Many of us come across the term AI almost on a daily basis. It is a term that we associate with self-driving cars, talking robots and with future in general. Many of us in the medical fraternity are already looking at this as a potential threat.
But, what exactly does this mean to medical practice? Extreme scenarios like robots taking over surgery without human interaction is unlikely. But there are certain aspects of medicine that computers are quite good at.
One of the things computer algorithms can do well is to analyse large volume of data. One of the practical applications where AI is being used is in oncology. Most patients will have several years of medical records. They will have specific medication preferences based on previously recorded patient responses.
Evolving research evidence might require analysis of entire patient profile and then selecting the best possible treatment. IBM Watson has evolved into a platform helping clinicians decide on treatment based on clinical trial data of medicines and matching it with appropriate patient profile.
More than 40 hospitals in China are using the same for assisted clinical decision making. AI is assisting doctors in getting the right diagnosis on first patient visit.
Computer vision and machine learning
Image recognition and comparing that with a database is an existing technology. This is familiar to the users of Pinterest and face recognition used in Facebook. But emergence of machine self-learning has made further progress in image recognition.
This is useful in radiology and pathology. How do they do it? One example from radiology is its application in helping to sort out abnormal digital X rays. This is being applied in mass screenings like immigration medicals and cancer screenings.
Here machine learning enables computers to understand normal images by scanning thousands of normal X rays. Once normal is understood, images are homogenised for exposure and artifacts, artificial intelligence highlight the abnormality to human interpreter.
This has enabled computer aided simple triage in emergency rooms. CT images of aortic dissection, pulmonary embolism and intracranial hypertension are given immediate alerts enabling rapid channeling of resources.
This is also used in many large scale labs to enable human experts to focus on the abnormal while machine auto-reports and excludes the normal. Cervical cancer screening programs have also used this technology.
But, in mainstream histopathology, well-differentiated cancers that almost mimic normal tissue has caused pitfalls for even human eyes. So far, machines have not been successful to distinguish near normal appearing cancers.
How about consultations?
Babylon app is one of the pioneer apps using AI-powered chatbots to engage patients seeking primary care consultations. The app provides remote video consultation for minor illnesses. It uses chat interface which helps the user deal with simple ailments. If consultation is required, it also collects data on symptoms prior to chatting with a doctor.
Brexa app is solving the issue of poor reach of health counsellors to rural population in the developing world. It is being developed to include AI mediated chatbot in regional languages. General cancer screening and vaccination advice which were provided by health care staff.
In a nutshell
It is doubtful if AI is mature enough to understand the complexities of human behavior. However, it is evolving fast. The ability to learn from experience is what makes it all the more a reality.
Medical professionals are going to be better equipped in providing more personalised care. It can enter some specialties like radiology and pathology earlier than others. Autonomous robotic surgery is not yet here, but remote surgery and remote patient monitoring are viable today.
The message is that AI is here to stay. The profession can evolve to embrace it as an assistance tool. But those who ignore it cannot expect to continue practicing medicine in to the next decade as our forefathers did.