Knee Cartilage Extraction and Bone-Cartilage Interface Analysis from 3D MRI Data Sets Jos´ e G. Tamez-Pe˜ na, Monica Barbu-McInnis and Saara Totterman VirtualScopics, 350 Linden Oaks, Rochester NY, 14625, USA ABSTRACT This works presents a robust methodology for the analysis of the knee joint cartilage and the knee bone-cartilage interface from fused MRI sets. The proposed approach starts by fusing a set of two 3D MR images the knee. Although the proposed method is not pulse sequence dependent, the first sequence should be programmed to achieve good contrast between bone and cartilage. The recommended second pulse sequence is one that maximizes the contrast between cartilage and surrounding soft tissues. Once both pulse sequences are fused, the proposed bone-cartilage analysis is done in four major steps. First, an unsupervised segmentation algorithm is used to extract the femur, the tibia, and the patella. Second, a knowledge based feature extraction algorithm is used to extract the femoral, tibia and patellar cartilages. Third, a trained user corrects cartilage miss- classifications done by the automated extracted cartilage. Finally, the final segmentation is the revisited using an unsupervised MAP voxel relaxation algorithm. This final segmentation has the property that includes the extracted bone tissue as well as all the cartilage tissue. This is an improvement over previous approaches where only the cartilage was segmented. Furthermore, this approach yields very reproducible segmentation results in a set of scan-rescan experiments. When these segmentations were coupled with a partial volume compensated surface extraction algorithm the volume, area, thickness measurements shows precisions around 2.6%. Keywords: Cartilage extraction, Multispectral Segmentation, Cartilage-Bone Interface, Deformable Models 1. INTRODUCTION Quantitative analysis of cartilage morphometry from magnetic resonance imaging (MRI) is becoming the stan- dard for the assessment of osteoarthritis(OA) for progression evaluation and treatment monitoring; but care has to be taken to ensure the validity of the end-points. 1 OA modifying drugs can take advantage of the sensitivity and specificity of the new cartilage quantification techniques which report inter class correlations(ICC) above 0.98 and coefficients of variations below 4.0% 23456 . 7 Although these new techniques have the potential to be applied into large population studies, most of them requires highly trained users to achieve the optimal perfor- mance; therefore, limiting their applicability. 1 The requirement of highly trained users can be avoided by the development of more automated segmentation algorithms. This automatization can be achieved in the future by the use MR images with very good contrast and high spatial resolution. In the mean time the use of use of advanced imaging techniques like image fusion and multi-spectral segmentation are helping in the creation of complex algorithms for the analysis of the human knee cartilage. Furthermore, the progression of computer graphics has allowed the creation of powerful editing tools that simplify the interactive segmentation required by semi-automated approaches like the one presented in this work. The use of only a single pulse sequence for cartilage segmentation limits the amount of automatization that can be achieved. Furthermore, it has been shown that a single pulse sequence is not sufficent for the accurate evaluation of the stage of OA on the knee joint 8 . 9 The best sequences that reflects the stage of structural cartilage damage are not the best for the quantification of subchondral bone, 2 fluid, or bone marrow edema. On the other hand, the applicatively of multiple pulse sequences for cartilage segmentation has been hinder by the lack of good registration techniques that address the articulated nature of the joints. Other factor that affects the applicability of multiple pulse sequences is the chemical shift artifact that make the registration Further author information: (Send correspondence to J.G.T) J.G.T.: E-mail: tamez@virtualscopics.com, Telephone:(585) 249-6231 M.B.: E-mail: monica@virtualscopics.com S.T.: E-mail: saara totterman@virtualscopics.com