1
$\begingroup$

I am working on a project with the aim of assisting Doctor in his Medical Diagnosis. The data which is available is the set of questions for various diseases asked by doctors who are expert in diagnosing the respective diseases. Data available is in this format:

The learning algorithm is expected to learn the pattern of question asked for a particular disease and when a patient comes,depending on his answer for a particular question the system will tell which question to ask next and so on. Idea is that doctor has some idea of the disease patient might have and this kind of system will simulate second opinion from expert doctor on that disease. This will help the doctor in confirming his diagnosis.

My problem is how to formulate this in Machine Learning model? I can think of Learning a Decision Tree but how to generate questions from it?

$\endgroup$
  • $\begingroup$ If doctor is thinking of diagnosis X, he/she will ask questions related to X and may also give diagnosis a to patient. A model trained on such conversations may only confirm that what the doctor is thinking of, is what the doctor is thinking of. $\endgroup$ – Soren Havelund Welling Oct 8 '15 at 7:49
  • $\begingroup$ Not all relevant questions are asked by all doctors. This system might be useful in education or for assisting inexperienced doctors if it includes questions that are relevant for many differential diagnoses. $\endgroup$ – JonB Oct 8 '15 at 7:59
  • $\begingroup$ @JonB That's the idea, it will assist inexperienced doctors. $\endgroup$ – SHASHANK GUPTA Oct 8 '15 at 8:16
1
$\begingroup$

This is probably more appropriate as a comment but I don't have enough reputation points to do that. Last summer's Amer Statistical Assoc Joint Statistical Meeting in Seattle hosted several sessions devoted to machine learning and health care. Here's a link to one of the programs but if you poke around on the site, there were more:

https://www.amstat.org/meetings/jsm/2015/onlineprogram/ActivityDetails.cfm?SessionID=211265

As well as a link to U Chicago's Center for Health Statistics...

http://www.healthstats.org/aboutCHS.html

| cite | improve this answer | |
$\endgroup$
0
$\begingroup$

This kind of expert system is often built by using Bayesian Networks which model the probabilities of questions, answers and diagnoses and their dependencies. As a basic first read whether this is helpful for you, you can check out the Wikipedia article.

If you want to learn these techniques in depth, I recommend the superb book by Koller and Friedman, "Probabilistic Graphical Models".

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.