UROP Openings

Have a UROP opening you would like to submit?

Please fill out the form.

Submit your UROP opening

Analyzing Novel Data To Calibrate Cognitive Health


Term:

Fall

Department:

CSAIL: Computer Science and Artificial Intelligence Lab

Faculty Supervisor:

Randall Davis

Faculty email:

davis@csail.mit.edu

Apply by:

31 August 2020

Contact:

Randall Davis, davis@csail.mit.edu

Project Description

Are you interested in the intersection of machine learning and healthcare? Do you want to work on one of the most pressing problems in worldwide healthcare? Read on. Populations around the world are “greying,” i.e., the high end of the age distribution is becoming a larger percentage of the total. This is in part the consequence advances in healthcare that allow people to live longer. But there is also a less pleasant side of this – more and more people are living long enough to be susceptible to the diseases of mental decline (e.g., Alzheimer’s) that occur in the late 60’s and beyond. The toll these diseases take is astonishing. In the US alone it is estimated that 5.8 million people suffer from some form of dementia and their care costs $290 billion annually. This cost is projected to top 1 trillion dollars by 2050. While there is as yet no cure for thesse ailments, there are ways of slowing the progress of decline. That in turn means that early detection takes on special importance – the earlier the problem can be detected, the earlier steps can be taken toward mitigation. Cognitive decline is typically measured with a battery of tests that are both verbal and written. Our research group – at MIT/CSAIL and Lahey Clinic – has been developing novel versions of traditional pen and paper neuropsychological tests, taking advantage of digital technology to extract considerably more information the test. To date we have done this with a digitizing ballpoint pen, but have been transitioning to an iPad, which offers additional interesting and novel capabilities. Where the paper form of any test is of course static, the dynamic display provided by a tablet makes possible building guidance into the testing app that ensures more valid data collection (the subject of another UROP recently posted). A tablet also enables collecting what is known as ecological data, i.e., small behaviors that, while not the major focus of the test, are also indicative of cognitive status. The test routinely gives written instructions about what to do next; unknown to the subject the system measures how long it takes the subject to read and react to those directions. We will be collecting data using the iPad app and in doing so will have a unique collection of novel data – detailed, time-stamped information about almost everything someone does when taking the test. We are looking for someone with more than an introductory level of experience in machine learning to help us find the valuable diagnostic information in this data. The two primary qualifications are previous experience in machine learning and some knowledge of signal processing. As in many machine learning projects, the chances of success here will surely be improved by careful de-noising and other cleanup of the data. While the data collection will be done on an iPad, the data will be exported and can be analyzed on any platform, with any appropriate tools. This is an Experiential Learning Project that offers an hourly wage. It requires a serious commitment of 10-12 hours per week during the term and will give you hands-on experience with an application that will see real use, and could make a substantial difference in people's lives. You will be part of a dedicated team and will have the chance to learn about the larger context in which this application sits.

Pre-requisites

Previous experience in machine learning Some experience in signal processing An interest in the intersection of ML and healthcare Enthusiasm and a willingness to push through anything that gets in the way.