How can I implement the definition of the disease by symptoms?

Question

Python

Hello everyone, How can I implement the definition of the disease by symptoms?
To implement what you need to choose neural networks or data mining?
Where should you start? What literature should be read?
Thanks for the replies and advice

Answer 1

IBM has long been concerned about this issue. Her brainchild is called Watson. He has already graduated from a medical university and went to work in some medical organizations.
Do you think you can do something like this on your knee and make it work on a regular PC?

Answer 2

diseases are determined not by symptoms, but by specific tests, and only for a small volume of well-studied diseases.
again, if there is a desire to determine something by the symptoms, then the neural network is useless there.
You can make a simple decision tree, based on a medical encyclopedia that lists the symptoms of diseases.
It goes without saying that if a person says that he has a headache and a temperature, then the number of possible diseases can exceed thousands of options.

Answer 3

Any implementation of supervised learning is suitable for such a case.
Put up a sign with signs and symptoms: temperature, pressure, if there is a cough, blood test, other and a verdict. Placeholder table with real patient data. Then experiment with different classification or regression algorithms.
As I see, you are the first in this topic. I highly recommend this article to figure out what exists and what's what: https://vas3k.ru/blog/machine_learning/

Answer 4

It may be useful to look at the document www.ict.edu.ru/ft/003873/neiro.pdf Chapter 5 Medical Neuroinformatics.. "Researchers trained a huge number of neural networks (167 networks for diagnosing anterior wall myocardial infarction and 139 networks for infarction of the inferior wall) on dataset of 360 electrocardiograms." And also look at the algorithms "Method of dynamic kernels with and without a teacher".

Answer 5

There is a good example of a similar site already implemented - https://healthchecker.net/ . No special technologies are needed for implementation. We need a large knowledge base, symptoms of diseases, and a lot of manual work. Therefore, if I were you, I would think a little if you need it. According to technology, neural networks can of course be adapted. I think at some stage they will even be needed. But for starters, php, mysql and some simple server are enough to make it all work.