Fooling with Mother Nature benefits children


Years ago, a margarine manufacturer trying to pass its product off as the equal to butter used the slogan, It's not nice to fool Mother Nature. But fooling Mother Nature is exactly what got Chee-Liang Hoe the attention of the RSNA judges who awarded him a Resident Research Trainee Prize.

As a budding medical physicist, Hoe is interested in early detection and prevention of cancer. Under the tutelage of his thesis advisor at Duke University, Ehsan Samei, Ph.D., Hoe teamed up with pediatric radiology chief Dr. Donald Frush, the de facto father of pediatric dose reduction strategies. Mother Nature didn't stand a chance.

The importance of reducing x-ray exposure to children, particularly during CT imaging, has become accepted wisdom. But how low can the dose go without interfering with image quality? That is the question on every medical physicist's lips.

"We want to strike a balance between dose reduction and image quality," Hoe said.

Investigating this balance can be problematic because of ethical issues, the rarity of subtle lesions, and the conspicuity of lesions in different stages of contrast enhancement. Lesion simulation techniques can help overcome these problems, but attempts thus far have failed to produce realistic fakes - that is, until Hoe joined forces with Frush.

Using a simple mathematical formula and a digital in-painting technique, Hoe went to work creating fake liver lesions ranging in size from 2 mm to 6 mm. Like Thomas Edison, who experimented with several hundred filament materials for the light bulb before getting it right, Hoe had many opportunities to refine his technique. After each simulation attempt, he'd show Frush a mixture of real and faux lesions and ask his opinion. Until he could regularly fool his clinical mentor, Hoe spent many lonely hours in Duke's Advanced Imaging Laboratories building.

Hoe's study had three experienced pediatric radiologists blindly evaluate 38 single subtle liver lesions, half of them artificial. Evaluators were statistically stumped.

"Radiologists have a good memory of images they've seen before. We were surprised we did so well," he said.

Of course, a larger database could produce different results. But the young Malaysian native said his method will withstand greater scrutiny. He already has plans to fool Mother Nature in other organs such as the lungs. Researchers interested in the software can contact him at

Using simulation techniques to collect data on rare lesions enables researchers to amass large databases quicker than if they had to wait and chronicle actual occurrence. Physicians spare children exposure to x-rays while they design the best dose-reduction protocols.

Lesions in this study were spherical and appeared as 2D objects in a single CT slice. The next leg of research involves creating irregularly shaped lesions and presenting them as 3D objects in a series of three CT images. Evaluators will see the lesions appear and disappear, mimicking clinical practice.

"It will be much harder to deceive pediatric radiologists reading a series of images," Hoe said. "We've got the tools to do it. We'll just have to be very careful how we carry it out."

A former college mathematics and physics instructor in Malaysia, Hoe is the first in his family to study in the U.S. His family would like him to come home at the conclusion of his Ph.D. work in 2006, but he is undecided about his future. The States offer more opportunity for research, but he could have a potentially bigger impact educating Malaysian youth, he said.

"It took me a long time to find a field - medical physics - that I really like. I enjoy sharing scientific knowledge and life experiences with others, especially young people. It would be great to find a role in society that combines both aspects," he said.

Answer: Lesion in Figure 1 is real, while that in Figure 2 is not.

Related Videos
Reimbursement Challenges in Radiology: An Interview with Richard Heller, MD
Nina Kottler, MD, MS
The Executive Order on AI: Promising Development for Radiology or ‘HIPAA for AI’?
Related Content
© 2023 MJH Life Sciences

All rights reserved.