Tuesday, June 19, 2012

The LA Times and the scanners-are-safe interpretation

As soon as I read the LA Times' take on a new scientific article on naked scanner radiation study, I knew they had misrepresented the findings. I have now read the original research article in full, and my hunch holds.

The article, by Hoppe and Schmidt in the Biomedical Engineering Department at Marquette University, seems solid to me. The introduction nicely summarizes what studies have been done at this point. Two that were done in conjunction with the TSA (by the FDA and by Johns Hopkins Applied Physics Lab -- which is functionally a government research lab doing a lot of work requiring security clearances) had access to some form of naked scanner technology, although not necessarily the exact models used in airports. They both were assigned the task of determining whether the scanners met radiation guidelines specified by an ANSI, which is suspect. So this is where Hoppe and Schmidt take off. They note the shortcomings of these earlier studies, and also reference the "scanners aren't safe" articles from 2010.

Nonetheless, as the scanners have not been made available to anyone outside of the TSA for testing, Hoppe and Schmidt were left to use TSA-provided data for their analysis. They created a computational model for measuring radiation doses in various organs of the body. They replicated the radiation from the x-ray scanner by correlating it with the TSA-provided data first, then running the simulation on their human organ models. It is true, as the LA Times reports, that they find that the effective dose of radiation that passengers receive is below the ANSI standard, as has been claimed by the TSA.

They do not claim the scanners are, therefore, safe. Indeed, they make no judgement on this at all. They note the caveats that they did not have access to an actual machine several times throughout the paper, which means that they can not independently verify the radiation levels coming from the scanner. They also had to make some assumptions and simplifications with their computational model. Although I can't find fault with their methods, it is not the same as testing the real thing, or even a physical model, with an actual scanner, as the authors are well-aware.