For example, it is futile to differentiate between tuberculosis and cancer from an image of the chest where the answer may never lie in the image, but rather in the symptoms of the patient. Even if AI tells the physician that the ECG it is reading is normal (which the physician probably noticed on their own), the physician still has many reasons to refer the patient to a higher center.
These are isolated examples. But it is the isolation that makes these good examples. AI's role is not in isolation. AI's role is in integration. AI (or computers) should come in and fill in where humans struggle - processing large amounts of data. (Processing data, not for the sake of figuring out patterns that humans have easily learned, but for the sake of figuring out patterns, perhaps within an individual, that a human cannot easily learn by going through information)
AI can be a very good physician assistant. I have previously written about an intelligent EMR. The only barrier to using digital EMRs is the user interface. There are ways to optimize that interface. An intelligent combination of predictive suggestions, tapping rather than typing, reading data from text, etc will help.
Once physicians can start using EMRs the possibilities are endless. Here is a list of things that come to the top of my mind:
- Intelligent to-and-fro symptom/sign/examination suggestion (that physicians can use to not miss important symptoms)
- Standard treatment guidelines based suggestions on medications and investigations
- Drug interaction checker
- Locally relevant and contextual antibiotic resistance patterns
- Patient's past reports based insights, trends, analytics, etc
- Medical records exporting, highlighting important information, etc.
If you are interested in building something like this with me, let me know.
No comments:
Post a Comment