Study On The Evaluation Of Knee OA Cartilage By MRI-based Cartilage Automatic Segmentation

Background and purpose:Osteoarthritis(OA)is the most common chronic progressive arthrosis caused by imbalance between breakdown and regeneration processes of articular cartilage and non-cartilage tissue,which pathologically,is characterized by articular cartilage degeneration and progressive loss,including other features such as subchondral bone marrow edema,subchondral bone sclerosis,subchondral cysts,synovitis,abnormalities of meniscus,ligament and muscle,etc.The clinical manifestations are arthralgia and progressive loss of joint function,which can lead to disability in severe cases.Its morbidity increases with obesity,population aging and sports injuries.OA seriously affects human quality of life.OA tends to occur in joints with heavy-bearing and great range of motion.The knee joint is the largest joint with greatest mobility and bears the strongest force in human body.The incidence of OA in knee joint is much higher than other joints.In the early stage of osteoarthritis,there is only cartilage components change without obvious morphological change in articular cartilage,and there is no obvious clinical symptoms from early patients,which makes it difficult to identify the disease using conventional radiological examinations.When the clinical symptoms appear,the disease often has progressed to the middle and late stage.Only conservative treatment can be used to relieve the symptoms or choose joint replacement.Magnetic resonance imaging component quantitative data and morphological quantitative data have become the main imaging indicators for studying the long-term efficacy of OA.The cartilage component quantitative data and morphological quantitative data of magnetic resonance imaging have become the main imaging indicators for the long-term efficacy of OA,among which the quantitative data of the cartilage volume and cartilage thickness based on magnetic resonance imaging have been extensively studied.Therefore,the purpose of this study was to first determine the accuracy of KneeCaP automatic cartilage segmentation method,and to compare the differences in cartilage volume between normal knee joints and various degrees of knee OA(KOA)using KneeCaP software.Materials and Methods:62 osteoarthritis and 29 control healthy subjects were recruited in this study.The cartilage lesion patients were divided into a mild-to-moderate OA group(n=29)and severe OA group(n=33)based on arthroscopy and surgery.All MR imaging was performed on a3.0 T MR scanner(MAGNETOM Spectra,Siemens AG,Erlangen,Germany)with an18-channel knee coil.Sagittal 3D PD-weighted image was acquired using a double-echo steady-state(DESS)sequence with water excitation(TR/TE of 14.8/5.3 ms,FA of 25,slice thickness of 0.6 mm,FOV of 160?160 mm~2,and matrix of 256?256).Cartilage segmentation was performed using a prototype KneeCaP software(Siemens Healthcare,Erlangen,Germany)automatically.The bone-cartilage interface(BCI)was firstly determined using image information and an elastic bone-cartilage model which has a bone statistical shape model and cartilage thickness statistics.Femur,patella and tibia were then segmented after BIC extracted and were used to initialize cartilage segmentation.The cartilage volume of femur,patella,and tibia was reported after the segmentation by KneeCaP software.The relative cartilage volume percentage was determined by using the cartilage volume value divided by the total cartilage volume value.Statistical significance was determined with the sample t-test,one-way analysis of variance(ANOVA).Result:Comparing manual segmentation and automatic segmentation,there was no statistically significant difference between the two groups(P>0.05).The relative cartilage volume percentages of the femur,tibia,and patella in the normal control/mild-to-moderate OA/severe OA groups were 57.28%/59.30%/62.45%(femur),25.35%/23.46%/21.84%(tibia),and 17.37%/17.24%/15.71%(patella),respectively.Compared to the normal control group,the relative tibia cartilage volume percentage was lower in the mild-to-moderate OA group and the severe OA group(normal vs.mild-to-moderate OA:P=0.000;normal vs.severe OA:P=0.000).The corresponding index showed a similar difference between the mild-to-moderate OA group and the severe OA group(P=0.000).Compared with normal control group,the relative patella cartilage volume percentage in the severe OA group was lower,and the corresponding index showed a similar difference between the mild-to-moderate OA group and the severe OA group(normal vs.severe OA:P=0.003;mild-to-moderate OA vs.severe OA:P=0.003).Conclusions:There was no statistical difference between manual and automatic segmentation results in normal and mild to moderate OA groups,which showed that KneeCaP software is reliable for automatic segmentation of normal or mildly deformed cartilage.For patients with severe OA,manual editing of automatic segmentation results can be used to correct critical cartilage areas.Quantitative cartilage volume was obtained by using automatic segmentation software and combined with semi-quantitative scoring,which is helpful for clinical classification of OA.