Course syllabus

Jump to today

Lp2 HT24 (7.5 hp)
The course is offered by the Department of Computer Science and Engineering

 

Welcome to DAT635: Machine learning in healthcare!

Healthcare is facing tremendous challenges with an ageing population, a lack ofmedical staff and increased cost. At the same time, machine learning & AI are developing rapidly and are well suited to tackle some of the repetitive tasks of general and specialized care. The aim for this course is to give knowledge and understanding of learning problems in the healthcare domain and machine learning methods to solve them. It should also provide experience with applying these tools in practical problem solving on real-world health data.

Note: This is the first run of this course. Please help us improve it by being patient and leaving constructive feedback if you have it. 

 

Teaching team

Examiner:

Fredrik Johansson (Chalmers, CSE)
fredrik.johansson@chalmers.se

Teaching Assistants: 

Herman Bergström (Chalmers, CSE)
hermanb@chalmers.se

Alireza Bordbar (Chalmers, E2)
bordbar@chalmers.se

For contacting the course staff, please prioritize direct emails over Canvas-based communication.
We are rarely notified of Canvas messages and are therefore prone to late responses to these. 

Guest lecturers

Magnus Kjellberg (SU)

Patrick Royer (GU/SU)

Arvid Sjölander (Karolinska Institute)

Alexandros Rentzos (GU/SU)


Announcements

  • November 19: Recordings from lecture 3.1 have now been posted
  • November 18: The lecture on December 9 will not be in-person, only prerecorded
  • November 15: (A draft of) Chapter 3 for the lecture notes has been uploaded. More to come. 
  • November 15: A correction to Assignment 1 has been posted. The coding for the status variable was written in the wrong order. Please correct this in your submissions. 
  • November 13: Assignment 1 has now been posted.
  • November 12: Recordings from lecture 2.1 have been posted
  • November 5: Lecture notes for Module 2 have now been added.
  • November 5: The lecture on November 25 clashes with the labor market fair. I will give the lecture in person, but it will also be recorded and shared on Canvas. 
  • November 4: Recordings of the first lecture have been posted on the Module page. Apologies for the random placement of the camera in Part b. 
  • October 25: A short introduction quiz has been added


Learning objectives

The learning objectives are specified in the syllabus on Studieportalen.


Schedule

Overview: Lectures Office hours Notes:
TimeEdit
  • Mondays 10:00–11:45
  • Wednesdays 13:15-15:00
  • Before the deadline of each assignment:
    • November 20, 10:00-11:45
    • December 4, 10:00-11:45
    • December 18, 10:00-11:45
  • Room: EDIT 5128
  • Come to the office hours prepared with specific questions for the teaching assistants
  • The lecture on November 25 clashes with the labor market fair. The lecture will be recorded and shared on Canvas. 
  • The lecture on December 9 will only be pre-recorded

Course literature


Course contents

  • The course is structured as 8 different modules corresponding to different aspects of machine learning in healthcare. 
  • Each module is taught through 2 lectures

1. Introduction

2. Patient-level learning & prediction

3. Medical text & administrative data

4. Epidemiology & Population-level prediction

5. Observational epidemiology

6. Machine learning in clinical workflow

7. Uncertainty & missing values

8. Toward integration (remote)

 

Assignments & exam

  • The course contains a total of 3 graded take-home assignments (see criteria below) submitted individually through Canvas.
    • The hand-in assignments will test your ability to implement machine-learning solutions to healthcare problems.
    • This will involve analyzing provided medical data sets in Python.
    • To complete the assignments, you are expected to have access to a computer. These will not require access to a GPU. 
  • At the end of the course, in January, there will be a written exam.
    • The exam will test your knowledge of the concepts taught in the course, as well as your mathematical ability associated with solving them.
    • No aids will be permitted during the exam


Grading

  • Each completed assignment will be awarded a score in the range 0–10 points—the maximum total number of assignment points is 30. The written exam is also awarded between 0–10 points. 
  • To pass the course (grade 3), the student must obtain at least 40 % of the total points in each of the hand-in problems and at least 40 % of the total points in the exam assignment. Higher grades require, in addition to the above, that the combined score from the hand-in problems and the exam, weighted by 60 % and 40 % respectively, exceeds 60 % for grade 4 or 80 % for grade 5.

LaTeX: \mbox{Combined score} = 100 \cdot \left(0.6 \cdot \frac{\mbox{Total assignment points}}{30} + 0.4\cdot\frac{\mbox{Exam points}}{10} \right)

  • Example: Student A receives a total of 21 points on the assignments (70%) and 5 points on the exam (50%). The combined score is 0.6*70% + 0.4*50% = 62%. The student will receive the grade "4". 
  • Example: Student B receives a total of 21 points on the assignments (70%) and 4 points on the exam (40%).  The combined score is 0.6*70% + 0.4*40% = 58%. The student will receive the grade "3". 
  • Example: Student C receives a total of 24 points on the assignments (80%) and 3 points on the exam (30%).  The exam score is below 40%. The student will receive the grade "U". 


Failed assignments, resubmissions and late submissions

  • Failed assignments can be replaced by a single resubmission during each run of the course. Twice failed assignments can be completed in the next exam period. 
  • Resubmissions will be awarded a reduced number of points. The first 4 points (pass) are awarded as-is, any points beyond 4 are counted as half. For example: 8p will be counted as 4+(8-4)/2 = 6p. 
  • Only failed assignments can be resubmitted. 
  • Late assignments will be counted as resubmissions: I.e., there can only be one (no resubmission after the first), and the points are reduced in the same way. 
  • Blank assignment submissions are counted as no assignment being submitted. 


A note on cheating and use of AI tools

  • Handing in solutions to assignments copied verbatim or near-verbatim (e.g., only changing notation or variable names) from another student or from resources on the web is considered cheating. 
  • You are allowed to use AI tools, such as Chat-GPT or Copilot, to assist you in your implementation tasks provided that you disclose the nature of the use in your hand-in.

Course summary:

Date Details Due