mimic-iv-ecg-ve

Vector Embedding Pipeline for MIMIC-IV-ECG

This repository contains the Python files to create vector embeddings from 12-lead ECG data and to evaluate the effectiveness of vector embedding with downstream tasks.

Step 1

Data handling and data curation for VE model training and evaluation with downstream tasks.

001: EDA for MIMIC-IV-ECG 
002: Diagnosis extraction from MIMIC-IV using ICD 
003: Extraction of demographic information from MIMI-IV-hosp
004: Data clustering for race
005: Data selection for VE model training
006: Creating TF dataset for faster training and evaluation
007: Splitting data for training and evaluation

Step 2

011: VE model creation using VAE approach
012: VE creation using trained VE model

Step 3

021: VE Evaluation with Downstream Tasks