Using the ANN with mobile applications

Question

Conor O' Sullivan on 16 Oct 2014

0
Link

Direct link to this question

https://se.mathworks.com/matlabcentral/answers/158935-using-the-ann-with-mobile-applications

Answered: Greg Heath on 17 Oct 2014

Hi all,

This might sound a bit of a stupid question, but I'm doing a project where I'm using an ANN to diagnose Sepsis in patients using various vital signs as inputs. The aim is to then develop an Android app that allows a user to put in the inputs and will return a risk score (% likelihood of sepsis, etc)

I'm constructing my ANN using Matlab (using their pattern recognition network wizard), and I'm using MIT App Inventor 2 to build the App itself.

Where I'm struggling, is understanding how one would actually put the ANN to work with the app.

I've recently been informed that App Inventor 2 isn't powerful enough to host an ANN, and that a possible solution may be to use an online source, such as Google Spreadsheet, however I'm not sure how this would work either.

Another small complication is the nature of the ANN - I'm using 8 inputs to determine one of 2 outputs. However, as these are vital signs, and more useful over a time period (to get a trend), I've actually got something like 24 inputs (8x3) for each patient. Now, if I train the ANN, will that mean that I can't query it unless I have 24 inputs again?

Lots of questions here, I know, but I would really appreciate any feedback, no matter what it might be.

Thank you again,

Conor.

0 Comments
Show -2 older commentsHide -2 older comments

Sign in to comment.

Sign in to answer this question.

Answer 1

Star Strider on 17 Oct 2014

0
Link

Direct link to this answer

https://se.mathworks.com/matlabcentral/answers/158935-using-the-ann-with-mobile-applications#answer_155468

For clinical questions such as yours, I always start with a PubMed search to see what other investigators have already done. When I searched on (sepsis & diagnosis & (classifier | "neural net")) & english[la], I got these six articles. None mention neural nets but several mention support vector machines, principal component analysis, and other techniques.

Click on ‘Display Settings’, the check the box to display the abstracts. It’s an interesting site to surf, and clicking on ‘Related citations’ beneath each citation to bring up similar articles.

As a physician and biomedical engineer, I found the abstract for ‘Classification of low systemic vascular resistance using photoplethysmogram and routine cardiovascular measurements’ particularly interesting.