Data filtering for corrupted MIMIC III dataset with deep learning

Abstract

In this paper, we propose a corrupted data filtering method for MIMIC III dataset based on the convolutional autoencoder. The convolutional autoencoder is employed to restore the corrupted data, and using the restoration error, the degree of data contamination is judged. Based on this function, a corrupted data filtering algorithm is constructed, and arterial blood pressure (ABP) and photoplethysmogram (PPG) signals are filtered. The experimental results show the effectiveness of the proposed method.