| 陈泽政,郝磊,朱丽静,赵捷,赵鑫,王博娟,土继政,张凯,王兴华.超声影像组学模型鉴别周围型肺腺癌与肺鳞状细胞癌[J].中国医学影像技术,2024,40(10):1529~1532 |
| 超声影像组学模型鉴别周围型肺腺癌与肺鳞状细胞癌 |
| Grey-scale ultrasound-based radiomics models for differentiating peripheral pulmonary adenocarcinoma and squamous cell carcinoma |
| 投稿时间：2024-02-01 修订日期：2024-05-20 |
| DOI：10.13929/j.issn.1003-3289.2024.10.015 |
| 中文关键词: 肺 腺癌 癌,鳞状细胞 超声检查 影像组学 |
| 英文关键词:lung adenocarcinoma carcinoma,squamous cell ultrasonography radiomics |
| 基金项目: |
| 作者 | 单位 | E-mail | | 陈泽政 | 山西医科大学第二医院超声科, 山西 太原 030001
山西医科大学医学影像学院, 山西 太原 030001 | | | 郝磊 | 山西医科大学第二医院超声科, 山西 太原 030001 | | | 朱丽静 | 山西医科大学第二医院超声科, 山西 太原 030001 | | | 赵捷 | 山西医科大学医学影像学院, 山西 太原 030001 | | | 赵鑫 | 山西医科大学第二医院超声科, 山西 太原 030001 | | | 王博娟 | 山西医科大学第二医院超声科, 山西 太原 030001 | | | 土继政 | 山西医科大学第二医院超声科, 山西 太原 030001 | | | 张凯 | 山西医科大学第二医院超声科, 山西 太原 030001 | | | 王兴华 | 山西医科大学第二医院超声科, 山西 太原 030001 | wangxhus@163.com |
|
| 摘要点击次数: 411 |
| 全文下载次数: 132 |
| 中文摘要: |
| 目的 观察基于灰阶超声声像图的影像组学模型鉴别周围型肺腺癌与肺鳞状细胞癌(简称肺鳞癌)的效能。方法 回顾性分析经穿刺活检病理确诊且肺超声清晰显示的88例单发周围型肺腺癌及58例单发周围型肺鳞癌患者资料,按照7∶3比例将患者分为训练集(n=103)与测试集(n=43)。基于训练集灰阶超声声像图提取、筛选可用于鉴别诊断的影像组学特征;分别以支持向量机(SVM)、线性判别分析(LDA)、logistic回归(LR)及最小绝对收缩和选择算子结合logistic回归(LASSO-LR)4种分类器建立模型,并以10折交叉验证下表现最佳者作为相应类型影像组学模型;绘制受试者工作特征曲线,计算曲线下面积(AUC),评估各模型鉴别诊断效能,并以DeLong检验进行比较,获取最大约登指数下最佳截断值的鉴别诊断准确率。结果 SVM、LDA、LR及LASSO-LR最佳影像组学模型鉴别测试集周围型肺腺癌与肺鳞癌的AUC分别为0.864、0.867、0.880及0.844,差异均无统计学意义(P均＞0.05);各模型在相应最佳截断值下鉴别诊断的准确率分别为86.05%、83.72%、88.37%及86.05%。结论 基于灰阶超声影像组学模型可用于鉴别周围型肺腺癌与肺鳞癌。 |
| 英文摘要: |
| Objective To observe the efficacy of gray-scale ultrasound-based radiomics for differentiating peripheral pulmonary adenocarcinoma and squamous cell carcinoma. Methods Data of 88 patients with single peripheral lung adenocarcinoma and 58 patients with single peripheral lung squamous cell carcinoma proved pathologically with puncture biopsy and clearly visualized with lung ultrasound were retrospectively analyzed. The patients were divided into training set (n=103) and test set (n=43) at the ratio of 7∶3. Based on gray-scale ultrasound of training set, radiomics features associated with differential diagnosis of peripheral lung adenocarcinoma and lung squamous cell carcinoma were extracted and screened. Using 4 different classifiers, including support vector machine (SVM), linear discriminant analysis (LDA), logistic regression (LR) and the least absolute shrinkage and selection operator combined with logistic regression (LASSO-LR), 4 corresponding radiomics models were obtained, and the relative best models were selected according to their performances under 10-fold cross validation. The receiver operating characteristic curves were drawn, the areas under the curve (AUC) were calculated to evaluate the differentiating efficacy of each model, and DeLong test was used for the comparison. The differentiating accuracy of models were obtained under the best cutoff value with the maximum Youden index. Results The AUC of SVM, LDA, LR and LASSO-LR radiomics models for differentiating peripheral lung adenocarcinoma and lung squamous carcinoma in test set was 0.864, 0.867, 0.880 and 0.844, respectively, and no significant difference was found among 4 models (all P＞0.05). Under the best cutoff value of each model, the corresponding accuracy of SVM, LDA, LR and LASSO-LR radiomics models for differentiating peripheral lung adenocarcinoma and lung squamous cell carcinoma was 86.05%, 83.72%, 88.37% and 86.05%, respectively. Conclusion Radiomics models based on gray-scale ultrasound could be used to differentiate peripheral lung adenocarcinoma and lung squamous cell carcinoma. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
