Artificial intelligence (AI) is set to disrupt almost every field imaginable, from transportation to finance and beyond. One key field that AI will turn on its head is health care. In the medical field, AI will transform such areas as personalized medicine, clinical decision making and even medical insurance.

Perhaps the one area of health care which will change the fastest due to AI is radiology. AI will be key to interpreting those key medical images which look deep within us, such as CT scans, MR and X-ray images, helping doctors do what they do best: diagnose.

Why will radiology be among the first areas of medicine to be completely revolutionized by AI? What is it about medical imaging that lends itself to the magic of deep learning?

1. Radiology is visual. Medical scans are, of course, inherently visual, and AI is particularly powerful at analyzing visual images — thanks at least in part to AI technology breakthroughs developed in the service of both security and social media to recognize our faces and pick us out from crowds. The fact that radiology relies so heavily on the interpretation of visual data makes it a better fit for deep learning technologies than some of its medical counterparts — meaning that radiologists can immediately benefit from AI in ways that, for example, a psychiatrist or gastroenterologist cannot.

2. There’s an acute need. The amount of medical imaging (CT and MR) continues to increase dramatically — they accounted for 7.9% and 8.9% of all tests in 2016, respectively. However, while more scans are being carried out, the number of radiologists has plateaued. In addition, with technological advances, the resolution and number of images per scan is growing exponentially. As a result, the number of minute details to be considered grows as well. This creates an enormous need for technologies which can break through the dangerous bottleneck caused by growing workloads — and, as we know, necessity is the mother of invention. Deep learning can help assess CT and MRI scans, quickly highlighting areas of concern for radiologists to then check further, while allowing urgent scans to be assessed quicker — improving patient outcomes.

3. Radiology is tech-centric. Beyond its visual nature, radiology is already a technologically-focused field. Radiologists depend every day on a host of advanced technologies — every examination involves various advanced software systems, diagnostic monitors and workstations. Due to the tech-driven nature of their day-to-day work, radiologists are considered “early adopters.” That’s exactly why they are far more likely to adopt additional technologies powered by AI. The move from film to digital images in the 80s and the early adaptation of speech-to-text in the industry are just two early examples of the way in which radiologists have always been more adept at embracing innovation than many of their colleagues.

4. There are huge amounts of accessible data. All deep learning requires copious amounts of data to be truly effective — and in the case of radiology, this data exists in the form of endless imaging of all kinds of indications carried out over the past few decades. The challenge, of course, is gaining access to these images in a form accessible to AI algorithms. The recent openness of some medical institutions to share their anonymized data has led to a boom in this space. A good example is the recent X-ray dataset released by the National Institutes of Health containing over 100,000 images with annotations.