Hi. I want to develop an app that can generate the most likely diagnosis from set of multiple signs/symptoms entered into output and app would predict the Most likely diagnosis by counting most common differential diagnosis (against each sign/symptom) from each input.
(Note; I dont know anything except how AI and ML works)
From where to begin?
Hi Dr. Hashim Ali Khan,
One of the good places to start learning how to develop such intelligent app is
https://www.mathworks.com/discovery/neural-network.html
https://www.mathworks.com/products/deep-learning.html
https://www.mathworks.com/help/deeplearning/getting-started-with-deep-learning-toolbox.html
Deep Learning networks are very powerful for learning representations.
In your use case I think the data set containing all the examples is the big challenge.
Do you already have the extensive data records?
Hi Hashim Ali Khan ,
You can either train model using neural networks in matlab with sample data, this way you can give the multiple signs/symptoms to the trained model. it can predict the resulted diagnosis based on the sample data you used prior for its training.
You can check this matlab tool box for deep learning :
https://www.mathworks.com/products/deep-learning.html
Other than that you can also write your own program to learn from previous predictions and improve its results. But it will take some time, as Roman Liessner mentioned the data set containing all the examples might be a challenge in your case.
Prediction accuracy might depend on the data set you use for training your model. The bigger it is, the better model works.
Well. I don't know any basics of Maltab at all. I'm a Clinician and don't have much knowledge of Maltab or Python etc.
Let me elaborate it a bit further.
I'm working on specific group of Diseases and The user would answer 7-8 questions (which system would use to generate the diagnosis)
For example age (8-10 Diseases entities associated with ppl in 20s) and acute onset (another 8 Diseases are acute) and patient is having cough (4 Disease entities associated with cough in our case). Lets say for each of the three answered question, the most common Disease entity is upper Tract infection and so on.
Nowadays, the access to machine learning is easy thanks to libraries like
scikit-learn and keras.
I still see the challenge in the data set.
How many user surveys do you expect?
I will train the model first and the app would help users in reaching a diagnosis. Users won't be doing a survey rather they would use it as helping tool in decision making.
Don't worry its not too hard, try training the model first. For the training do get a bigger sample data set so that you can have more accurate results.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence