Speeding up segmentation

Prior to extended hepatectomy, the tedious process of liver volume assessment must be conducted. Automated and semiautomated liver segmentation methods, however, stand to greatly increase the efficiency of this task, as evidenced by a recent study.

The standard method of estimating liver volume by manually segmenting on consecutive CT images is time-consuming, and estimation using patient height and weight is not specific to patient anatomy. Because of these limitations, Akshat Gotra, MD, of McGill University in Montreal, and colleagues sought to validate the use of a semiautomated liver segmentation method using CT. They published their results online in Academic Radiology.

The authors described the novel semiautomated segmentation method they developed as a “3D extension of a technique developed for segmentation of femoral heads in biplane radiography.” The software first roughly delineates liver contour on several CT slices, then uses variational interpolation and 3D minimal path-surface segmentation to generate and refine the shape.

Gotra and colleagues validated the method retrospectively in a group of 41 patients who underwent liver CT for preoperative planning. Total and subsegmental liver volumes were segmented from contrast-enhanced CT images in venous phase. Two pairs of image analysts independently performed segmentations, with one pair using the semiautomated method and the other performing a manual segmentation.

Results showed the semiautomated process correlated well with the manual volume measurements. The semiautomated method also had high interreader and intrareader repeatability.

But by far the biggest advantage of a semiautomated process is time savings. Manual segmentation took an average of longer than 34 minutes, while the semiautomated method used just 8 minutes of the readers’ time.

It almost goes without saying that in the ever-present search for value, tools that can improve productivity by 75 percent will be greatly prized.

-Evan Godt
Editor – Health Imaging