liuxiaoliXRZS/TOP-Net

TOP-Net: Tachycardia Onset Early Prediction Using Bi-Directional LSTM in a Medical-Grade Wearable Multi-Sensor System

TOP-Net: Tachycardia Onset Early Prediction Using Bi-Directional LSTM in a Medical-Grade Wearable Multi-Sensor System

Data

• MIMIC-III Waveform Database Matched Subset
• MIMIC-III Database

Code (main)

• Preprocess:

[1] demographic_info.sql: acquire the basic information of patients, including age, gender, first_careunit, admission_type, history of heart diseases
[2] data_process.ipynb: extract the required physiological time series, downsampling, imputation, indicate the tachycardia events by definition, extract the positive and negative sample sets, calculate statistic features

• Model:

topnet_mimic.ipynb: including the main part of the models (TOP-Net, CNN, LSTM, XGBoost, MLP, RF)
topnet_gw.ipynb: keep the same as topnet_mimic