Curtis Caldwell PhD, Presenter: Research Consultant, Bayer AG Research collaboration, Claron Technology Inc

Katherine Anne Zukotynski MD, Abstract Co-Author: Nothing to Disclose

Christopher Sweeney, Abstract Co-Author: Research Consultant, Johnson & Johnson Research Consultant, Astellas Group Research Consultant, BIND Therapeutics, Inc Research Consultant, Blend Therapeutics Research Consultant, Bionomics Limited Research Consultant, Exelixis, Inc

Chun Ki Kim MD, Abstract Co-Author: Nothing to Disclose

Victor H. Gerbaudo PhD, Abstract Co-Author: Nothing to Disclose

Trinette Wright BEng, Abstract Co-Author: Nothing to Disclose

Gregory M. Szilagyi BS, Abstract Co-Author: Nothing to Disclose

Ingmar Bitter PhD, Abstract Co-Author: Employee, Claron Technology Inc

Sameer Zaheer MSc, Abstract Co-Author: Employee, Claron Technology Inc

Assessment of change in skeletal lesions in the spine, e.g., progression over time or response to treatment, is a common, but somewhat tedious task for the radiologist. We have developed means of automated assessment of change in vertebrae over time, both in terms of uptake of F-18-NaF on PET scans and in terms of blastic and lytic sub-region volumes on CT.

F-18-NaF PET/CT scans were acquired pre- and post-therapy for 9 men with prostate cancer. Analysis for the current study was limited to the sub-region of the PET and CT images corresponding to the spine. The software automatically aligned each patient's post-therapy spine images to the corresponding pre-therapy images. The process of spine segmentation was completely automated, with the software automatically locating the vertebrae and pedicles for the cervical, thoracic and lumbar regions of the spine using an atlas based registration. Landmarks are placed by the software adjacent to each vertebra that allow the user to identify the edges of the vertebra. These landmarks can be manipulated if necessary to change where the program has defined the edges. In the current project, we were most interested in change to trabecular bone. Thus, a step was added to automatically strip away the cortical shell of each vertebral body and define trabecular bone volumes of interest (VOIs). These VOIs were used to calculate statistical information from both CT and co-registered PET images. On average, the time taken to align the images, derive the VOIs, and extract the data for each patient from the input, non-aligned CT images, was 45 seconds using an ordinary PC. For the 9 patient scans in question, no user manipulation of the automatically defined landmarks was necessary. Overall change in F-18-NaF standard uptake values (SUVs) in metastatic lesions appeared to have significantly reduced in 4, to have increased in 3, and to have stayed stable in 2 patients.  

The technique appeared to have promise for rapid, automatic and objective assessment of change in skeletal lesions.

Further assessment in a larger patient cohort and correlation of the results with clinical outcome measures are needed.

http://abstract.rsna.org/uploads/2014/14019103/14019103_xq1w.jpg

Caldwell, C, Zukotynski, K, Sweeney, C, Kim, C, Gerbaudo, V, Wright, T, Szilagyi, G, Bitter, I, Zaheer, S, Technique for Quantitative Assessment of Change in PET and CT Images of Spine Lesions.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14019103.html