Pancreatic ductal adenocarcinoma (PDAC) is a highly lethal cancer with a 5-year survival rate of less than 8%. For patients with resectable disease, the survival rate is only marginally better at 20%, reflecting our current inability to predict the biological aggressiveness of this cancer. A hallmark of PDAC that contributes to its aggressive biology is the variable and often extensive stromal involvement, which has previously hampered advances in molecular subtyping as well as chemotherapy delivery. Recent RNA sequencing studies highlight prognostic subtypes of PDAC based on separate stromal (normal or activated) and tumoral (classical or basal-like) gene-expression signatures. While RNA expression-determined subtypes may better capture the molecular landscape of PDAC, they may not adequately capture the intratumoral heterogeneity of PDAC in vivo. Thus validated prognostic biomarkers of survival are of paramount importance to improving patient survival of this deadly disease.
The challenge focuses on the quantitative assessment of pancreas cancer using a consecutive series of 212 patients undergoing pancreas resection at Memorial Sloan Kettering Cancer Center with clinical variables and high-quality annotated CT imaging. The aim of this challenge is to:
- predict overall survival based on predictors derived from contrast-enhanced pancreas CT scans and patient clinical variables.
Survival analysis is used to analyze data in which the time until the event is of interest. A good overview of survival analysis can be found here.
Our challenge is inspired by the Prostate DREAM Challenge subchallenge 1, in which overall survival was predicted from clinical variables. An overview paper for subchallenge 1 has been published at Lancet Oncology. You can find the accepted manuscript here.
Our challenge is different as it utilizes imaging data in addition to clinical variables.
Participants will build a preoperative model to predict overall survival (OS), defined by “OS_months” (months from surgery to last followup date or death) and “OS_status” (alive or dead) for patients with PDAC. Participants will provide a global risk prediction score. For example, in a linear predictor model, risk score is defined as the exponential of the predicted value. All information in the training data may be used for analysis. Models will be designed using the training data sets and evaluated with the independent test dataset.
Statistical evaluation will be performed by the Department of Epidemiology and Biostatistics at Memorial Sloan Kettering Cancer Center, led by Mithat Gönen. Concordance index is the metric that will be used for evaluation. See the submission instructions page for details on how to submit your models and submission formats. Results will be compared across participants.
Terms and Conditions
The data collection is a limited access data set. By joining the challenge you agree:
- not to transmit this data to any third parties until after MICCAI 2018,
- not to utilize the data for research beyond the scope of this challenge without the written permission of Memorial Sloan Kettering Cancer Center, and
- to appropriately acknowledge Memorial Sloan Kettering Cancer Center in any publications or presentations of results derived from participation in this challenge.
Comma-delimited (.csv) data files must be zipped and uploaded to the submission system. A user account is required for submission - if you don't have one, you can create one following the link on the front page. The following guidelines for submitting result file must be followed.
- Results must be saved as a comma-delimited (.csv) file and zipped for upload.
- The first column must contain the subject ID with the column heading "SNO", e.g. "1". This ID is the prefix for all of the image filenames. For example, 1 is the ID from 001_pdac_tumor.mhd image file.
- The second column must contain the global predicted risk of survival for the patients with the column heading "GlobalRiskScore". Global risk will be interpreted as equal risk and evaluated accordingly (example below).
- All cells should have finite numeric values (no NaN or inf or blank).
- No additional rows or columns may be present in the .csv file.
- Teams will be allowed three submissions in the training phase and three submissions in the testing phase. We will take the top submission for your team as your final team score.
-  Attiyeh MA, Chakraborty J, Doussot A, Langdon-Embry L, Mainarich S, Gönen M, et al. Survival Prediction in Pancreatic Ductal Adenocarcinoma by Quantitative Computed Tomography Image Analysis. Ann Surg Oncol. 2018;25(4):1034–42.
-  Brennan MF, Kattan MW, Klimstra D, Conlon K. Prognostic nomogram for patients undergoing resection for adenocarcinoma of the pancreas. Ann Surg. 2004 Aug;240(2):293–8.
-  Distler M, Pilarsky E, Kersting S, Grutzmann R. Preoperative CEA and CA 19-9 are prognostic markers for survival after curative resection for ductal adenocarcinoma of the pancreas - a retrospective tumor marker prognostic study. Int J Surg. 2013;11(10):1067-1072.