DAIM: Images Overview

Let’s Do Digital Team

Introduction

• Welcome to the Data Analytics in Medicine (DAIM) course!
• In the coming weeks, we will be covering a range of topics from basic data visualisation all the way up to machine learning
• Please let one of the team know if you have any questions during the course or any suggestions for improvement.

Overall Course Aim

• To broaden basic Python skills into fully-fledged, practical experience that can be used to:
  • Do research
  • Build new clinical tools
• And most importantly, to directly improve patient care

Overall Course Philosophy

• I set up this course as I want to accelerate the process of learning to code as a clinician after my own experiences.
• The massive range of resources online can be intimidating.
• Knowing how to improve can be difficult.
• We want people to leave this course with a toolkit to solve real world clinical problems.
• We also want to improve the quality of cross-disciplinary work

Tips to gain the most from DAIR

• Practice, practice, practice
• Be prepared for frustration!
• Finding resources
  • Documentation
  • Stack Overflow
  • GitHub

5-minute open discussion

• Questions:
  • What reasons does everyone have for signing up for this course?
  • What is the one skill that you are hoping to gain?

Break!

Course breakdown

• The course is split into multiple modules which each cover a different area.
• Each module is split into two types of weeks:
  • Seminars - Lecture-style discussions of the topic for the week.
  • Workshops - Workbooks that you will complete with the direct or indirect support of a tutor.
• The workshops will use Jupyter Notebooks within Google Colab to deliver the content.

Learning Levels

• We’ve created a set of learning levels (LLs) for this course:
  1. Level 1 - Basic Practical Skills (coding skills)
  2. Level 2 - Theory (modelling skills)
  3. Level 3 - Project Planning and Management
  4. Level 4 - Clinical Translation

Module 2 - Course Introduction

• Seminar content will include:
  • Aims and objectives of the course
  • Overview of resources to use throughout the course
  • Clinical applications of programming

Module 2 - Core Python for Image Processing

• Seminar content will include:
  • How images are digitally represented
  • Basics of multiple Python packages
    • PIL, a Python library for working with images
    • NumPy, a Python library for working with large structured data
• Workshop content will include:
  • How to open and render different types of image data

Module 3 - Python and DICOM

• Seminar content will include:
  • What the DICOM standard is
  • How computers represent scan data
  • What data is in a DICOM file
• Workshop content will include:
  • Opening and rendering DICOM with Python
  • Extracting demographics from DICOM

Module 4 - AI for Medicine

• Seminar content will include:
  • Different types of “machine learning”
  • The practical steps needed to take an idea to a system
  • Discussing advantages and limitations of these algorithms
• Workshop content will include:
  • Preprocessing data for training
  • Constructing, training, and evaulating a basic neural network for pneumonia detection on CXR

Course Timeline

• TODO (Make this when the updated course layout is better specified)

Course Feedback

• We will be asking for course feedback to allow us to refine the course for further cohorts
• Please fill this out at the start and end of the course
• Please include honest feedback!

Note - Terminology

• We will be using the term “programming” in this course to prevent confusion with:
  • Clinical coding - SNOMED-CT, ICD-10, etc…
  • Patients coding - crash trollies, etc…

Thank you!

Any questions?