Radiologists often use Bayesian logic as a way to analyze the marginal contribution a positive finding may have on the overall diagnostic yield. Humans are notorious for being poor Bayesian thinkers, however, according to Dr. Howard Chen, chief radiology resident at the Hospital of the University of Pennsylvania.
To combat common biases, the group created the Adaptive Radiology Interpretation and Education System (ARIES), a web-based educational decision-support application. ARIES enables radiologists to interact with a Bayesian network, a form of artificial intelligence that mathematically chains multiple Bayesian equations together and can provide probability-ranked differential diagnoses, as well as suggest further imaging or testing. However, ARIES is useful for more than just decision support, Chen said.
In the talk, the group will share how Bayesian reasoning can be utilized as a teaching tool. For example, a radiologist user could submit three differential diagnoses after looking at an MR image and identifying the image features. The Bayesian network can then process the user's input of features and the differential diagnoses out of more than 100 possible choices. Next, the network provides instant feedback on where the user is making errors, such as during feature selection or in making a differential diagnosis based on certain image features, Chen said.
"In this way, the learner must think like a radiologist compared to the traditional hard-coded multiple-choice learning tools where the test taker only has to consider four to five diagnoses and can operate by process of elimination," Chen told AuntMinnie.com. "This allows a much richer learning experience."
You can learn more for yourself by attending this late morning talk on Monday.
