Artificial intelligence (AI) is making fast progress in the field of Medical Imaging. Clinical adoption of AI by radiologists has gone from none to 30% from 2015 to 2020, according to a study by the American College of Radiology. In medical imaging, this might, for example, consist of telling the computer which images contain breast cancer and allowing it to learn from this dataset to find features common to those images but absent in the non-cancer images. This enhances the radiologist’s ability to find new details in the images related to a pathology that previously might not have been identified by the human eye. AI programs can therefore develop entirely new ways of interpreting medical images, sometimes in ways that are incomprehensible to humans.

This situation has triggered a duality of emotions among radiologists. On the one hand, some show enthusiasm for the potential of AI to reduce their workload and on the other hand there is a fear of losing their jobs to machines. However, what we have seen since 8 years ago, when some experts predicted that AI would replace radiologists, is that AIworks hand in hand with them toincrease their intelligence, automating redundancies and optimizing the way radiologists practice. Not just saving time but improving the diagnosis potentially preventing what could have been an easy miss.

A computer scientist at Massachusetts Institute of Technology, Regina Barzilay, developed a machine-learning algorithm that uses computers’ superior visual analysis to spot subtle patterns in mammograms that the human eye might miss.  In this way, it can predict very accurately from a mammogram whether a patient is likely to develop breast cancer up to five years later. This is highly relevant because the uncertainties of a mammogram reading could delay the diagnosis and a later diagnosis implies aggressive treatments, uncertain results and more medical expenses.

This is just one of many examples of how incorporating AI software into the radiologist’s workflow can improve the accuracy and timeliness of diagnoses, while enabling radiologists to practice their profession with maximum efficiency, accelerating their ability to deliver optimal value and enable the best patient care possible. In addition, AI systems could expand the boundaries and physician’s reach to provide better treatment in developing countries and remote regions that lack radiologists. Running an algorithm can be a cheap way to close the gap between the level of care provided in the richest and poorest areas and can even be done on a mobile phone.

Nevertheless, artificial intelligence applied to medical imaging also has many limitations. The disconnect between the way computers and humans think presents several trust issues that escalate to a legal concern. For example, if an AI gets a diagnosis wrong, it can be hard to determine whether the doctor or the program is at fault and how to regulate a machine that is constantly learning and changing.

Moreover, algorithms can be very sensitive to the characteristics of the data they were trained on; if any of the factors change, it is possible that the accuracy of the model will be reduced. Current standards for assessing breast cancer risk are much less accurate in African American women, Barzilay says, because those standards were developed primarily using scans of white women.

AI limitations should reassure radiologists who worry that machines will take their jobs, even if an algorithm is better at diagnosing a particular problem, combining it with a physician’s experience and knowledge of the patient’s individual story will lead to a better outcome. In the short term, AI algorithms are more likely to assist doctors than replace them and the best-case scenario would be for artificial and human intelligence to work synergistically.

References

1. Reardon, S. (2019). Rise of robot radiologists. Nature, 576(7787), S54-S54.
2. Wallis, C. (2019). How artificial intelligence will change medicine. Nature, 576(7787), S48-S48.
3. Siwicki, B. (2021). Mass General Brigham and the future of AI in radiology. Healthcare IT News.