实用老年医学

实用老年医学 ›› 2022, Vol. 36 ›› Issue (9): 933-937.doi: 10.3969/j.issn.1003-9198.2022.09.016

• 临床研究 • 上一篇    下一篇

基于临床及影像组学资料建立模型预测前列腺癌生化复发的初步研究

杨浩, 施颉   

  1. 225300 江苏省泰州市,泰州市中医院超声科
  • 收稿日期:2022-03-07 出版日期:2022-09-20 发布日期:2022-09-21
  • 通讯作者: 施颉,Email: 405067467@qq.com
  • 基金资助:
    泰州市科技支撑计划(社发)项目(TS201908)

Study on a clinical and radiomics data-based model for predicting biochemical recurrence of prostate cancer

YANG Hao, SHI Jie   

  1. Department of Ultrasound, Taizhou Hospital of Traditional Chinese Medicine, Taizhou 225300, China
  • Received:2022-03-07 Online:2022-09-20 Published:2022-09-21

PDF (PC)

960

摘要： 目的 基于临床资料-超声造影-磁共振影像组学数据,为前列腺癌 (PCa) 病人开发预测PCa治疗后生化复发的模型,并验证其临床有效性。 方法 回顾性分析169例2018年1月至2020年2月在泰州市中医院通过MRI及超声靶向性活检确诊并进行PCa根治术病人的临床资料,随访截至2022年3月,若前列腺特异性抗原(PSA)监测连续2次≥0.2 ng/mL提示PCa生化复发,将其分为生化复发组(n=65)和正常组(n=104)。以7/3的比例建立训练集和测试集,基于上述病人的临床影像资料建立一般资料模型、超声造影模型、磁共振组学模型和组合模型。在训练集中,利用R语言4.1.3版本制作ROC曲线和决策曲线分析评估预测模型的能效,然后在测试集中进行外部验证。 结果 多因素回归证实肿瘤直径、术前PSA浓度、治疗方式、弹性成像分级、灰度区域大小矩阵特征-小区域高灰度强调范围、灰度行程矩阵特征-行程方差是预测生化复发的独立风险因素(组合模型)。在训练集中,基于临床影像资料建立的组合模型预测PCa生化复发的能效最高,明显高于一般资料模型(淋巴结转移、治疗前PSA、治疗方式)、超声造影模型(弹性成像分级)、磁共振组学模型(灰度区域大小矩阵特征-小区域高灰度强调范围、灰度行程矩阵特征-行程方差、邻域灰度差矩阵特征-信噪比对比度)。R语言决策曲线证实组合模型比其他模型净收益大,测试集验证结果相同。 结论 对于通过MRI及超声造影靶向活检诊断的PCa病人,单纯临床、影像学资料预测生化复发均不理想,组合模型预测工具为临床筛选生化复发人群提供了契机并有助于改善PCa预后。

关键词: 前列腺癌, 生化复发, 超声造影, 磁共振, 影像组学, 预测模型

Abstract: Objective To develop a model based on clinical data and contrast-enhanced ultrasound (CEUS) and MRI results for predicting biochemical recurrence of prostate cancer (PCa) in the elderly patients confirmed by targeted biopsy. Methods The clinical data of 169 elderly patients who underwent radical prostatectomy from January 2018 to February 2020 in Taizhou Hospital of Traditional Chinese Medicine were analyzed retrospectively. The cases were divided into biochemical recurrence group (n=65) and normal group (n=104) according to the results of 2-4 years of follow-up. PCa biochemical recurrence was defined as two consecutive detection of PSA ≥0.2 ng/mL . General data model, CEUS model, MRI radiomics model and combined model to predict the biochemical recurrence of PCa were established based on the clinical-imaging data of these patients. The predictive effects were compared among the four models. Results Logistic regression analysis showed that the tumor diameter, preoperative PSA concentration, treatment mode, elastography grade, small area high gray level emphasis and run variance were the independent risk factors of biochemical recurrence. In the training set, the combined model based on clinical-imaging parameters showed the highest predictive efficiency of PCa biochemical recurrence than general data model, contrast-enhanced ultrasound model and MRI radiomics model. Conclusions The predictive value of clinical data or imaging data alone for PCa biochemical recurrence is not ideal. The combined model provides an opportunity for clinical screening of biochemical recurrence population and helps to improve the prognosis of PCa.

Key words: prostate cancer, biochemical recurrence, contrast-enhanced ultrasound, magnetic resonance imaging, radiomics, predictive model

中图分类号: