目的

通过构建双模态超声特征和影像组学联合的列线图模型，评估其在鉴别周围型肺局灶性病变良恶性中的价值。

方法

回顾性收集2017 年10 月至2023 年10 月于山西医科大学第二医院经超声引导下穿刺活检病理证实的周围型肺局灶性病变患者306 例，其中良性病变患者143例，恶性病变患者163 例，所有患者均行二维超声及超声造影检查。按照7 ∶3 的比例将患者随机分为训练集（214 例）和验证集（92 例），对双模态超声图像进行影像组学特征提取。采用最小绝对收缩和选择算子（LASSO）回归筛选最优特征并构建影像组学评分（Rad-score）；使用单因素及多因素Logistic 回归分析筛选鉴别诊断的独立危险因素，分别建立临床模型、Rad-score 及联合列线图模型。其中联合列线图模型是在多因素分析的基础上同时纳入Rad-score 和临床独立危险因素，而临床模型仅纳入临床独立危险因素。采用受试者工作特征曲线、校准曲线及临床决策曲线评估各模型的鉴别效能。

结果

通过特征降维和筛选，共得到14 个影像组学特征。多因素Logistic 回归分析表明，年龄、病灶形态、病灶与肺组织增强时间差、增强模式及Rad-score 是判断周围型肺局灶性病变良恶性的独立危险因素，基于以上因素建立的联合列线图模型在训练集和验证集中的曲线下面积分别为0.894、0.917，均高于临床模型和Rad-score 单独应用时的诊断效能。校准曲线显示列线图模型预测概率与病理结果间具有良好的一致性；临床决策曲线表明列线图模型在较大阈值范围对鉴别诊断周围型肺局灶性病变良恶性提供了更高的临床价值。

结论

将二维超声、超声造影特征与影像组学联合，可较好地鉴别周围型肺局灶性病变的良恶性，为临床诊断提供依据。

Objective

To develop a nomogram based on dual-modal ultrasound features and radiomics, and to evaluate its value in differentiating benign and malignant peripheral pulmonary focal lesions.

Methods

A total of 306 patients with peripheral pulmonary focal lesions confirmed by ultrasoundguided biopsy in the Second Hospital of Shanxi Medical University from October 2017 to October 2023 were retrospectively collected, including 143 patients with benign lesions and 163 patients with malignant lesions.All patients underwent two-dimensional ultrasound and contrast-enhanced ultrasound.At a ratio of 7:3, the patients were randomly divided into a training set (214 cases) and a validation set (92 cases),and extraction of radiomics feature from dual-modal ultrasound images was performed.The least absolute shrinkage and selection operator (LASSO) regression was used to screen the optimal features and construct the radiomics score (Rad-score).Univariate and multivariate Logistic regression analyses were performed to screen independent risk factors for differential diagnosis, and a clinical model and a nomogram model were developed.The nomogram model included both Rad-score and independent clinical risk factors identified by multivariate analysis, while the clinical model only included independent clinical risk factors.Receiver operating characteristic curve, calibration curve, and clinical decision curve analyses were used to evaluate the efficiency of each model.

Results

A total of 14 radiomics features were obtained by feature dimension reduction and screening.Multivariate Logistic regression analysis showed that age, lesion morphology,lesion-lung tissue enhancement time difference, enhancement mode, and Rad-score were independent risk factors for judging the benign and malignant nature of peripheral pulmonary focal lesions.The area under the curve of the nomogram based on the above factors in the training set and validation set was 0.894 and 0.917, respectively, which was higher than that of the clinical model and Rad-score alone.The calibration curve showed a good consistency between the predicted probability of the nomogram and the pathological results.The clinical decision curve showed that the nomogram model provided higher clinical value in the differential diagnosis of benign and malignant peripheral pulmonary focal lesions in a larger threshold range.

Conclusion

The combination of two-dimensional ultrasound, contrast-enhanced ultrasound features, and radiomics can better identify the benign and malignant peripheral pulmonary focal lesions and provide a basis for clinical diagnosis.