Future Care - Book Summary - Page 3 of 4 - Dr. Doug Green Dr. Doug Green

Future Care – Book Summary

AI has been in use for some time in implantable defibrillators. AI is already present in our everyday lives in the guise of Alexa , Siri, spam detection, self-driving cars, auto trading, and fraud detection. It seems ideal for medicine as it is good at handling large amounts of unstructured data.
AI image recognition has made inroads in dermatology, radiology, ophthalmology, oncology, and cardiology. Natural language processing can save doctors a lot of time at the keyboard. Doctors can use AI analysis as a second opinion when dealing with rare diseases and tricky diagnoses. In short, a doctor with an algorithm is much smarter than one without.

In addition to other AI abilities already mentioned, AI can also analyze faces and voices for a number of diseases. It still faces problems, however, as far as data quality and collection are concerned. An AI culture is developing, but it’s nowhere near where it needs to be for AI to be a routine player in diagnosis and treatment.

Beyond standard clinical data, there is also a world of environmental data that sensors can harvest such as sleep patterns, emotional data, and others. This world of big data analysis is subject to the four V’s, which are velocity, volume, variety, and veracity. More data is fine as long as it’s clean, relevant, interpretable, and actionable. You can have data without information, but you can’t have information without data.
Four-fifths of the data in EMRs are in the form of unstructured notes. This is something that the natural language processing abilities of AI can turn into actionable data. Bias is another matter that needs to be addressed. Traditionally, clinical studies have disproportionately featured data from white males. Ai based on this data is as biased at the studies themselves. Automated bias detectors are in the works and are sorely needed.

When it comes to heart implants, only one in twenty save a life. There are also many people without an implant who succumb. This means that this effort has a long way to go to even approach perfection. More than half of heart attacks and strokes occur in people who have never been predicted to have one. This is where population-specific machine learning algorithms can surpass conventional wisdom.
A simple twelve-lead ECG can forecast which of us may have a heart attack, develop atrial fibrillation, or die suddenly in the years to come. An Ai algorithm can read minor changes in the electrical signals that are not perceptible or meaningful to the human eye. We have started to use smartwatch signals to do this as well. Brain-computer interfaces have also helped paralyzed people regain function.

AI is being phased in. One thing that is more accepted by older physicians is using AI to manage patient flow through the system. This is vital as beds are often in short supply. Online symptom checkers have also gained popularity even though their accuracy is about 50%. See Bright.MD for example. By using AI up front to gather information, clinical visits can be shorter at the expense of the human touch.
Some surgeries already allow surgeons to use joysticks to manipulate cutting tools. The next step will be to turn AI surgeons loose to do the cutting. COVID gave a big boost to AI as it was used to track outbreaks and monitor social media trends.