新型半自动CT身体成分分析软件验证研究
本研究在30例患者L3层面腹部CT上,由两名阅片者分别使用新型半自动软件BodyCompSlicer与商用验证软件Slice-O-Matic测量骨骼肌、皮下脂肪和内脏脂肪面积。BodyCompSlicer的观察者间ICC为0.997-1.000、COV为0.8%-1.5%,软件间ICC为0.991-0.998、COV为2.0%-3.9%,分析30例耗时70-100分钟,短于Slice-O-Matic的150-180分钟。结论认为BodyCompSlicer与Slice-O-Matic相当。
要点
- BodyCompSlicer半自动分割L3层面皮下脂肪、骨骼肌和内脏脂肪,观察者间ICC达0.997-1.000。
- 软件间ICC为0.991-0.998,COV为2.0%-3.9%,与商用软件Slice-O-Matic结果相当。
- 分析30例耗时BodyCompSlicer 70-100分钟 vs Slice-O-Matic 150-180分钟。
核心结论
- BodyCompSlicer在CT身体成分分析方面与Slice-O-Matic相当。
研究设计与样本
- 验证研究,纳入30例患者L3层面腹部CT,由两名阅片者分别使用半自动软件BodyCompSlicer和手动分割软件Slice-O-Matic测量骨骼肌、皮下脂肪和内脏脂肪面积,以ICC和COV评估观察者间及软件间变异性。
诊断标准与关键数值
- BodyCompSlicer观察者间:骨骼肌ICC 0.997、COV 1.5%;皮下脂肪ICC 1.000、COV 0.8%;内脏脂肪ICC 1.000、COV 1.0%。
- Slice-O-Matic观察者间:骨骼肌ICC 0.993、COV 2.5%;皮下脂肪ICC 0.995、COV 3.1%;内脏脂肪ICC 0.999、COV 2.3%。
- 软件间:骨骼肌ICC 0.995-0.995、COV 2.0-2.1%;皮下脂肪ICC 0.991-0.994、COV 3.4-3.9%;内脏脂肪ICC 0.998-0.998、COV 2.8-3.3%。
临床应用与阅片要点
- BodyCompSlicer自动分割皮下脂肪、骨骼肌和内脏脂肪,阅片者可按需用鼠标手动修正分割。
- 分割后采用CT衰减阈值计算骨骼肌、皮下脂肪和内脏脂肪面积。
局限性
- 原文未报告局限性。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
英文原文(摘要)· 点击展开
BACKGROUND: Computed tomography (CT) has been increasingly used to quantify abdominal muscle and fat in clinical research studies, and multiple studies have shown importance of body composition in predicting clinical outcome. The purpose of study is to compare newly developed semi-automated software (BodyCompSlicer) to commercially available validated software (Slice-O-Matic) for CT body composition analysis.
METHODS: CT scans of abdomen at L3 level in 30 patients were analyzed by two reviewers and using two softwares (BodyCompSlicer and Slice-O-Matic). Body composition analysis using BodyCompSlicer was semi-automated. The program automatically segmented subcutaneous fat (SF), skeletal muscle (SM), and visceral fat (VF) areas. Reviewers manually corrected the segmentation using computer-mouse interface as necessary. Body composition analysis using Slice-O-Matic was performed by manually segmenting each area using computer-mouse interface (brush tool). After segmentation, SM, SF, and VF areas were calculated using CT attenuation thresholds. Inter-observer and inter-software variability of measurements were analyzed using intraclass correlation coefficients (ICC) and coefficient of variation (COV).
RESULTS: Inter-observer ICC and COV using BodyCompSlicer were 0.997 and 1.5% for SM, 1.000 and 0.8% for SF, and 1.000 and 1.0% for VF, whereas those using Slice-O-Matic were 0.993 and 2.5% for SM, 0.995 and 3.1% for SF, and 0.999 and 2.3% for VF. Inter-software ICCs and COV were 0.995-0.995 and 2.0-2.1% for SM, 0.991-0.994 and 3.4-3.9% for SF, and 0.998-0.998 and 2.8-3.3% for VF. Time to analyze 30 cases was 70-100 min and 150-180 min using BodyCompSlicer and Slice-O-Matic, respectively.
CONCLUSION: BodyCompSlicer is comparable to Slice-O-Matic for CT body composition analysis.
引用
- Takahashi N, Sugimoto M, Psutka SP, et al. Validation study of a new semi-automated software program for CT body composition analysis. Abdom Radiol (NY). 2017. PMID: 28389787. doi:10.1007/s00261-017-1123-6
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。