| 陈海静,杨亚英,赵卫,何波,吴莉,胡继红,李青青.增强CT影像组学-CT特征联合模型鉴别鼻腔鼻窦鳞状细胞癌与淋巴瘤[J].中国医学影像技术,2024,40(7):1003~1008 |
| 增强CT影像组学-CT特征联合模型鉴别鼻腔鼻窦鳞状细胞癌与淋巴瘤 |
| Enhanced CT radiomics-CT feature model for differentiating sinonasal squamous cell carcinoma and lymphoma |
| 投稿时间：2024-01-09 修订日期：2024-03-13 |
| DOI：10.13929/j.issn.1003-3289.2024.07.010 |
| 中文关键词: 鼻窦肿瘤 癌,鳞状细胞 淋巴瘤 列线图 体层摄影术,X线计算机 |
| 英文关键词:paranasal sinus neoplasms carcinoma, squamous cell lymphoma nomograms tomography, X-ray computed |
| 基金项目:云南省放射与治疗临床医学研究中心专项基金子课题(202102AA100067)、云南省科技厅科技计划项目(202101AY070001-103)。 |
|
| 摘要点击次数: 656 |
| 全文下载次数: 273 |
| 中文摘要: |
| 目的 观察增强CT影像组学-CT特征联合模型(联合模型)鉴别鼻腔鼻窦鳞状细胞癌(SNSCC)与鼻腔鼻窦淋巴瘤(SL)的价值。方法 回顾性收集68例SNSCC及63例SL患者,按7∶3比例分为训练集(n=92,含48例SNSCC及44例SL)与验证集(n=39,含20例SNSCC及19例SL)。以单因素分析及logistic回归分析训练集临床资料及病灶CT表现,筛选鉴别SNSCC与SL的独立预测因素并建立CT特征模型;基于训练集增强静脉期CT提取和筛选病灶最佳影像组学特征,建立影像组学模型,计算影像组学标签;基于二者构建联合模型并绘制列线图。绘制受试者工作特征(ROC)曲线,计算曲线下面积(AUC),评估各模型鉴别SNSCC与SL的效能;以校准曲线、决策曲线分析评估联合模型的校准效能及其临床获益。结果 CT所示原发病灶部位及骨质侵犯均为鉴别SNSCC与SL的独立预测因素(P均＜0.05);于增强静脉期CT筛选出3个最佳影像组学特征;分别以之构建CT模型及影像组学模型;并基于二者构建联合模型。CT、影像组学及联合模型在训练集的AUC分别为0.895、0.730及0.925,差异均有统计学意义(Z=－3.964~－1.833,P均＜0.05);在验证集的AUC分别为0.845、0.684及0.868,联合模型的AUC大于影像组学模型(Z=－2.568,P=0.010)。联合模型校准度良好,其在训练集以15%~62%及85%~92%为阈值时、在验证集以88%~95%为阈值时临床净获益较高。结论 所获增强CT影像组学-CT特征联合模型可有效鉴别SNSCC与SL。 |
| 英文摘要: |
| Objective To investigate the value of enhanced CT radiomics combined with CT features model (combined model) for differentiating squamous cell carcinoma (SNSCC) and sinonasal lymphoma (SL). Methods Totally 68 patients with SNSCC and 63 patients with SL were retrospectively collected and divided into training set (n=92, including 48 SNSCC and 44 SL) and verification set (n=39, including 20 SNSCC and 19 SL) at the ratio of 7∶3. Univariate analysis and logistic regression were used to analyze clinical data and CT manifestations in training set, and the independent predictive factors for differentiating SNSCC and SL were screened and used to construct a CT features model. Based on enhanced venous phase CT of training set, the best radiomics features of lesions were extracted and screened. The radiomics model was then established, and the radiomics label was calculated. The combined model was finally constructed based on CT model and radiomics labels, and its nomogram was drawn. Receiver operating characteristic (ROC) curve were drawn, and the areas under the curve (AUC) were calculated to evaluate the efficacy of each model for differentiating SNSCC and SL. Calibration and decision curve analysis were used to evaluate the calibration efficacy and clinical benefit of the obtained combined model. Results The primary location of the lesion and bone invasion showed on CT were both independent predictive factors for SNSCC and SL (both P＜0.05), and CT model was constructed. Based on enhanced venous phase CT, 3 best radiomics features were selected to establish the radiomics model. The AUC of CT, radiomics and combined model in training set was 0.895, 0.730 and 0.925, respectively, and significant differences of AUC were found among 3 models (Z=－3.964 to －1.833, all P＜0.05). The AUC of CT, radiomics and combined model in verification set was 0.845, 0.684 and 0.868, respectively, of combined model was greater than of radiomics model (Z=－2.568, P=0.010). The combined model had good calibration. Taken 15%-62% and 85%-92% as the thresholds in training set and 88% to 95% in validation set, the clinical net benefit of combined model was high. Conclusion The obtained enhanced CT radiomics combined with CT features model could be used to effectively differentiate SNSCC and SL. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
