Computer generated illustration of a brain.

Artificial Intelligence and Radiology

WHAT DOES THE FUTURE LOOK LIKE?

Amber Nason • Jan 13, 2017

Will artificial intelligence (AI) make radiologists’ jobs easier and improve patient outcomes? With the hype and controversy surrounding AI and what it means for the future of radiology, it was a hot topic at RSNA 2016, held from November 27 to December 2, 2016 in Chicago, Illinois. In fact, information sessions on applications of AI and a related principle called deep learning were packed, some with standing room only.

WHAT ARE ARTIFICIAL INTELLIGENCE AND DEEP LEARNING?

Artificial intelligence describes computer systems which are able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, and decision-making.

Deep learning is a branch of machine learning which uses a set of algorithms to try to mimic the structure and function of the human brain. It is composed of many layers which are sometimes called artificial neural networks. Basically, it’s a process by which a computer figures out why something is what it is after being shown several examples.

WHAT TYPES OF APPLICATIONS COULD AI BE USED FOR IN RADIOLOGY?

Artificial intelligence has the potential to improve diagnosis and achieve better patient outcomes. By teaching a computer how to read images and what to look for, AI could potentially help:

  • Identify abnormalities and signs of disease.
  • Tap into massive streams of data to quantify the images and create reports that are effective and useful.
  • Take over during a scan to optimize the parameters and home in on pathology.
  • Recognize and make sense of visual patterns to manage large numbers of images.
  • Comb through electronic medical records for patient information that can provide context to the images.
  • Improve efficiency by prioritizing cases.
  • Provide peer-review for quality assurance.

There are a lot of potential uses for AI in the radiology industry, but currently the technology is still in its infancy.

WHAT ARE SOME OF THE BARRIERS TO AI IMPLEMENTATION?

In fact, there are a number of hurdles that AI systems are going to need to overcome in order to live up to its “hype,” something that is unlikely to happen in the near future. Below are a few of the challenges:

  1. The IT infrastructure capable of supporting such complex uses of AI is not in place, nor is it consistent throughout the industry.
  2. Deep learning is a complicated concept which will present a challenge for the testing and approval process, particularly in regards to the U.S. Food and Drug Administration (FDA) since these machines would be the first of their kind.
  3. Current versions of AI technology are being under-leveraged in terms of clinical applications. To prepare for future advances, it will be necessary to fully understand present technology.
  4. The time need to gather curated data, which will necessitate cooperation across the industry, is significant.

While these challenges are significant, there are a few companies who have already successfully tested this technology.

For example, deep learning healthcare company Enlitic fed several hundred musculoskeletal images, which were either normal or had fractures, to their machine. Not only did it flag the images with fractures, but also the site of the fracture. The machine was not told what to do; it learned by trial and error.

Another AI contender is IBM’s Watson Health, which can find clots in the pulmonary arteries. They are also looking into tasking it with summarizing electronic medical records to provide context for radiologists and looking for things that have been missed.

According to their presentations at RSNA 2016, both companies expect their technology to have a symbiotic relationship with radiologists which will help them save time and provide better care, possibly freeing them up to better engage with referring physicians.


REFERENCES

Brownlee, J. (2016) What is Deep Learning? Taken from www.machinelearningmastery.com

Freiherr, G. (2016) “Artificial Intelligence May Hold Key to Radiology's Future.” Imaging Technology News. February. Taken from www.itnonline.com

Freiherr, G. (2016) “Artificial Intelligence: Promise or Pitfall for Radiology?” Imaging Technology News. February. Taken from www.itnonline.com

Jackson, W.L. (2016) “In Radiology, Man Versus Machine.” Diagnostic Imaging. February. Taken from www.diagnosticimaging.com

Jha, S. (2016) “Will Computers Replace Radiologists?” Medscape Radiology. May. Taken from www.medscape.com

Ridley, E. L. (2016) “Will AI soon put radiologists out of a job?” AuntMinnie.com. July.

Techopedia. What does Artificial Intelligence (AI) mean? Taken from www.techopedia.com

Yeager, D. (2016) “Paging HAL: What Will Happen When Artificial Intelligence Comes to Radiology?” Radiology Today. May, Vol. 17(5): p. 12