Medicine and Technology Part 1

Technology is inextricably linked with medicine. Since the advent of the stethoscope, technological innovations have been rapidly applied to medical concepts to improve care, improve outcomes and make the process more efficient. As time goes on, medicine becomes more and more dependent on technology. And this trend is only going to continue. However, never has there been a technological advance that has been able to exceed the capacity of knowledge, experience, insight and judgement. That is, until now. The advent of artificial intelligence and machine learning has finally begun to impinge on the territory of clinical judgement. This is important to know because it should factor in to what you choose to do within the profession should you go into medicine.

Whilst it is still early days in the field of machine learning and artificial intelligence, we are already seeing the potential these programs have to eventually replace clinicians We have already seen machines surpass humans in games such as chess and go, what is at play here is essentially pattern recognition with limited inputs and options. This is essentially what also goes into being a good doctor, taking a number of data points from a patient and recognising the patterns that emerge from those data points. Whilst machines will be some time away from interpreting the many shades of grey in a patient’s story, they are much closer to cracking visual patterns, such as can be found in radiology.

Looking at images and recognising what is not normal is perfect for machines. What’s more, the involvement of computers in radiology has been predicted from as early as 1959 by Lee Luster when he described the value of "an electronic scanner-computer to look at chest photofluorograms, to separate the clearly normal chest films from the abnormal chest films. The abnormal chest films would be marked for later study by the radiologists."

The significant difference between artificial intelligence is its capacity to learn. That means that with greater exposure, the program will become better. For example, one such prototype has been tested by being presented with several hundred X-Rays of limbs, which were either normal or had fractures. The machine began naïve, it had no approach with which to analyse these images and indeed no pre-determined outcome. However, after seeing all the images the machine was not only able to decide which radiographs showed fractures and which ones did not, but was also able to describe the site of the fracture. Here lies the great advantage of computer learning, it capacity to improve its success with time.

The second great advantage is the rate. The analysis of those several hundred films would be done in a matter of seconds. A computer can read thousands of images a minute if it is powerful enough. A professional radiologist may read twenty thousand in a year.

The final and perhaps greatest advantage of a computer is the absence of human error. Looking at hundreds of similar images a day in a dark room results in eye strain, fatigue and boredom. Not only this, but the human brain is subject to a swathe of cognitive biases that alter our interpretations of data points including anchoring and confirmation bias. A machine however is not subject to these. It will look at each image as if it was the first and apply the same systematic approach to each one.

There is hope however. The last bastion of medicine is still subjective judgement. It is all well and good to identify when an abnormality is present, but the collection of normal variations that herald a life-threatening syndrome is another matter. So too is it one thing to identify a sub-clinical pneumonia, but in the immune-compromised patient due to chemotherapy, that can be life-threatening as well. The last refuge of medicine remains judgement and not detection.

Keep this in mind when deciding on your future profession. The one thing that is certain is that machine learning will continue to improve. So be sure to keep yourself ahead of the change.