Adversarial Learning on Incomplete and Imbalanced Medical Data for Robust Survival Prediction of Liver Transplant Patients

Ehsan Hallaji,Roozbeh Razavi-Far,Mehrdad Saif,Vasile Palade

doi:10.1109/access.2021.3081040

Ehsan Hallaji, Roozbeh Razavi-Far + Show 2 more

Open Access

https://doi.org/10.1109/access.2021.3081040

Copy DOI

Abstract

The scarcity of liver transplants necessitates prioritizing patients based on their health condition to minimize deaths on the waiting list. Recently, machine learning methods have gained popularity for automatizing liver transplant allocation systems, which enables prompt and suitable selection of recipients. Nevertheless, raw medical data often contain complexities such as missing values and class imbalance that reduce the reliability of the constructed model. This paper aims at eliminating the respective challenges to ensure the reliability of the decision-making process. To this aim, we first propose a novel deep learning method to simultaneously eliminate these challenges and predict the patients’ survival chance. Secondly, a hybrid framework is designed that contains three main modules for missing data imputation, class imbalance learning, and classification, each of which employing multiple advanced techniques for the given task. Furthermore, these two approaches are compared and evaluated using a real clinical case study. The experimental results indicate the robust and superior performance of the proposed deep learning method in terms of F-measure and area under the receiver operating characteristic curve (AUC).

Highlights

Liver failure may occur suddenly or chronically under various conditions, such as viral infections, intolerance to certain medications, chronic hepatitis, liver cirrhosis, and hepatocellular carcinoma [1]–[3]
SCORING SYSTEMS Traditionally, patients were listed and prioritized based on their blood type, BMI, and medical condition. They were prioritized based on three simple tests: creatinine, bilirubin, and International Normalized Ratio (INR), which later led to the emergence of other scoring systems called MELD
Inspired by the famous Generative Adversarial Networks (GAN) [45], we propose a deep learning model, called Adversarial Imputation-Classification Network (AICN) to predict survival chance of patients when missing values and class imbalance exist in the training data

Summary

Introduction

Liver failure may occur suddenly or chronically under various conditions, such as viral infections, intolerance to certain medications, chronic hepatitis, liver cirrhosis, and hepatocellular carcinoma [1]–[3]. Considering that these conditions creates a major challenge for the medical team to properly determine the liver transplantation risks for each patient, a careful decision is usually made in two steps. A. SCORING SYSTEMS Traditionally, patients were listed and prioritized based on their blood type, BMI, and medical condition (degree of disease). Due to the increased mortality rate in the liver transplantation waiting list, the MELD system was no longer used, and researchers were searching for more efficient ways to allocate organs to the recipients [15]

Objectives

Methods

Results

Conclusion