Blood test using neural networks?

Question

Neural networks

The task is to write a neural network that, based on the results of a general blood test, would make a recommendation about the state of the blood.
The input is given numerical parameters that characterize their element in the blood (leukocytes, hemoglobin, etc.). At the output (as I imagine) - result = one of the categories of blood conditions (normal, medium inflammatory process, high inflammatory process)
The essence of the question is, what architecture of the neural network should be used in this case?
How to organize it correctly?
I would be glad for any information or a link on the topic

Answer 1

To solve this problem, a neural network is not needed, several inequalities are needed.

Answer 2

For starters, it would be good to do an exploratory data analysis and build a baseline using linear models or decision trees. Then it will be possible to return to the question of neural networks - whether they are needed and whether you have enough data to train the neural network.

Answer 3

In order to recognize only "normal, medium inflammatory process, high inflammatory process" - as you have already correctly noticed here - not only neural networks - no ML is needed at all, three or four rules are enough, which any medical student should know by heart around the second year.
In real life, the problems of medical diagnostics using machine learning methods arise when the problem is solved extremely difficult and / or inaccurately by traditional medical methods. Well, for example - according to the results of laboratory tests, certain types of cancer are detected, according to the results of monitoring, they predict the risk of heart attacks / strokes in the near future. According to the results of the analysis of genetic material - the predisposition (risk) of morbidity with certain types of diseases or reactions to certain medications, according to information about the state of the body and the region of residence - the likelihood of catching certain infections and, accordingly, the need for preventive treatment, etc. I'm not talking about diagnosing by analyzing x-rays, ultrasound, etc.,
And according to the analysis of blood - the tasks that are actually solved - are more complicated than "there is or is not inflammation." Well, for example - determining a person's age:
https://habr.com/company/spbifmo/blog/350624/
or recognizing diabetes and its varieties:
https://datascienceplus.com/machine-learning-for-d...
Only it is extremely arrogant to hope that even the simplest task of this class can be solved without having a team of specialists with medical education.

Answer 4

www.biogenmed.uz here is a good clinic