Abstract：
BACKGROUND: Measurements of knee cartilage thickness derived from MR images are attractive biomarkers for osteoarthritis research. Although some cross-sectional multivendor studies exist, none have employed fully automatic three-dimensional MRI analysis. Our objective was to evaluate the variations in knee cartilage thickness measurements obtained using automated methods and MRI instruments from five different vendors.
METHODS: The subjects were 10 healthy volunteers aged 22-60 years. MRI models with 3 Tesla strength from five different companies were used. Cartilage thickness was quantified fully automatically for seven regions. We hypothesized that \"the MRI model influences cartilage thickness measurements.\" Inter-measurement error, defined as the absolute difference between the targeted and median thicknesses determined by the five MRI models, was analyzed using histograms. The factors generating the largest inter-measurement error were also examined.
RESULTS: No exceptional trends attributable to a specific instrument model were observed, and the p-value from the Kruskal-Wallis test exceeded 0.05 in all seven regions. Therefore, the study hypothesis was rejected. Of the 350 measurements, the inter-measurement error was ≤0.05 mm in 53 %, ≤0.10 mm in 75 %, and ≤0.20 mm in 95 %. Analysis of the medial tibial cartilage, which had the largest inter-measurement error, revealed mis-extraction of synovial fluid as cartilage.
CONCLUSIONS: The choice of MRI model did not influence cartilage thickness measurements. Overall, 95 % of the inter-measurement errors were within 0.20 mm. The greatest error resulted from mis-extracting synovial fluid as cartilage.
METHODS: The subjects were 10 healthy volunteers aged 22-60 years. MRI models with 3 Tesla strength from five different companies were used. Cartilage thickness was quantified fully automatically for seven regions. We hypothesized that \"the MRI model influences cartilage thickness measurements.\" Inter-measurement error, defined as the absolute difference between the targeted and median thicknesses determined by the five MRI models, was analyzed using histograms. The factors generating the largest inter-measurement error were also examined.
RESULTS: No exceptional trends attributable to a specific instrument model were observed, and the p-value from the Kruskal-Wallis test exceeded 0.05 in all seven regions. Therefore, the study hypothesis was rejected. Of the 350 measurements, the inter-measurement error was ≤0.05 mm in 53 %, ≤0.10 mm in 75 %, and ≤0.20 mm in 95 %. Analysis of the medial tibial cartilage, which had the largest inter-measurement error, revealed mis-extraction of synovial fluid as cartilage.
CONCLUSIONS: The choice of MRI model did not influence cartilage thickness measurements. Overall, 95 % of the inter-measurement errors were within 0.20 mm. The greatest error resulted from mis-extracting synovial fluid as cartilage.
摘要：
背景：来自MR图像的膝关节软骨厚度的测量是骨关节炎研究的有吸引力的生物标志物。尽管存在一些横截面多厂商研究，没有人采用全自动三维MRI分析。我们的目标是评估使用自动化方法和来自五个不同供应商的MRI仪器获得的膝关节软骨厚度测量值的变化。
方法：受试者为10名健康志愿者，年龄22～60岁。使用了来自五家不同公司的具有3特斯拉强度的MRI模型。对于七个区域全自动定量软骨厚度。我们假设MRI模型影响软骨厚度测量。“相互测量误差，定义为由五个MRI模型确定的目标厚度和中值厚度之间的绝对差异，使用直方图进行了分析。还检查了产生最大测量间误差的因素。
结果：没有观察到归因于特定仪器型号的异常趋势，在所有七个区域中，Kruskal-Wallis检验的p值超过0.05。因此,研究假设被拒绝.在350次测量中，测量间误差≤0.05mm，为53%，在75%中≤0.10mm,95%≤0.20mm。胫骨内侧软骨的分析，具有最大的测量间误差，显示滑液作为软骨的错误提取。
结论：MRI模型的选择不影响软骨厚度测量。总的来说,95%的测量间误差在0.20mm以内。最大的误差是由于误提取滑液作为软骨。
方法：受试者为10名健康志愿者，年龄22～60岁。使用了来自五家不同公司的具有3特斯拉强度的MRI模型。对于七个区域全自动定量软骨厚度。我们假设MRI模型影响软骨厚度测量。“相互测量误差，定义为由五个MRI模型确定的目标厚度和中值厚度之间的绝对差异，使用直方图进行了分析。还检查了产生最大测量间误差的因素。
结果：没有观察到归因于特定仪器型号的异常趋势，在所有七个区域中，Kruskal-Wallis检验的p值超过0.05。因此,研究假设被拒绝.在350次测量中，测量间误差≤0.05mm，为53%，在75%中≤0.10mm,95%≤0.20mm。胫骨内侧软骨的分析，具有最大的测量间误差，显示滑液作为软骨的错误提取。
结论：MRI模型的选择不影响软骨厚度测量。总的来说,95%的测量间误差在0.20mm以内。最大的误差是由于误提取滑液作为软骨。
