healthcare ai ideas(disease diagnosis) - how to go about it?
hi all Below are few ideas I have. Please help me to give ideas and how to take first step if you also think these are good ones (I am based out of mumbai, India). 1) Diagnosis of diabetes using android app. As you would have already noticed, Google had collaborated with Shankara eye hospital Bangalore to effectively diagnose diabetes probability using machine learning and retina database scan. How do I take this forward? It might be very useful for everybody not only in developing but also in developed countries. 2) On similar themes, both Thyroid as well as effective radiology scan can be better done using ai/ml. Please help and comment for inputs and help. Thank you.
thanks for the needful
For now I am mostly stuck at data collection for ai. Whom and where to approach. All other parts are easy and cheap. It might be similar to https://healthitanalytics.com/news/how-artificial-intelligence-is-changing-radiology-pathology
Get in touch with hospital administrators at places that work with non-medical students. Try contacting chief medical officers at small city hospitals--they don't get much fan mail and have less resources than their big city counterparts, so will be more interested in your amateur tech project. With any luck, they'll push your project to their healthcare network and you'll get a lot more data Expect that data will still be limited though, as sharing information is legally difficult in healthcare. They'll need to make a significant effort to clean up personal identification stuff from it, or you'll need to jump through some very expensive hoops yourself to be allowed to see it
Healthcare is a hard industry. Lots of regulation and red tape. I would join a big company that has lawyers otherwise you’ll be spending most of your time dealing with regulators and not doing interesting technical work.
Both ideas are horrible. The average consumer doesn’t download an app for healthcare if they are healthy. So you collecting enough data to warn them that they have diabetes or thyroid problems is kinda funny. Sure you can go and say something about using smart watch or passive collection of data... but you can only collect so much. If you can do a hardware that actually diagnoses because it’s able to sense new bio markers, you will be very very rich. Current info picked up by smart watches are laughable for diabetes and thyroid problems. Even cardiology is very limited. When they feel something funny, they go to see the doc or google search. Diagnosing diabetes and thyroid issues are not that hard. It takes some blood tests. And if they don’t have access to basic blood test, they probably don’t have access to retina scans either. Diagnosing by symptoms is not strong enough because in medicine, shit tons of diseases have the same symptoms so without lab tests and scans, it’s hard to give accurate differentials. Radiology is one area where ML could be very good because you have repeated large sample of data and the logic behind differentials can be standardized. So look into that. There are tons of companies in that area though. Healthcare is really hard because people solving the problems a lot of times don’t really understand the real problems and don’t understand workflow or medicine. Regulations can be overcome. Product market fit is harder than technicals.
The hard part is getting data of course. You should think about kidnapping millions of people, infecting them with random diseases, send hitting them with random treatments to build a training set
Are you an ML engineer or what? You didn’t say much about your own background. I’m working on the same topic right now actually
Use 'ai', along with a mobile phone app, to let people take pics of their dick, and tell them if it's small or not. Here's a twist though, then we blackmail them for $599 in cryptocurrency, if they don't want us to send their dick pick to all their contacts.
idea would not be sustainable as google/apple would soon ban this app from playstore/appstore :-p