Abstract Archives of the RSNA, 2013
SSK20-08
Classification of Osteoarthritic and Healthy Chondrocyte Patterns in Human Patellar Cartilage on Phase Contrast Computed Tomography through Topological and Geometric Features
Scientific Formal (Paper) Presentations
Presented on December 4, 2013
Presented as part of SSK20: Physics (Quantitative Imaging II)
Mahesh Nagarajan, Presenter: Nothing to Disclose
Paola Coan, Abstract Co-Author: Nothing to Disclose
Markus B. Huber PhD, Abstract Co-Author: Nothing to Disclose
Paul Claude Diemoz PhD, Abstract Co-Author: Nothing to Disclose
Christian Glaser MD, Abstract Co-Author: Nothing to Disclose
Axel Wismueller MD, PhD, Abstract Co-Author: Nothing to Disclose
Phase-contrast X-ray computed tomography (PCI-CT) has been demonstrated at achieving soft-tissue contrast with micrometer scale resolution while imaging cartilage. This study proposes to quantitatively evaluate the performance of topological and geometrical approaches in characterizing chondrocyte patterns as observed in PCI-CT of human patellar cartilage as healthy or osteoarthritic.
Five osteochondral cylinders (7 mm diameter, 3 osteoarthritic, 2 healthy) extracted from post-mortem human patellae were subject to PCI-CT at 26 keV (European Synchrotron Radiation Facility, Grenoble, France). From reconstructed CT images of the cartilage, 842 regions of interest (ROI) of size 51x51 pixels capturing chondrocyte patterns were then annotated in the radial zone of the cartilage matrix from high resolution images (voxel size: 8 x 8 x 8 μm3). Two texture analysis techniques - (1) Scaling Index Method (SIM), that estimates local scaling properties and (2) Minkowski Functionals (MF), that evaluates topological properties, were used to extract features from the ROIs. Random sub-sampling cross-validation was utilized in optimizing a support vector regression model with a radial basis function kernel for the classification task. Performance was measured using area under the Receiver-Operator Characteristic (ROC) curve (AUC) for each feature.
With the experimental conditions used in this study, the best classification performance was observed with the SIM histogram (0.95 ± 0.06) which was significantly better than the performance achieved by all Minkowski Functionals – Area (0.61 ± 0.07), Perimeter (0.85 ± 0.10) and Euler Characteristic (0.88 ± 0.09).
Our study investigates the use of advanced texture analysis techniques in images acquired with PCI-CT to quantitatively evaluate their ability in distinguishing between healthy and osteoarthritic cartilage. Our results show that geometrical features derived from SIM can capture differences in chondrocyte patterns annotated in the radial zone of knee cartilage matrix extracted from healthy and osteoarthritic specimens with high accuracy, and significantly outperform topological features derived from MF at the same task.
Computer-aided feature analysis can distinguish between osteoarthritic and healthy chondrocyte patterns in knee cartilage as seen on Phase Contrast CT imaging studies with micrometer scale resolution.
Nagarajan, M,
Coan, P,
Huber, M,
Diemoz, P,
Glaser, C,
Wismueller, A,
Classification of Osteoarthritic and Healthy Chondrocyte Patterns in Human Patellar Cartilage on Phase Contrast Computed Tomography through Topological and Geometric Features. Radiological Society of North America 2013 Scientific Assembly and Annual Meeting, December 1 - December 6, 2013 ,Chicago IL.
http://archive.rsna.org/2013/13016775.html