Diagnostics is probably the first frontier for AI in healthcare. Much of what happens in healthcare is about collecting data (symptoms, exam data, labs, genetics, etc) and interpreting it to make determinations about a patient’s health or medical issues. We have developed great capabilities in the last century in diagnostic testing. Often, a piece of data can clinch a diagnosis: a CT scan that shows a stroke or bleed in the context of a patient with neurological issues; a urinalysis that shows a urinary tract infection in someone with lower abdominal pains; and many other examples.
As the amount of information generated by these diagnostic modalities has exploded, the work of interpreting the large number of tests and identifying all of the diverse types of abnormalities in them has become harder and longer. There is a shortage of specialists and more opportunities to miss key findings on these tests since the quality of these tests has improved and there is more detailed information available to review and interpret. While the advancement of technologies that make such high-quality diagnostics possible is very good news indeed, we need new technologies that help in interpreting results, both in terms of improving the quality of interpretation and the speed. AI to the rescue!
Medical tests like lab results often have a normal range and if something is out of the range, an abnormal result is detected and used to diagnose the cause of an issue. Sometimes, the interpretation is much more subjective and requires careful examination of an X-ray or an MRI scan, looking for abnormalities in different areas such as the bones, kidneys, abdomen, everything visible with the scan. This is a prime area for AI algorithms to assist clinicians. Another area would be when there is large amounts of information to interpret (e.g., genetics) or combination of information from different sources to figure out the right diagnosis. It is my opinion that with the explosion of medical information about each individual, soon AI will be a necessity and not a luxury to interpret what all of this information means for each individual.
In the next few posts, we will examine the applications of AI in diagnostics in different specialties such as radiology, cardiology, pathology, and more.
