Machine Learning for Healthcare Project 1

This repository contains different deep learning models for classifying ECG time series. Our models are trained and tested on the well-known MIT-BIH Arrythmia Database and on the PTB Diagnostic ECG Database. For a detailed discussion of the models and their performances on the given data we refer to the written report.

---

Getting started

We use a Conda environment for dependency management. Therefore the minimal requirements are Conda. To create an environment, open an Anaconda prompt and run the following:

conda env create -f environment_ml.yml

If antivirus related permission issues or os access errors occur, running the Anaconda prompt on administrator mode avoids such issues.

Add your files

The data input files should be added into a new folder input/. For our model training and evaluating, we used the same data as the baselinemodel and the same train test split.

---

Models

All evaluated models can be found in this package. To import the models to your program use the following import statements:

from models import *

The models are scikeras.Wrappers.KerasClassifier classes. For a detailed instruction on how to use them the reader is referred to the SciKeras documentation. The following models are implemented (see /models/) in the respective self-explanatory files:

CNNs

• Baseline
• DoubleConvCNN
• VanillaCNN

ResNets

• ResNetDS
• ResNetSmall
• ResNetStandard

UNET

• Unet

RNNs

• BaseRNN
• BiDirLSTM
• ConvLSTM
• VanillaLSTM
• VanillaRNN

Ensembles

• LogRegEnsemble
• MeanEnsemble
• PretrainedEnsemble
• VotingEnsemble

Testing and Evaluation

The code that we use for evaluating different models is in the respective subfolders. Further the code for the final models are also included. The following gives a brief overview over the corresponding files. To reproduce our results from the report, follow the exact order of the tasks and their corresponding scripts as they are described in the following.

Task 1

For the steps taken to implement and evaluate the vanilla models, refer to the following files. or a demonstration of model performance, VanillaCNN.ipynb and Final_RNN_Based_Models.ipynb notebooks must be run.

Code	Description
/CNNs/VanillaCNN.ipynb	Vanilla CNN training and testing, hyperparameter tuning
/RNNs/Final_RNN_Based_Models.ipynb	RNN based models training and testing
/RNNs/RNN_Models_Hyperparam_Search.ipynb	RNN based models hyperparameter tuning

Task 2

For the steps taken to develop the models improving those of Task 1, refer to the following files. For a demonstration of model performance, refer again to Final_RNN_Based_Models.ipynb as it contains extended RNN based models as well. For a demosntration of ResNet model performance, ResNet.ipynb notebook must be run.

Code	Description
/ResNet/ResNet.ipynb	ResNet training and testing, hyperparameter tuning
/RNNs/Final_RNN_Based_Models.ipynb	RNN based models training and testing
/RNNs/RNN_Models_Hyperparam_Search.ipynb	RNN based models hyperparameter tuning

Task 3

Code	Description

Task 4

For transfer learning models to run without issues, having the Task 2 models run and saved already is required. Refer to the notebooks in the Task 2 section. To display ResNet transfer learning model training and performance, run TransferResNet.ipynb. To display the training and performance of RNN based models, run Transfer_Learning_RNN_Based_Models.ipynb notebook.

Code	Description
/CNNs/TransferResNet.ipynb	Transfer Learning for ResNet
/RNNs/Transfer_Learning_RNN_Based_Models.ipynb	Transfer Learning for RNN based models

---

Authors

Mert Ertugrul
Johan Lokna
Nora Schneider