| 魏逸,翟建,阮元梓,吴兴旺.常规胸部CT影像组学筛查代谢不健康正常体重[J].中国介入影像与治疗学,2026,23(7):403-408 |
| 常规胸部CT影像组学筛查代谢不健康正常体重 |
| Routine chest CT radiomics for screening metabolically unhealthy normal weight |
| 投稿时间：2026-06-05 修订日期：2026-07-01 |
| DOI：10.13929/j.issn.1672-8475.2026.07.005 |
| 中文关键词: 代谢性疾病 体质量指数 体层摄影术，X线计算机 影像组学 |
| 英文关键词:metabolic diseases body mass index tomography, X-ray computed radiomics |
| 基金项目:皖南医学院校中青年科研基金（WK2022F08）、安徽省临床医学研究转化专项项目（202304295107020003）。 |
|
| 摘要点击次数: 47 |
| 全文下载次数: 6 |
| 中文摘要: |
| 目的 观察基于常规胸部CT影像组学筛查代谢不健康正常体重（MUNW）的价值。方法 回顾性纳入A中心接受胸部CT检查的1 186名体检者，按7∶3比例划分训练集［n=830，含127名MUNW及703名代谢健康正常体重（MHNW）］与校准集（n=356，含55名MUNW及301名MHNW）；另以B中心254名体检者为外部验证集（n=254，含 33名MUNW及221名MHNW）。于CT图中提取并筛选胸肌、肝脏、内脏脂肪及皮下脂肪影像组学特征并构建影像组学模型；以多因素logistic回归分析基于MUNW独立影响因素构建临床-CT模型，结合影像组学模型构建联合模型。利用受试者工作特征曲线下面积（ROC-AUC）评估各模型在训练-校准集与外部验证集的效能，以精确率-召回率（PR）曲线分析其精确性。结果 影像组学模型在训练-校准集中筛查MUNW的ROC-AUC及PR-AUC分别为0.931及0.745，在外部验证集分别为0.876及0.671，均高于临床-CT模型（训练-校准集的ROC-AUC及PR-AUC分别为0.772及0.383，外部验证集分别为0.767及0.324，P均＜0.05）而与联合模型差异均无统计学意义；其在训练-校准集ROC-AUC及PR-AUC分别为0.938及0.776，在外部验证集分别为0.890及0.682（P均＞0.05）。结论 利用常规胸部CT影像组学能有效筛查MUNW。 |
| 英文摘要: |
| Objective To observe the value of routine chest CT radiomics for screening metabolically unhealthy normal weight （MUNW）. Methods Totally 1 186 physical examinees who underwent chest CT examination at center A were retrospectively enrolled and divided into training set （n=830， including 127 MUNW and 703 metabolically healthy normal weight ［MHNW］） and calibration set （n=356， including 55 MUNW and 301 MHNW） at a ratio of 7∶3. Meanwhile， 254 physical examinees who underwent both chest and abdominal CT at center B were served as external validation set （n=254， including 33 MUNW and 221 MHNW）. Radiomics features of pectoral muscle， liver， visceral fat and subcutaneous fat were extracted and selected from chest CT to construct a radiomics model， while a clinical-CT model was constructed based on independent impact factors of MUNW screened by multivariate logistic regression analysis. Finally a combined model was established by integrating radiomics model with clinical-CT model. The area under the receiver operating characteristic curve （ROC-AUC） was used to evaluate the performance of models in training-calibration set and external validation set， while precision-recall （PR） curve analysis was performed to assess model's precision. Results Radiomics model achieved ROC-AUC and PR-AUC of 0.931 and 0.745 in training-calibration set while of 0.876 and 0.671 in external validation set， respectively， all higher than that of clinical-CT model （ROC-AUC and PR-AUC of 0.772 and 0.383 in training-calibration set， and 0.767 and 0.324 in external validation set； all P＜0.05）. ROC-AUC and PR-AUC of combined model was 0.938 and 0.776 in training-calibration set， 0.890 and 0.682 in external validation set （all P＞0.05）. Conclusion Routine chest CT radiomics could effectively screen for MUNW. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
