Artificial Intelligence for the Practicing Radiologist: Understand AI in Five Lessons

July 6 – 7, 2019 | 5:00pm – 8:00pm | Washington D.C.
Add to Calendar

This module explains basic AI concepts, helping radiologists to understand how the algorithms behind AI work and their limitations. Understanding AI is a critical first step before trusting AI and knowing which scenarios it can be applied to. This course offers the opportunity to learn, understand and apply AI concepts to specific problems within radiology. Throughout this course, you'll have the opportunity to learn how to:
  • Recognize foundational machine learning algorithms
  • Describe the strengths and weaknesses of each algorithm
  • Describe commonly used metrics to evaluate the performance of algorithms
  • Design a machine learning solution to a specific problem in radiology assigned to a predefined dataset

Topics Covered:

Lindqwister headshot_square.jpg

Alex Lindqwister, MD | Course Developer

Alex Lindqwister will begin a Diagnostic Radiology Residency at Stanford University Medical Center in 2023. As the 2019 Dartmouth-Hitchcock Robert Jeffery Fellow in Radiology, Alex wrote and deployed a course on artificial intelligence in radiology specifically targeting radiology residents with limited backgrounds in engineering/mathematics. Prior to medical school, Alex studied computational biology at Stanford University, writing a course on Computational Frontiers in Biology for non-scientists as a graduate student. His interests include experimental imaging, computational risk stratification, medical education and digital art.


Computing Basics

Learning Objectives:
  • Numeric representation of images.
  • Data types, feature spaces.
  • AI terminology and vernacular.


A Statistical Approach to Data: Naive Bayes

Learning Objectives:

  • Probability basics.
  • Feature selection.
  • The Naive Bayes Algorithm.
  • On Bias and Limitations.
  • Naive Bayes in radiology.


Data dimensionality: K-NN and Principal Component Analysis

Learning Objectives:

  • Data visualization and feature selection revisited.
  • Nearest Neighbor approach to data similarity.
  • Introduction to the Curse of Dimensionality.
  • Consolidation and principal component analysis.
  • Limitations.
  • Applications of KNN + PCA in radiology.


Ensemble Algorithms: Random Forest and Gradient Boosting 

Learning Objectives: 
  • Data types, revisited.
  • Ensembles and management of mixed data.
  • Random Forest and Decision Trees.
  • Error and data purity.
  • Gradient Boosting.
  • Curse of Dimensionality Revisited.
  • Limitations.
  • Applications of Ensembles in Radiology.


Linear Algorithms and Early Neurons

Learning Objectives:
  • Data visualization, revisited (hyperplanes and high dimensional partitions).
  • Linear separation: The perceptron.
  • Linear separation: Support Vector Machines.
  • Curse of Dimensionality, Revisited.
  • Limitations.
  • Applications of SVM in Radiology.