基于灰阶超声的影像组学模型预测乳腺癌新辅助化疗效果

目的 探讨基于灰阶超声的影像组学模型预测乳腺癌新辅助化疗（NACT）效果的应用价值。方法 选取53例乳腺癌患者，根据NACT疗效分为临床应答与临床无应答组，比较二组临床资料及灰阶超声特征。提取基于灰阶超声的乳腺癌影像组学特征，采用Logistic回归分析建立基于上述特征的模型，采用ROC曲线评价模型预测乳腺癌NACT后临床应答的效能。结果 NACT后临床应答组32例、临床无应答组21例，2组间年龄、绝经比例、分期及分子分型差异均无统计学意义（P均>0.05），声像图所示病灶最大径、内部回声、钙化、边缘、后方回声、形态差异均无统计学意义（P均>0.05）。共6个影像学特征纳入Logistic回归模型，该模型预测乳腺癌NACT后临床应答的AUC为0.8895% CI（0.78，0.99）]，敏感度0.88，特异度0.81。结论 基于灰阶超声的影像组学模型对评价乳腺癌NACT效果有一定价值。

Ultrasound-based radiomics model in predicting efficacy of neoadjuvant chemotherapy in breast cancer

Objective To investigate the clinical value of ultrasound-based radiomics model in predicting clinical efficacy of neoadjuvant chemotherapy (NACT) in breast cancer. Methods Totally 53 breast cancer patients were divided into clinical response group and clinical non-response group according to clinical outcomes of NACT. Clinical data and grayscale ultrasound characteristics were compared between the two groups. Moreover, radiomic features of greyscale ultrasonic images of breast cancers were extracted and selected to build an influential prediction model using Logistic regression. ROC curve was used to evaluate the efficacy of the model for diagnostic clinical response after NACT. Results Totally 32 patients were enrolled in clinical response group, while 21 in clinical non response group. No significant difference was found in age, menopause ratio, cancer staging and molecular typing (all P>0.05), nor in longest diameter of lesion, echo pattern, calcification, boundary, shadow, shape between the two groups (all P>0.05). Logistic regression model was successfully performed with 6 adapted features, which predicted the clinical response after NACT with an AUC of 0.88 (95% CI0.78, 0.99]), the sensitivity and specificity was 0.88 and 0.81, respectively. Conclusion Ultrasound-based radiomics model has certain value in predicting clinical efficacy of NACT in breast cancer.