Date Published: November 30, 2018
Publisher: Public Library of Science
Author(s): Ahmed Hosny, Chintan Parmar, Thibaud P. Coroller, Patrick Grossmann, Roman Zeleznik, Avnish Kumar, Johan Bussink, Robert J. Gillies, Raymond H. Mak, Hugo J. W. L. Aerts, Atul J. Butte
Abstract: BackgroundNon-small-cell lung cancer (NSCLC) patients often demonstrate varying clinical courses and outcomes, even within the same tumor stage. This study explores deep learning applications in medical imaging allowing for the automated quantification of radiographic characteristics and potentially improving patient stratification.Methods and findingsWe performed an integrative analysis on 7 independent datasets across 5 institutions totaling 1,194 NSCLC patients (age median = 68.3 years [range 32.5–93.3], survival median = 1.7 years [range 0.0–11.7]). Using external validation in computed tomography (CT) data, we identified prognostic signatures using a 3D convolutional neural network (CNN) for patients treated with radiotherapy (n = 771, age median = 68.0 years [range 32.5–93.3], survival median = 1.3 years [range 0.0–11.7]). We then employed a transfer learning approach to achieve the same for surgery patients (n = 391, age median = 69.1 years [range 37.2–88.0], survival median = 3.1 years [range 0.0–8.8]). We found that the CNN predictions were significantly associated with 2-year overall survival from the start of respective treatment for radiotherapy (area under the receiver operating characteristic curve [AUC] = 0.70 [95% CI 0.63–0.78], p < 0.001) and surgery (AUC = 0.71 [95% CI 0.60–0.82], p < 0.001) patients. The CNN was also able to significantly stratify patients into low and high mortality risk groups in both the radiotherapy (p < 0.001) and surgery (p = 0.03) datasets. Additionally, the CNN was found to significantly outperform random forest models built on clinical parameters—including age, sex, and tumor node metastasis stage—as well as demonstrate high robustness against test–retest (intraclass correlation coefficient = 0.91) and inter-reader (Spearman’s rank-order correlation = 0.88) variations. To gain a better understanding of the characteristics captured by the CNN, we identified regions with the most contribution towards predictions and highlighted the importance of tumor-surrounding tissue in patient stratification. We also present preliminary findings on the biological basis of the captured phenotypes as being linked to cell cycle and transcriptional processes. Limitations include the retrospective nature of this study as well as the opaque black box nature of deep learning networks.ConclusionsOur results provide evidence that deep learning networks may be used for mortality risk stratification based on standard-of-care CT images from NSCLC patients. This evidence motivates future research into better deciphering the clinical and biological basis of deep learning networks as well as validation in prospective data.
Partial Text: Cancer’s ever-evolving nature and interaction with its surroundings continue to challenge patients, clinicians, and researchers alike. One of its deadliest forms appears in the lungs, leading to the most cancer-related mortalities worldwide . Lung cancer is the second most commonly diagnosed cancer in both men and women , with non-small-cell lung cancer (NSCLC) comprising 85% of cases . The ability to accurately categorize NSCLC patients into groups structured around clinical factors represents a crucial step in cancer care. This stratification allows for evaluating tumor progression, establishing prognosis, providing standard terminologies for effective clinical communication, and, most importantly, identifying appropriate treatment plans from chemotherapy and surgery to radiation and targeted therapy. In addition to clinical factors (including performance status) and, to a lesser extent, age and sex , tumor stage—as evaluated through the predominant tumor node metastasis (TNM) staging manual —is often regarded as a universal benchmark for performing such stratification.
In this study, we assessed the utility of deep learning networks in predicting 2-year overall survival of NSCLC patients from CT data. We trained a 3D CNN end to end on patients treated with radiotherapy, and employed a transfer learning approach for those treated with surgery. We demonstrated the CNN’s ability to significantly stratify patients into low and high mortality risk groups, as well as its stability in test–retest and inter-reader variability scenarios. In addition to benchmarking against feature engineering methods, we also highlighted regions with the largest contributions to the captured prognostic signatures, both within and beyond the tumor volume. Finally, our preliminary genomic association studies suggested correlations between deep learning features and cell cycle and transcriptional processes.