By Mark Watts
The sun was setting as we were driving across the desert on a long stretch of highway on a hot night. As the road dipped in front of us there appeared to be a pool of water in the road. As we drove closer the mirage disappeared and only the dry pavement greeted us. In this case the mirage was an additive to the surroundings. For radiologists it is difficult to appreciate all that is visible. When they are looking for one abnormality they can completely miss another. Researchers call this phenomenon “inattentional blindness.”
In 2013, Trafton Drew, Ph.D., conducted research and published a study on the topic. The study was “The invisible gorilla strikes again; Sustained inattentional blindness in expert observations” and it was published in Psychological Science. You can do a web search for “CT and Gorilla” to find it online. In the study, the researchers asked 24 radiologists to look for lung nodules in five chest CTs that had an average of 10 nodules. In the last case they presented, the researchers inserted a gorilla that was 48 times the size of the average nodule. Eighty-three percent of the radiologist did not notice the gorilla. The researchers blamed the misses largely on “inherent limitations of human attention and perception.”
I think the fact that after reading a patient’s notes for a referring physician it is possible that focusing narrowly on one task – such as looking for lung nodules – could lead radiologists to miss a significant breast mass or abnormal lymph node.
Inattentional blindness is one of many biases that come into the reading room. Satisfaction of search or conformational bias is where you come up with an idea for what the patient has and evidence to support it. Once this bias is formed you do not keep looking for any other evidence than to support your original hypothesis.
An example would be if a patient had chest pain and you find a left lobe mass you could be blinded to the true emergency of a perforated ulcer in the abdomen.
I was recently asked by a radiologist to invent a lung screening artificial intelligence algorithm that would screen out the normal chest X-rays and present the ones with findings. He estimated that 80 percent of the chest X-rays could be diverted to a lower priority list of “Computer Assisted Normal.”
While it’s true that AI does not have the same cognitive limitations that humans do, AI is only as good as its training, programming and area of application.
The use of CAD in mammography has helped improve accuracy and consistence of treatment.
The diagnostic error rates for radiologist have been consistent since the 1940s when it was first measured by Dr. Henry Garland. A striking fact is that just about everything about medical imaging has changed since the study yet the error rate has stayed the same (3 to 4%).
When faced with a long-standing error rate, I ask myself what Dr. Edwards Deming would do. Deming, a renown instructor of quality principles of management, thought errors should be studied and that organizations can increase quality and simultaneously reduce costs (by reducing waste, rework, staff attrition and litigation) while increasing customer loyalty.
All these are worthy goals for radiologists, but I feel that we are dealing with a fundamental, neurobiological and phycological mechanism. Our human visual systems, including our eyes and brain, evolved over millions of years before radiology was ever invented. I think a human’s visual system is not optimized for radiology. With years of training, a radiologist can perform the task at an impressively high level. This in no way reflects poorly on radiologist. The critical point is that the phenomenon of inattentional blindness is experienced by everyone, and we need to better understand the situations that lead to these misses in order to avoid them.
I think that 97% is an impressive success rate, but this is health care not baseball. Radiologists face a number of challenges, including sorting through the vast amount of information available to them. The growing reliance on cross-sectional imaging easily translates into thousands of images per day that the radiologist is confronted with. When radiologists have access to accurate clinical information and specific exam indications, it can help narrow their focus and make their tasks more manageable. The original gorilla study demonstrated that focusing on a specific task can lead to significant costs, even when the unexpected findings are highly unusual. Radiologists do not actually need to keep an eye out for gorillas at their clinical practices. I would encourage additional follow-up studies, because it is important to demonstrate that unintentional blindness phenomenon applies to more realistic circumstances as well.
The heat created a perceived pool of water on the highway, it was a harmless mirage. For radiologists it is the human visual system and the known limitations that could put patients at risk and are worthy of a “sharp curve ahead” sign.
Mark A. Watts is the enterprise imaging director at Fountain Hills Medical Center.