Automated polyp measurements may support CT colonography

Most researchers believe that the majority of colorectal cancers develop from polyps. However, endoscopic polypectomy — the treatment method for surgically removing these polyps — can lead to substantial bleeding and can be life-threatening. Because the risk that a polyp will lead to cancer is strongly correlated with its size (less than one percent of those smaller than 1 cm in diameter lead to cancer), a physician’s decision whether or not to perform an endoscopic polypectomy on smaller polyps can be difficult.

As reported in the January issue of Radiology, Dr. Stuart A. Taylor and his colleagues from University College Hospital in London investigated the accuracy and reproducibility of automated computer software measurements of polyps to see whether the method could be used to support CT colonography surveillance of small polyps.

A specimen containing 27 polyps was obtained from a 16-year-old male and scanned using a CT scanner from GE Medical Systems of Milwaukee. An experienced histopathologist measured the polyps. Digital photographs were taken of each 15-cm-long section, and maps were drawn to show the relative position of each polyp within the segment to aid in CT-histologic measurement correlation.

Next, both a novice and an experienced observer estimated the diameter of each polyp using three methods: manual measurement in both 2- and 3-D with software calipers, and automated measurement with software on a stand-alone workstation from Medicsight of London.

The software could not measure seven of the polyps (all ≤6 mm in diameter). Automated measurement size was not significantly different from the true histologic size for either reader, although manual 2-D and 3-D measurements significantly underestimated polyp size. Inter- and intra-observer agreement was similar for all three methods, with measurement differences of up to 2.5 mm within the expected limits of agreement for repeated polyp measurements.

The researchers believe that automated computer software could prove beneficial in the future as a way of improving polyp measurement accuracy and of reducing observer variation. However, they feel that more research is needed on the software.