基于深度网络术前评估膀胱癌患者淋巴结状态的模型构建

Construction of a deep learning model for preoperative evaluation of lymph node status in patients with bladder cancer

目的针对术前膀胱癌患者淋巴结状态影像学检查敏感度低、病理检查有创等问题,构建基于影像与深度学习的膀胱癌淋巴结转移预测模型,实现对术前淋巴结状态的准确预测。方法首先,从癌症影像档案数据库下载膀胱癌患者的影像数据和临床数据,经数据清洗最终纳入80例膀胱癌患者数据,其中,淋巴结阳性27例,阴性53例;其次,基于Pytorch深度学习框架构建用于术前评估膀胱癌淋巴结转移的ResNet18卷积神经网络;最后,将数据集按照7∶3的比例划分为训练集和测试集,通过调整网络结构和超参数,提高网络预测效能。结果本研究构建的深度学习模型在膀胱癌淋巴结转移预测方面取得良好效果,在测试集中受试者工作特征曲线下面积达到94%,敏感度达到98%,明显高于既往研究结果。结论本研究基于术前CT影像与ResNet18构建的深度学习模型,能够实现对膀胱癌淋巴结转移的准确预测,有望为临床医师制定最佳治疗决策提供重要依据。

Objective Aiming at the problems of low sensitivity of imaging examination of lymph node status and invasive pathological examination in patients with bladder cancer before operation,a prediction model of lymph node metastasis of bladder cancer based on imaging and deep learning is constructed to realize the accurate prediction of preoperative lymph node status.Methods Firstly,imaging data and clinical data of bladder cancer patients were downloaded from the cancer imaging archive database,and 80 cases were finally included after data cleaning,including 27 cases with positive lymph nodes and 53 cases with negative lymph nodes.Secondly,a ResNet18 convolutional neural network was constructed based on Pytorch deep learning framework for preoperative evaluation of bladder cancer lymph node metastasis.Finally,the data sets were divided into training set and test set according to the ratio of 7∶3,and the network structure and hyperparameters were adjusted to improve the network prediction efficiency.Results The deep learning model constructed in this study has achieved a good prediction effect on lymph node metastasis of bladder cancer.In the test set,the area under the receiver operating characteristic curve reached 94%and the sensitivity reached 98%,which were significantly higher than the results of previous studies.Conclusion The deep learning model based on preoperative CT images and ResNet18 can realize the accurate prediction of lymph node metastasis of bladder cancer,which is expected to provide an important basis for clinicians to make the best treatment decisions.