Abstract: Proposed a deep-learning model for estimating short-term life expectancy of metastatic cancer patients by analyzing clinical visit narratives while maintaining temporal sequence of the data. The model was trained on data from 10,239 patients and validated on general group of metastatic cancer patients or patients treated with palliative radiation therapy (1818 patients). The model achieved PR-curve AUC score 0.97. We developed an interactive graphical tool that may improve physician understanding of the basis for predictions.
Learning Objective 1: Develop understanding about creation of deep learning model for survival prediction using temporal unstructured data.
Imon Banerjee (Presenter)
Michael Gensheimer, Stanford University
Douglas Wood, Stanford University
A. Solomon Henry, Stanford University
Daniel Chang, Stanford University
Daniel Rubin, Stanford University