Space tech in cancer detection

What is the link between satellite imagery of the earth's surface and breast cancer diagnosis? It is called the MED-SEG system, which uses US space agency NASA's satellite imaging technology in medical imaging to provide an accurate diagnosis of breast cancer.

The widely used mammograms often give a false negative when checking for breast abnormalities.

Women with either high breast density (excess tissue) or a family history of breast cancer are usually advised to get an MRI, which is costly, cumbersome and gives a high rate of false positive results leading to unnecessary biopsies.

The MED-SEG technology enables the doctor to study the image in greater detail and obtain more information. It does away with the need for additional and costly tests.

In response to an emailed query on how breast density affects the accuracy of mammograms, Dr Molly Brewer, a professor with the Division of Gynaecologic Oncology at the University of Connecticut Health Centre in Farmington, US, replied: “When breast density is high, the density may block out the area of increased density due to cancer or pre-cancer. We think that MED-SEG may increase our ability to see through the density and thus use a less-expensive option and potentially fewer biopsies.”

The satellite imaging software used in MED-SEG was developed over the past 25 years by NASA computer engineer James C. Tilton.

At the core of the software is a computer algorithm, the Hierarchical Segmentation Software (HSEG), which processes satellite images into a wide range of resolutions — from coarse to detailed and, finally, very finely detailed.

Digital images are made up of thousands of pixels and, similar to a single piece of a jigsaw puzzle, each pixel cannot usually provide enough information about where it fits in the overall picture.

Tilton focused on image segmentation, which groups together an image's pixels at different levels of detail, and took it a step further by grouping together spatially separated objects into region classes.

“You intuitively know about regions and region relationships because your eyes and brain form a sophisticated image processing and analysis system, which we can only very poorly replicate with an image sensor and computer. RHSEG (Recursive hierarchical segmentation) software, which is used for mammograms, is intended to be a part of a system that we hope will better replicate the sophisticated human eye/ brain image processing system by providing the computer with information about image regions and the relationships between these regions, instead of just isolated, disassociated information about individual image pixels,” explains Tilton.

For very large images and even moderately large images (such as mammograms), he uses a parallel processing “divide and conquer” approach in which the large image is divided recursively into equal smaller images.

“One recursive division divides the image into four subsections, two recursive divisions divides the image into sixteen subsections, etc. When the HSEG processing is completed on the subsections, the results are recombined and blended back together using a NASA patented approach,” he says in an email response.

This enables very rapid and accurate analysis to detect breast tumours. A 3-D image can be sliced into 2-D planes and the software applied for a detailed and accurate analysis, Tilton says.