Monthly Subscription Starting at AED 99 + VAT

Deep learning in Electronic Health Records - CDSS 2
  • Level Professional
  • Duration 32 hours
  • Course by University of Glasgow
  • Offered by

About

Overview of the main principles of Deep Learning along with common architectures. Formulate the problem for time-series classification and apply it to vital signals such as ECG. Applying this methods in Electronic Health Records is challenging due to the missing values and the heterogeneity in EHR, which include both continuous, ordinal and categorical variables. Subsequently, explore imputation techniques and different encoding strategies to address these issues. Apply these approaches to formulate clinical prediction benchmarks derived from information available in MIMIC-III database.

Modules

Welcome

    • 1 Videos
Show info about module content

1 Videos

  • Welcome Video - Deep Learning in Electronic Health Records

Deep Learning and Artificial Intelligence

    • 1 Videos
    • 1 Readings
Show info about module content

1 Videos

  • Deep Learning and Artificial Intelligence

1 Readings

  • Artificial Intelligence

Training Multi-Layer Perceptron Networks

    • 4 Videos
    • 1 Readings
Show info about module content

4 Videos

  • Multi-Layer Perceptron
  • Training a Multi-Layer Perceptron
  • Optimization of a Multi-Layer Perceptron (Part 1)
  • Optimization of a Mutli-Layer Perceptrion (Part 2)

1 Readings

  • Deep Learning for Health Informatics

Multi-Layer Perceptron Networks for ECG Classification

    • 1 Videos
    • 3 Readings
Show info about module content

1 Videos

  • Preprocessing of ECG Signal

3 Readings

  • Practical Exercise: Pre-process ECG data for arrythmia detection
  • Practical Exercise: Split and resample ECG data
  • Practical Exercise: Classify beats using MLP and SVC models and the holdout beats validation protocol

End of Week 1

    • 1 Discussion
    • 1 Quiz
Show info about module content

1 Discussions

  • Week 1 - Your experience

1 Quiz

  • Week 1 summary quiz

Week 1 - Interactive notebook examples

    • 4 Lab
Show info about module content

4 Labs

  • Pre-process ECG data for arrythmia detection
  • Split and resample ECG data
  • Classification of beats using MLP and SVC models and beat holdout method
  • Classification of beats using MLP and SVC models and leave out patients method

Validation strategies of Machine Learning Models

    • 1 Videos
    • 2 Readings
Show info about module content

1 Videos

  • Validation of Machine Learning Models

2 Readings

  • Evaluating Learning Algorithms: A
  • Practical Exercise: Classify beats using MLP and SVC models and the leave out patients validation protocol

Convolutional Neural Networks in Time-Series Classification

    • 1 Videos
    • 2 Readings
Show info about module content

1 Videos

  • Convolutional Neural Networks

2 Readings

  • Practical Exercise: Classify beats using a CNN and the beat holdout validation protocol
  • Practical Exercise: Classify beats using a CNN and the leave-out patients validation protocol

Recurrent Neural Networks in Time-Series Classification

    • 1 Videos
    • 2 Readings
Show info about module content

1 Videos

  • Recurrent Neural Networks

2 Readings

  • Practical Exercise: Classify beats using an LSTM and the beat holdout validation protocol
  • Practical Exercise: Classify beats using an LSTM and the leave-out patients validation protocol

End of Week 2

    • 1 Discussion
    • 1 Quiz
Show info about module content

1 Discussions

  • Week 2 - Your experience

1 Quiz

  • End of week 2 quiz

Week 2 - Interactive notebook examples

    • 5 Lab
Show info about module content

5 Labs

  • Classification of beats using a CNN model and the holdout beats validation protocol
  • Classification of beats using a CNN model and the leave out patients validation protocol
  • Classification of beats using an LSTM model and the holdout beats validation protocol
  • Classification of beats using an LSTM model and the leave out patients validation protocol
  • Dimensionality reduction techniques to visualize ECG data

Data Extraction and Representation Pipeline

    • 1 Videos
    • 2 Readings
Show info about module content

1 Videos

  • Benchmark Deep Learning Models with EHR - Part 1

2 Readings

  • A Data Extraction and Representation Pipeline
  • Practical Exercise: Patients and time-series data extraction of MIMIC-III

In-hospital mortality prediction benchmark pipeline

    • 1 Videos
    • 4 Readings
Show info about module content

1 Videos

  • Benchmark Deep Learning Models with EHR - Part 2

4 Readings

  • Creation of Benchmark data for DNN
  • Practical Exercise: Pre-processing of MIMIC-III dataset
  • Practical Exercise: One-hot encoding and in-hospital mortality prediction
  • Practical Exercise: in-hospital mortality prediction using one-hot encoding and undersampling

Imputation Stategies

    • 2 Videos
    • 2 Readings
Show info about module content

2 Videos

  • Imputation Strategies
  • Deep Learning Imputation Strategies

2 Readings

  • Practical Exercise: Mean vs Joint modelling imputation
  • Imputation based on moments

End of Week 3

    • 1 Discussion
    • 1 Quiz
Show info about module content

1 Discussions

  • Week 3 - Your experience

1 Quiz

  • End of week 3 quiz

Week 3 - Interactive notebook examples

    • 5 Lab
Show info about module content

5 Labs

  • Study cohort selection and variable extraction
  • Imputation methods for in-hospital mortality prediction
  • In-hospital mortality prediction using one-hot encoding
  • In-hospital mortality prediction using one-hot encoding and undersampling
  • Pre-processing

Target Encodings

    • 2 Videos
    • 2 Readings
Show info about module content

2 Videos

  • Categorical and Continuous Variables
  • Bayesian Target Encoding

2 Readings

  • Practical Exercise: mean target encoding
  • Practical exercise: leave one out encoding

Natural Language Inspired Encodings

    • 1 Videos
    • 2 Readings
Show info about module content

1 Videos

  • Encodings Inspired from NLP

2 Readings

  • Representation Learning for Electronic Health Records
  • Practical Exercise: encoding using an autoencoder

Similarity Encodings

    • 1 Videos
    • 1 Readings
Show info about module content

1 Videos

  • Other Types of Embeddings

1 Readings

  • Similarity Encodings

End of Week 4

    • 1 Discussion
    • 1 Quiz
Show info about module content

1 Discussions

  • Week 4 - Your experience

1 Quiz

  • End of week 4 quiz

Week 4 - Interactive notebook examples

    • 4 Lab
Show info about module content

4 Labs

  • In-hospital mortality prediction using an autoencoder
  • In-hospital mortality prediction using Bayesian target encoding
  • In-hospital mortality prediction using leave one out encoding
  • In-hospital mortality prediction using mean target encoding

Deep learning in Electronic Health Records - Summative Quiz

    • 1 Quiz
Show info about module content

1 Quiz

  • End of course summative quiz

Auto Summary

Dive into the principles of Deep Learning with a focus on Electronic Health Records (EHR) in this professional-level course by Coursera. Tailored for data science and AI enthusiasts, it covers time-series classification, vital signals like ECG, and tackles challenges like missing values and data heterogeneity in EHR. Learn imputation techniques and encoding strategies while applying these methods to clinical prediction benchmarks using the MIMIC-III database. Available with Starter and Professional subscription options, this 1920-minute course is perfect for professionals aiming to advance their expertise in the healthcare data domain.

Fani Deligianni
Instructor

Fani Deligianni