Loading...
Paper Type
ERF
Abstract
Hepatocellular carcinoma (HCC) has high mortalities and warrants research efforts to examine its progression. Three components are central to the underlying cancer progression mechanisms: tumor, tumor microenvironment (TME), and their interactions. We develop a double-layer interactive recurrent neural network method that incorporates two interactive layers to account for the hidden tumor state and hidden TME state, and their interactions over time. Our method leverages deep learning to predict patient survivability with temporal clinical data, unlike existing data-driven techniques that primarily use static patient data. Our method can model both the hidden tumor state variable and the TME state variable to allow timely survivability predictions. A sample of 2,322 HCC patients from a major healthcare organization in Taiwan is used to evaluate the proposed method and several prevalent techniques. In line with previous research, we measure prediction performance with mean absolute error and concordance index.
Recommended Citation
Xu, Anqi; Wen, Bo; Hu, Paul Jen-Hwa; and Huang, Ting Shuo, "A Double-Layer Interactive Recurrent Neural Network Method to Predict Hepatocellular Carcinoma Patient’s Survivability" (2020). AMCIS 2020 Proceedings. 9.
https://aisel.aisnet.org/amcis2020/healthcare_it/healthcare_it/9
A Double-Layer Interactive Recurrent Neural Network Method to Predict Hepatocellular Carcinoma Patient’s Survivability
Hepatocellular carcinoma (HCC) has high mortalities and warrants research efforts to examine its progression. Three components are central to the underlying cancer progression mechanisms: tumor, tumor microenvironment (TME), and their interactions. We develop a double-layer interactive recurrent neural network method that incorporates two interactive layers to account for the hidden tumor state and hidden TME state, and their interactions over time. Our method leverages deep learning to predict patient survivability with temporal clinical data, unlike existing data-driven techniques that primarily use static patient data. Our method can model both the hidden tumor state variable and the TME state variable to allow timely survivability predictions. A sample of 2,322 HCC patients from a major healthcare organization in Taiwan is used to evaluate the proposed method and several prevalent techniques. In line with previous research, we measure prediction performance with mean absolute error and concordance index.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.