Abstract
1311
Objectives Selective internal radiation treatment (SIRT) with Y-90 microspheres is emerging as an effective liver-directed therapy. The treatment involves the accurate calculation of functional tumor volumes and anatomical volumes of liver for determination of the tumor to normal liver ratio and for calculation of the dose of Y-90 microspheres. The objective was to develop a semi-automatic algorithm to segment the liver region from CT scans in contrast to the fully manual, time consuming procedure which is currently used.
Methods The developed algorithm is a hybrid using the k-means segmentation as precursor step, the results of which are fed to an advanced contouring based segmentation. This algorithm requires the user to pick 5 points of varied intensity in the scan for the k-means segmentation and a rough outline of the liver in widely spaced slices of the CT dataset for the contouring algorithm. The selection of the slices for the contouring algorithm are adaptive based on an automatic selection of only those slices which show a ‘marked difference’ from the neighboring slices. To quantify the ‘marked difference’ image statistics are employed which include a rough change in number of pixels occupied by the liver in each slice. The algorithm calculates the volumes of the liver and creates a 3D rendering of the segmented liver.
Results The volumes obtained by the algorithm were compared to the volumes obtained by manual segmentation done by an expert. These calculated volumes differ by only as much as 1-2 % as compared to the volumes obtained manually.
Conclusions The high accuracies enable the use of the algorithm in hospital settings for accurate tumor to normal liver volume ratio that when combined with functional tumor volume estimation using nuclear medicine provides for accurate SIRT treatment planning