Erik B. Dam PhD, Abstract Co-Author: Shareholder, Biomediq A/S Employee, Biomediq A/S

Souhil Zaim MD, Abstract Co-Author: Employee, SYNARC Inc

Thomas Fuerst, Abstract Co-Author: Employee, SYNARC Inc

Harry Kenneth Genant MD, Abstract Co-Author: Stockholder, SYNARC Inc Member, SYNARC Inc Board of Directors, SYNARC Inc

Mads Nielsen PhD, Presenter: Stockholder, Biomediq A/S Research Grant, Nordic Bioscience A/S Research Grant, SYNARC Inc Research Grant, AstraZeneca PLC

Martin Lillholm MSc, PhD, Abstract Co-Author: Employee, Biomediq AS Shareholder, Biomediq AS

Comprehensive analysis of the NIH funded Osteoarthritis Initiative (OAI) data has been limited by the overwhelming task of scoring the 21,000 knee MRI in the multi-visit study with 4,769 participants. We investigate the feasibility of fully automatic, computer-based segmentation and cartilage quantification on an OAI sub-population including 500 subjects.

The MRI were acquired on a Siemens 3T Trio scanner using a sagittal 3D DESS WE sequence (25º flip angle, 16ms RT, 4.7ms ET, 0.36x0.36x0.7 mm voxels, scan time 10 min).  The analysis population was selected as the first 500 subjects (sorted by patient ID) to have completed the year 0 and year 1 visits. They had age 64±9 years, BMI 28±5, with 36% women. Their Kellgren and Lawrence scores were 38%, 17%, 28%, 15%, and 1% at KL0-4.  The KneeIQ segmentation framework (from Biomediq, Copenhagen) combining statistical texture and shape models was trained on 50 knee MRI with cartilage segmentations provided by OAI (courtesy of iMorphics) and validated on another 38 MRI. For each knee, cartilage thickness maps were quantified for the five compartments: patellar and medial/lateral tibial/femoral cartilages. Mean thickness was computed for each map.

On the validation set, the linear correlations to ground truth cartilage volumes were between 0.78 and .93 for the five compartments.  Cartilage loss was observed in all compartments. Limited mean thickness changes were observed for lateral tibial, lateral femoral, and patellar cartilages (-0.2%, -1.3%, and -0.4% respectively). Larger mean thickness changes were observed for medial tibial and medial femoral cartilages (-5.9% and -5.2%). Visualization of the thickness change maps revealed very non-homogeneous progression with large differences within each compartment.

The observed cartilage loss was highly non-homogeneous within the compartments and focused in the anterior femoral regions and the central medial tibial compartment. We demonstrated the feasibility of complete, automatic quantification of the OAI knee MRI by presenting results including year 0 and year 1 scans for 500 subjects.

Comprehensive, quantitative analysis of the large epidemiological OAI study can further the currently limited understanding of OA etiology.

Dam, E, Zaim, S, Fuerst, T, Genant, H, Nielsen, M, Lillholm, M, Automatic Quantification of the Osteoarthritis Initiative Knee MRI.  Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL. http://archive.rsna.org/2012/12043512.html