Course»Course 6»Fall 2013»6.872/HST.950»Homepage

6.872/HST.950  Biomedical Computing

Fall 2013

Instructors: Peter Szolovits, Gil Alterovitz

Lecture:  TR9.30-11  (34-304)        

Engineering the Future of Medical Care: 

The “Learning Health Care System” is envisioned by the Institute of Medicine as

“one in which progress in science, informatics, and care culture align to generate new knowledge as an ongoing, natural by-product of the care experience, and seamlessly refine and deliver best practices for continuous improvement in health and health care.”

In combination with the promise of ubiquitous genetic testing, we see an opportunity to create truly personalized medicine. Technically, both visions require

We will examine various machine learning methods as they apply to health data, and discuss what is known about scaling them to big-data settings. Socio-technically, they require establishing the patient as the driver of his/her care, sensitivity to safety, provider workflows, economic incentives, confidentiality concerns, and institutional constraints.

This class introduces the constellation of engineering challenges to make this vision a reality, enables students to work with real clinical data and analysis methods, and illustrates many applications, often relying on guest experts. Domains include intensive care, genomic medicine, public health, telemedicine and health care in the developing world, clinical decision support, and scientific discovery from clinical data.

Announcements

Project report notes

A few people were asking about the project report this week, so I wanted to add some comments. First, some thoughts from Pete on this (sent to a student) that may be helpful for all:
"There is no prescribed style except what you would normally expect in a research paper:
1. Intro describing the problem to be solved, previous state of the art, motivating problems, etc.
2. Methods describing what you did, perhaps showing demo/examples, analyses, ...
3. Discussion: what you learned, what remains problematic, ...
4. Conclusion
5. References

No page min or max; write as much as is needed to clearly describe the above, but no more. I am not looking for bulk, but for good ideas and techniques."

Second, I wanted to note that the final report does provide an opportunity for students to add more/change material planned for the report if desired (e.g. to implement suggestions mentioned in class during the presentations). In terms of the report structure, past years' reports suggest that more emphasis should be put on the motivation, how the project is different from what is out there, and explanation of the method/approach in detail (at conceptual (overview), technical (algorithm), and implementation (code) levels).

Third, please send your code over to me with some reasonable documentation on how to run it (e.g. appendix of report or as document with the code). For those using virtual machines or who desire to transfer directly, I can pick up the code on Tue or another day (just email me).

Feel free to email if any questions. I have met with a number of project groups and am happy to help...

Announced on 08 December 2013  10:23  p.m. by Gil Alterovitz

Project Presentation Schedule

December 5 and 10 are the days when students will present their projects in class. Remember that the write-ups are officially due on Dec 11 (last day of classes), but we will accept late submissions without penalty through Dec 16. (Sorry, there was a typo in the assignment that gave a much earlier due date.) Most students prefer to present on the second day, but of course that's not possible. Remember that getting feedback earlier gives you more time to revise any issues that are identified.

There are seven teams of two and two solo presenters. The schedule (drawn almost at random) is as follows:

Thu, Dec 5
Dawit Zewdie, Chris Pucci
Yongwook Bryce Kim, Joohyun Seo
Sean Burke, Neena Parikh
Iris Xu, Dalesh Dharamshi
Yasemin Gokce

Tue, Dec 10
Jaya Plmanabhan, Jenny Liu
Pranav Ramkrishnan and Wesam Manassra
Nahom Workie and Denzil Sikka
Will Drevo

Each pair presentation will be at most 15 minutes, and each single will be 10 minutes. For the pairs, I suggest that you split the presentation into two logical sections, but don't repeat material in each other's presentations. This schedule should leave us a bit of time for discussion as well. Please either bring a laptop computer. If that is not possible, get in touch to make other arrangements.

Announced on 02 December 2013  8:13  a.m. by Peter Szolovits

Availability of Shortliffe textbook for MIT

Gil reminded me that the "standard" textbook of biomedical informatics, by Shortliffe, is available for anyone with an MIT certificate, at
http://link.springer.com.libproxy.mit.edu/book/10.1007%2F0-387-36278-9

We have used this as a text in the past, but the current edition, from 2006, is a bit dated. There is supposed to be a new edition coming out very soon. For some of the material we have discussed in class, the text is still a very good reference.

Announced on 08 November 2013  7:20  p.m. by Peter Szolovits

HW4 postponed

I'm certain that you will all be devastated to learn that I was unable to complete the next homework assignment, which I had intended to hand out tomorrow (10/1). My trip to DC and a disk crash both contributed. Therefore, there will be no homework assignment this week. Please use the extra time to catch up on your readings.

Announced on 30 September 2013  9:21  p.m. by Peter Szolovits

Suggestion on using icsiboost under Windows

Jaya Pimanabhan notes that it's not necessary to install cygwin to run icsiboost under Windows:


Below are the steps I took to get it to run on Windows:

What I did was download and unzip data.zip

I then downloaded the win version of iciboost into the datafolder (even though the instruction states that we need requires cygwin1.dll in same directory there is no need for it)

http://code.google.com/p/icsiboost/downloads/detail?name=icsiboost-r147-win32-cygwin.exe&can=2&q=

Then you will be able to simply run the following instructions to validate the executable is working:

icsiboost.exe -n 10 -W 1 -N ngram -S sample -V

Announced on 19 September 2013  1:13  p.m. by Peter Szolovits

View archived announcements