Magnetic Resonance Image Analysis With Fine-granularity For Knee Osteoarthritis

In this work,we proposed a magnetic resonance image(MRI)analysis method with fine-granularity for knee osteoarthritis(OA).FatSat 3D fast field echo sequence(3D FFE)and 3D balanced fast field echo(3D B-FFE)sequence MRIs of 64 knees,which included 31 healthy subjects and 33 mild OA subjects,were enrolled in this study.Trabecular microarchitecture analysis indicated that degeneration of tibiofemoral subchondral bone begins from mild OA.For further investigation,we proposed the MRI analysis with fine-granularity.MRI scans were preprocessed and decomposed into several components with finer granularity.Plenty of texture features were then extracted from different subchondral localities.Statistic test revealed that the number of features with significant inter-group difference varied along the bone-cartilage interface(BCI).According to the number of features with significant inter-group difference,several locations were identified to be potential OA onset areas.This assumption was further validated by machine learning method with several SVM(support vector machine)models.The texture features extracted from aforementioned areas were used to build these models.The results of validation experiments further verified that the subchondral bone between anterior and central femur at medial condyle could be an early onset region of OA in tibiofemoral joints.This study also made some further exploration to some related techniques,such as tibiofemoral compartment segmentation and cascade multi-granularity image decomposition.The experimental result shows that our work is effective.