Use Case: Automated Extraction of Relevant Anatomical Structures |in Small Animal SPECT/CT Studies

Fast, accurate extraction of quantitative anatomical and functional information from 3D data sets is a critical, and sometimes rate-limiting, step in analyzing biological questions through in vivo imaging. The ability to apply a single analysis routine for the automatic extraction of major anatomical features while maintaining the flexibility to incorporate the extraction of application-specific features helps resolve the time- limiting aspect of such analysis, increases reproducibility, and reduces observer variability.

For automated extraction of relevant anatomical structures (i.e., bones, organs), with minimal dependency on radiopharmaceutical properties, a multi-purpose rule set was created using Definiens Developer XD.

For developing the solution, six small-animal SPECT/CT data sets were used, exhibiting uptake of different markers (Tc99m-HDP, Tl201, Tc99m-GCA, In111-labeled Peptide, Tc99m-MDP and Tc99m-labeled stannous colloids) in varying major organs. The data sets were chosen for their range in distribution properties as well as their range in image quality.

For testing the rule set, it was applied to eleven further mice, imaged with Tc99m-GCA under the same protocol as the animals used for development. The segmentation routine was applied without any modification. Except for some deficiencies in classifying the uptake in the bladder in 3 animals, the algorithm properly segmented all relevant organs in each data set: brain, lung, skeleton and body from the CT and kidney, gall bladder, bladder and intestines when relevant in SPECT. Organ volumes were also in line with hand drawn ROIs. These results prove the consistent quantification of organ volume across an entire population.

This use case is based on a study conducted in cooperation with Bioscan Inc., inviCRO, LLC. and the College of Pharmacy, University of New Mexico, Albuquerque, NM, USA.