MIT: Independent Activities Period: IAP

IAP 2016 Activities by Sponsor - Institute for Medical Engineering & Science



Machine Learning in Critical Care

Mohammad Ghassemi, PhD Candidate, Alistair Johnson, Postdoctoral Associate, Roger Mark, Distinguished Professor

Enrollment: Limited: Advance sign-up required
Sign-up by 12/24
Limited to 30 participants
Attendance: Participants must attend all sessions

Based on 2012 estimates, 15% of us will die in an intensive care unit (ICU). When admitted to an ICU, patients are connected to countless devices all dedicated to monitoring their physiology. Doctors are faced with the immense task of incorporating all this information in an extremely short amount of time to choose a treatment. At the moment, most of this is done in a doctor's brain, but with recent advances in machine learning, there is an unprecedented opportunity to help doctors make decisions about how to best care for their patients.

Over the course of four sessions, you will engage in hands on activities and learn how to:

(Session 1) Formulate a research question
(Session 2) Preprocess clinical data
(Session 3) Apply machine learning algorithms*
(Session 4) Compete with other attendees to build the best predictive model, with prizes for the winners

* Topics covered include discriminative and generative modelling, natural language processing, deep learning, among others.

Sponsor(s): Institute for Medical Engineering & Science
Contact: Alistair Johnson, E25-505, (617) 324-2092, AEWJ@MIT.EDU


Formulate a research question

Jan/05 Tue 01:00PM-03:00PM E25-101, Bring your laptop with pgadmin3 installed

Hour 1: Introduction to clinical data

Hour 2 - hands on activity: Extraction of data related to severity of illness

Tom Pollard - Postdoctoral Associate, Mohammad Ghassemi - PhD Candidate, Alistair Johnson - Postdoctoral Associate, Roger Mark - Distinguished Professor


Preprocess clinical data

Jan/12 Tue 01:00PM-03:00PM E25-101, Bring your laptop with MATLAB installed

Hour 1: Erroneous, missing, and imprecise data

Hour 2 - hands on activity: Preprocessing of data extracted in session 1

Mohammad Ghassemi - PhD Candidate, Alistair Johnson - Postdoctoral Associate


Apply machine learning

Jan/19 Tue 01:00PM-03:00PM E25-101, Bring your laptop with MATLAB installed

Hour 1: Machine learning techniques

Hour 2, hands on activity: Build a mortality prediction model using preprocessed data from session 2

Mohammad Ghassemi - PhD Candidate, Li-wei Lehman - Research staff, Alistair Johnson - Postdoctoral Associate


Class competition

Jan/26 Tue 01:00PM-05:00PM E25-101, Bring your laptop with MATLAB installed

Bring together all that you have learned in the past sessions to build a high performance mortality prediction model. Your algorithm will be judged using a hidden set of data, and prizes will be given out to top scorers.

Mohammad Ghassemi - PhD Candidate, Alistair Johnson - Postdoctoral Associate