Fun article.
Truth is that radiologists will look for things that make sense based on the H&P given to them by the ordering physician, and then they look at the whole image for things they really shouldn't miss. They have prototypes for what things look like, and would find them. A gorilla in someone's lung obviously wouldn't have the appearance of something that's bad, thus no reason to look closer at it. The researchers need to follow up with similar studies, where they use images that resemble some sort of pathology. I would bet money that more radiologists than not would find that.