The area of health care is also overflowing with data (although privacy issues are way stronger in this area) and the trend is towards more and more data being generated. The increase in number and adoption of wearable sensors detecting life parameters and the shift towards digitalisation in the whole supply/delivery chain associated with the digitalisation of medical exams is fuelling an unprecedented data creation and at the same time is creating a culture of sharing that multiplies by millions the data available enabling all sort of correlation. The best environment for AI.
Also the digitalisation (and sequencing) of the genome that will become pervasive in the coming decade is creating another stream of data where correlation is key to create meaning.
Here a few use of Artificial Intelligence in the health care space today that may not have been under your radar:
- Antibiotic resistance is becoming a critical issue, resulting in some 100,000 death in 2018 (in US). This is going to get worse. This is going to get worse. Machine learning is being used to identify the genes in bacteria that mutate leading to antibiotic resistance, trying to understand what changes could be made in the antibiotic to keep it effective (it is a never ending story, bacteria replicates quickly and among these replications, by pure chance, some mutations will give resistance and those bacteria will take the upper hand). This is a race and artificial intelligence is used to allow researchers to become as fast in modifying an antibiotic as bacteria are in mutating. Furthermore AI is used to extract meaningful correlation among Electronic Data Records, EHR, so that antibiotic resistance can be spotted earlier.
- Image recognition/behaviour analyses coupled with EHR is making use of artificial intelligence to detect if patients in cure for detox (from drugs like opiate) are faking pain in order to get more doses (or are not following the prescribed cure). It is interesting to see AI taking the upper hand in the detection of facial clues that can be missed by a medical professional.
- Another application of AI related to image recognition is the one of reading and understanding pathology images and radiographies as an example to detect cancer. The AI is based on Machine Learning, trained on hundred thousands of images. Once trained the system may be more accurate (and way faster) than an expert doctor.
I just picked up a few examples, there are many more. Overall the big question, that is crucial in healthcare is about accountability. Who is responsible for mis-diagnoses, failure in a complex surgery procedure, wrong, ineffective medication when AI is involved? Is it the medical doctor that has not supervised the AI system, is the creator of the system, is the one who trained the system. It is a fuzzy area, where boundaries a blurred. This is further complicated by the fact that in true AI, when the system is not “intelligent” but becomes intelligent through self learning, we, humans, may be losing the capability of evaluating what is going on. And this mey be scaring.
As a last thought, we are seeing plenty of claims of AI powered wearable, including watches and bands. My personal opinion is that these are more rooted in marketing than in true artificial intelligence science. Yes they may use some neural network algorithm in some of the computations performed but to me this is not enough to say that they are based on artificial intelligence. Your comments are appreciated!