Davis Derik L, Kesler Thomas, Gilotra Mohit N, Almardawi Ranyah, Hasan Syed A, Gullapalli Rao P, Zhuo Jiachen
Department of Diagnostic Radiology & Nuclear Medicine, University of Maryland School of Medicine, 22 S. Greene Street, Baltimore, MD, 21201, USA.
Department of Diagnostic Radiology & Nuclear Medicine, University of Maryland Medical Center, Baltimore, MD, USA.
Skeletal Radiol. 2019 Apr;48(4):535-541. doi: 10.1007/s00256-018-3057-7. Epub 2018 Sep 10.
Quantification of rotator cuff intramuscular fatty infiltration is important for clinical decision-making in patients with rotator cuff tear. The semi-quantitative Goutallier classification system is the most commonly used method, but has limited reliability. Therefore, we sought to test a freely available fuzzy C-means segmentation software program for reliability of the quantification of shoulder intramuscular fatty infiltration on T1-weighted MR images and for correlation with fat fraction by six-point Dixon MRI.
We performed a prospective cross-sectional study to measure visible intramuscular fat area percentage on oblique sagittal T1 MR images by fuzzy C-means segmentation and fat fraction maps by six-point Dixon MRI for 42 shoulder muscles. Intra- and inter-observer reliability were determined. Correlative analysis for fuzzy C-means and six-point Dixon intramuscular fatty infiltration measures was also performed.
We found that inter-observer reliability for the quantification of visible intramuscular fat area percentage by fuzzy C-means segmentation and fat fraction by six-point Dixon MRI was 0.947 and 0.951 respectively. The intra-observer reliability for the quantification of visible intramuscular fat area percentage by fuzzy C-means segmentation and fat fraction by six-point Dixon MRI was 0.871 and 0.979 respectively. We found a strong correlation between fuzzy C-means segmentation and six-point Dixon techniques; r = 0.850, p < 0.001 by individual muscle; and r = 0.977, p < 0.002 by study subject.
Quantification of intramuscular fatty infiltration by fuzzy C-means segmentation on T1-weighted sequences demonstrates excellent reliability and strong correlation with fat fraction by six-point Dixon MRI. Quantitative fuzzy C-means segmentation is a viable alternative to the semi-quantitative Goutallier classification system.
肩袖肌内脂肪浸润的量化对于肩袖撕裂患者的临床决策至关重要。半定量的Goutallier分类系统是最常用的方法，但可靠性有限。因此，我们试图测试一款免费的模糊C均值分割软件程序，以评估其在T1加权磁共振成像（MRI）上对肩部肌内脂肪浸润量化的可靠性，以及与六点Dixon MRI脂肪分数的相关性。
我们进行了一项前瞻性横断面研究，通过模糊C均值分割测量斜矢状面T1 MR图像上可见的肌内脂肪面积百分比，并通过六点Dixon MRI测量42块肩部肌肉的脂肪分数图。确定了观察者间和观察者内的可靠性。还对模糊C均值和六点Dixon肌内脂肪浸润测量进行了相关性分析。
我们发现，通过模糊C均值分割量化可见肌内脂肪面积百分比和通过六点Dixon MRI量化脂肪分数的观察者间可靠性分别为0.947和0.951。通过模糊C均值分割量化可见肌内脂肪面积百分比和通过六点Dixon MRI量化脂肪分数的观察者内可靠性分别为0.871和0.979。我们发现模糊C均值分割和六点Dixon技术之间存在很强的相关性；按个体肌肉计算，r = 0.850，p < 0.001；按研究对象计算，r = 0.977，p < 0.002。
在T1加权序列上通过模糊C均值分割对肌内脂肪浸润进行量化显示出极好的可靠性，并且与六点Dixon MRI的脂肪分数有很强的相关性。定量模糊C均值分割是半定量Goutallier分类系统的一个可行替代方案。
