PITCHME.md

Machine Learning: A Primer for Radiologists

+++

Contents

• Machine Learning Basics
• Machine Learning Workflow
• Commonly Used Models
• Notes and Tips
• Applications in Radiology

+++

Machine Learning Basics

• Unsupervised and Supervised
• Bias and Variance
• Model Specification and Training
• Training and Testing
• Example: Linear Regression
  • Input and Output
  • Parameter Estimation
  • Training and Testing

+++

Unsupervised and Supervised Learning

Unsupervised	Supervised
Aims to "summarize" data	Aims to "learn" a function between input data and output
Methods to reduce dimensionality of data	Methods to train a system for prediction
Clustering, Compression, Association Rules	Regression
+++

Unsupervised and Supervised Learning

These methods are not exclusive of each other and can compliment each other!
Example: reducing gene expression data to summarize the most important gene expression patterns of a dataset (unsupervised), which is then used to estimate survival (supervised).

+++

Firstname	Lastname	Age
Jill	Smith	50
Eve	Jackson
John	Doe

GIT Pitch +++

• Machine Learning Workflow

  • Questions, hypotheses and model choice
  • Data organization
  • Feature extraction
  • Model training and testing

• Commonly Used Models

  • Supervised
    • GLMs
    • Random Forests
    • Neural Networks
  • Unsupervised
    • PCA
    • K*means
    • t*SNE

• Notes on when to use Machine Learning

  • Prediction vs Interpretation
  • Variance and sample size
  • Pitfalls

• Applications in Radiology

+++

Next

+++

Time for a video!

+++

How about a picture?

+++

The End :)

+++

Ok, now it's the end!