Recent research documenting that CT scans increase the risk of cancer has biomedical engineers looking for new ways to reduce patients' exposure to ionizing radiation.
CT scans, which use multiple X-ray images to build up cross-sectional and 3-D pictures of structures inside the human body, have soared in popularity in recent decades. A study published in December's Archives of Internal Medicine found that the number of CT scans grew from 3 million in 1980 to roughly 70 million in 2007. Those 70 million scans could eventually lead to 29 000 cancers, according to the same study. For each year's use of today's scanning technology, the resulting cancers could cause about 14 500 deaths.
But the makers of CT scanners are already taking steps to reduce the radiation dose from a scan, and consequently the risks of cancer. For instance, last year GE Healthcare introduced scanners that incorporate a technique it calls Adaptive Statistical Iterative Reconstruction. An ASIR scan uses a less intense X-ray beam, which means that the resulting raw image can contain more noise. ASIR compares voxels—volumetric pixels—side by side, and if one looks too different from its neighboring voxels, it's assumed to be noise and is removed from the data. ASIR then constructs a high-quality image from the cleaned-up data.
Amy Hara, an associate professor of radiology at the Mayo Clinic in Arizona, studied liver scans made using ASIR. She found that the images were better than low-dose images made without ASIR and required 32 to 65 percent less radiation, depending on such factors as the patient's size. Taking out the noise didn't eliminate any diagnostically important details, Hara says: "We're not losing any actual data. What we're seeing is that it's removing what it's supposed to remove."
Another radiologist, James Earls of Fairfax Radiological Consultants, in Virginia, looked at heart scans done with ASIR and found high-quality images with up to a 90 percent reduction in radiation. Hara says that ASIR studies will have to be done for all the important parts of the body typically scanned—such as lungs and heads—to make sure nothing of diagnostic importance is lost.
Siemens has a similar approach that it calls Iterative Reconstruction in Image Space, which it plans to start selling in the second quarter of 2010. The general technique behind both the Siemens and GE technologies has been known for a while, but it's so computationally intensive that it wasn't practical until computers became more powerful and programmers came up with more streamlined algorithms.
Lower-dose CT scanning is a goal of researchers outside the major imaging firms, too. Ge Wang, director of the biomedical imaging division of the joint Virginia Tech–Wake Forest University School of Biomedical Engineering and Sciences, 'and his team are developing a new low-dose CT technique called interior tomography. While typical scans cover a whole area—scanning the chest to image the heart, say, or the whole head to map out a cochlear implant—interior tomography focuses on a much smaller region of interest. "Data means radiation dose, so if you reduce data, you reduce radiation dose," Wang says. Scanning only the heart, for instance, immediately cuts the X-rays by about half.
When today's CT scanners image the chest, for example, part of the data they collect includes where the body ends and the open air begins. That defined edge provides the image-building algorithm with the reference values that allow it to calibrate the rest of the image and figure out such information as how dense a bit of tissue is. With the more targeted scan used in interior tomography, that reference is missing. Instead, the technique looks for reference points within the image. An air pocket or a region of blood, even a medical implant, scatters or absorbs X-rays in known ways. Finding those points gives the algorithm a place to start building up the image.
Without the hard edges of empty space in the picture, the scanner cannot use the Fourier transform to reconstruct an image, the way standard image-processing systems do. Instead, Wang and his team turn to a trick called the truncated Hilbert transform. Ironically, although the Hilbert transform has less data to work with, the computation is more demanding. It can take 10 minutes to an hour to create an image, depending on the size, but Wang believes that in two or three years he'll find a way to do it much faster.
Researchers hope these and other efforts will cut the cancer risk from CT scans while retaining their value in finding and treating disease. The Mayo Clinic's Hara says there's a growing awareness of the radiation risk. "For a long time it was about image quality, with less of an emphasis on dose," she says. "We all know that CT's very valuable and helpful, and that's been proven. You just want to make sure there's the least risk to the patient."