Graduate Degree Type
School of Engineering
Advancements in medical imaging techniques provide biomedical researchers with quality anatomical and functional information inside preclinical subjects in the fields of cancer, osteopathic, cardiovascular, and neurodegenerative research. The throughput of the preclinical imaging studies is a critical factor which determines the pace of small animal medical research. The time involved in manual analysis of large amount of imaging data prior to data interpretation by the researcher, limits the number of studies in a time frame.
In the proposed solution, an automated image segmentation method was used to segment individual vertebrae in mice. Individual vertebrae of MOBY atlas were manually segmented and registered to the CT data. The PET activity for L1-L5 vertebrae was measured by applying the CT registered atlas vertebrae ROI. The algorithm was tested on three datasets from a PET/CT bone metastasis study using 18F-NaF radiotracer.
The algorithm was found to reduce the analysis time threefold with a potential to further reduce the automated analysis time by use of computer system with better specification to run the algorithm. The manual analysis value can vary each time the analysis is performed and is dependent on the individual performing the analysis. Also the error percent was recorded and showed an increasing trend as the analysis moves down the spine from skull to caudal vertebrae. This method can be applied to segment the rest of the bone in the CT data and act as the starting point for the registration of the soft tissues.
Radhakrishnan, Vineeth, "Registration and Segmentation of Multimodality Images for Post Processing of Skeleton in Preclinical Oncology Studies" (2016). Masters Theses. 809.