基于增强CT影像组学特征预测难治性恶性黑色素瘤肺转移患者的免疫治疗疗效

目的 探讨基于胸部增强CT影像组学特征预测免疫治疗用于难治性恶性黑色素瘤肺转移疗效的价值.方法 回顾性分析49例难治性恶性黑色素瘤肺内转移患者,均接受程序性死亡受体(PD-1)单抗免疫治疗,采用实体瘤疗效评价标准(RECIST)1.1评价疗效,并将患者分为进展组(n=17)和未进展组[n= 32,包括稳定组(n=16)...

Radiomics features based on enhanced CT in predicting efficacy of immunotherapy in patients with refractory malignant melanoma lung metastasis

Objective To explore the value of radiomics features based on chest enhanced CT in predicting the efficacy of immunotherapy in pulmonary metastasis of refractory malignant melanoma. Methods Data of 49 patients with intrapulmonary metastasis of refractory malignant melanoma treated with programmed death receptor (PD-1) monoclonal antibody were retrospectively analyzed. The patients were then divided into progression group (n=17) and non-progression group (n=32,including 16 stable cases and 16 partial remission cases) according to respond evaluation criteria of solid tumors (RECIST) 1.1 criteria. The whole lesion of lung metastasis was manually delineated on lung window enhanced CT images before immunotherapy with 3D-Slicer software. Pyradiomics program was used to extract the shape, first-order gray level, texture and wavelet features of the lesions. Then the features were reduced by Pearson correlation coefficient and recursive feature elimination. A classification model for predicting progress was established using support vector machine (SVM) method, and receiver operating characteristic (ROC) curve was drawn to evaluate the diagnostic efficacy of model for distinguishing progression and non-progression group. Results A total of 841 CT imaging features were extracted from each target lesion, and finally 3 texture features were selected after dimension reduction and were taken to establish the prediction model, including wavelet-HHH_glszm_Low Gray Level Zone Emphasis, wavelet-HHL_first order_Skewness and wavelet-LLL_gldm_Small Dependence High Gray Level Emphasis. The area under the curve (AUC) of the prediction model was 0.913 (95%CI) for the training group and 0.860 (95%CI) for the test group. The sensitivity, specificity, accuracy, positive predictive value and negative predictive value for the training group was 83.3%,95.5%,91.2%,90.9% and 91.3%, of the test group was 80.0%, 80.0%,80.0%,66.7% and 88.9%, respectively. Conclusion The radiomics model based on enhanced chest CT before immunotherapy demonstrated good performance for predicting immunotherapy efficacy in patients with lung metastasis of malignant melanoma.