Prescott Jeffrey W, Priddy Mike, Best Thomas M, Pennell Michael, Swanson Mark S, Haq Furqan, Jackson Rebecca D, Gurcan Metin N
Department of Biomedical Informatics, Ohio State University, Columbus, OH 43210, USA.
Annu Int Conf IEEE Eng Med Biol Soc. 2009;2009:6360-3. doi: 10.1109/IEMBS.2009.5333260.
In this paper we explore a method of segmentation of muscle interstitial adipose tissue (IAT) in MR images of the thigh. The objective is to apply the method towards research into biomarkers of osteoarthritis (OA). T1-weighted images of the thigh are intensity standardized through bias field correction and intensity normalization. IAT within the thigh muscles is then segmented using a threshold combined with morphological constraints applied on connected regions in the thresholded image. The morphological constraints can be adjusted to allow for highly sensitive or highly specific IAT segmentation. The use of the morphological constraints improved the specificity of IAT segmentation over a threshold segmentation method from 0.54 to 0.67, while retaining a nearly equivalent sensitivity of 0.82 compared to 0.84. We then present a preliminary statistical analysis to demonstrate the application of the automated IAT segmentation. Finally, we specify a protocol for further exploration of IAT by leveraging the massive imaging dataset of the Osteoarthritis Initiative (OAI).
在本文中，我们探索了一种在大腿磁共振成像（MR）图像中分割肌肉间质脂肪组织（IAT）的方法。目的是将该方法应用于骨关节炎（OA）生物标志物的研究。通过偏置场校正和强度归一化对大腿的T1加权图像进行强度标准化。然后，使用阈值结合形态学约束对阈值图像中的连通区域应用于大腿肌肉内的IAT进行分割。形态学约束可以进行调整，以实现高度敏感或高度特异的IAT分割。与阈值分割方法相比，形态学约束的使用将IAT分割的特异性从0.54提高到了0.67，同时与0.84相比，保持了近乎相同的0.82的敏感性。然后，我们进行了初步的统计分析，以证明自动IAT分割的应用。最后，我们通过利用骨关节炎倡议（OAI）的大量成像数据集，指定了进一步探索IAT的方案。
