利用k-means分叢演算法評估T2磁振造影影像中皮下脂肪與內臟脂肪之含量

Using k-means Cluster Algorithm for Abdominal Fat Segmentation of T2 Weighted Magnetic Resonance Images

孫靜儀(Jing-Yi Sun) ； 韓若平(Rou-Ping Han) ； 黃英明(Ing-Ming Hwang) ； 施政廷(Cheng-Ting Shih) ； 張烈賓(Lei-Bin Chang) ； 曾冠英(Guan-Ying Tseng) ； 吳杰(Jay Wu)

《中華放射線技術學雜誌》 35卷2期 (2011/06) Pp. 76-82

https://doi.org/10.30045/CJRT.201106.0002

摘要

在－高血壓、動脈硬化、糖尿病等新陳代謝症候群與腹部脂肪含量有關，因此，正確的分析腹部脂肪含量為監測代謝性症候群之重要指標。本研究設計一套自動化腹部脂肪分割演算法，應用於T2加權腹部磁振造影影像(magnetic resonance image, MRI)以分析腹部皮下脂肪與內臟脂肪含量。分割演算法可分為下列步驟：1.利用Canny演算法偵測邊緣資訊以製作外部與內部遮罩、2.採用直方圖均化法與內部遮罩區分皮下脂肪(subcutaneous adipose tissue, SAT)、3.使用k-means分叢技術區分腹部影像中軟組織、骨頭、空氣與內臟脂肪(visceral adipose tissue, VAT)之分佈。接著利用成對樣本T檢定(Paired-Sample T test)與共變數分析法(Analysis of Covariance, ANCOVA)評估手動與自動分割方法的一致性。SAT與VAT的分析結果顯示，手動與自動分割的皮爾森相關係數分別為0.99與0.97均有高度正相關。除此之外，ANCOVA的分析顯示在不同性別之下，自動分割與手動分割的一致性表現不盡相同。本研究所設計的自動分割演算法可有效測量腹部脂肪分佈並可減少手動分割的誤差，因此，此演算法應可作為脂肪計量分析的研究工具，進一步可提升新陳代謝相關疾病之診斷率。

關鍵字

脂肪 ； 影像分割 ； 磁振造影 ； k-means分叢

並列摘要

Metabolic syndrome, such as hypertension, arteriosclerosis, and diabetes, are related with the amount of abdominal fat. Thus, calculation of the abdominal fat provides an important index for preventing metabolic diseases. This research created an automatic segmentation algorithm for measurement of abdominal fat by applying T2-weighted magnetic resonance imaging (MRI). Segmentation procedure can be divided into the following steps. Step 1: fabricating the external and internal masks using Canny edge detection algorithm. Step 2: Using histogram equalization and internal mask to distinguish subcutaneous adipose tissue (SAT). Step 3: Using the k-means cluster algorithm to search the soft tissue, and the visceral adipose tissue (VAT) in the abdominal MR images. The paired-sample T test and analysis of covariance (ANCOVA) were used to analyze and compared the consistency between the manual and automatic segmentation methods. The results showed that Pearson correlation coefficients for SAT and VAT between manual and automatic segmentations were 0.99 and 0.97, respectively. The ANCOVA results showed that manual and automatic segmentations have different consistencies for male and female. The automatic segmentation algorithm can measure abdominal fat distribution efficiently and can lower the error of manual segmentation. Therefore, this approach could be used as an effective tool for quantitative research. Furthermore it could be used to prevent the metabolic related diseases.

並列關鍵字

abdominal fat ； segmentation ； magnetic resonance imaging ； k-means segmentation