SEER) Prostate Active Surveillance/Watchful Waiting database, unlike other databases, includes a quality-assured AS/WW variable. 5 The proposal for this study was approved by the SEER custom data group. All men with localized prostate cancer diagnosed between 2010 and 2015 and known management type were included.Patients designated by treating facilities as receiving AS or WW as management without any receipt of definitive therapy were coded by SEER as AS/WW. 5 If changes from AS/WW to definitive therapy occurred within 1 year of diagnosis for reasons other than disease progression, the cases were coded as the definitive therapy used. Definitive therapy types were defined by SEER as either definitive radical prostatectomy or radiotherapy (including external-beam radiotherapy, brachytherapy, or any combination thereof); the positive predictive value and specificity of both variables are high.Baseline characteristics, stratified by year of diagnosis, were summarized by descriptive statistics. Use of initial