基于磁共振动态增强的影像组学及深度学习在肺癌脊柱转移鉴别诊断中的应用

目的探讨基于动态增强磁共振成像(DCE-MRI)的影像组学及深度学习在肺癌脊柱转移鉴别诊断中的应用价值。方法回顾性分析61例确诊为脊柱转移患者的DCE-MRI,绘制感兴趣区域的时间-信号强度曲线,根据曲线定义3个参数,用区域增长算法对病灶进行标准化分割,通过影像组学提取分析3个DCE-MRI参数图的特征,用随机森林算法挑选出与鉴别疾病最相关的特征用于构建分类器进而进行诊断;研究包含2种深度学习算法,3个DCE-MRI参数图作为卷积神经元网络(CNN)的输入,将DCE-MRI每个层面的图像集视为一个时间序列,12层DCE图像作为卷积长短时间记忆(CLSTM)神经元网络的输入。结果影像组学诊断的准确率为0.71,CNN和CLSTM的平均诊断准确率分别为0.71、0.81。结论基于DCE-MRI的影像组学及深度学习在鉴别诊断肺癌脊柱转移方面具有可行性,可为临床诊断提供有价值的信息。

The Application of Radiomics and Deep Learning Based on Dynamic Contrast-Enhanced MRI in Differential Diagnosis of Lung Cancer Spinal Metastasis

Objective To explore the value of radiomics and deep learning based on dynamic contrast enhanced magnetic resonance imaging(DCE-MRI)in the differential diagnosis of spinal metastases in lung cancer.Methods The DCE-MRI of 61 patients diagnosed with spinal metastases were retrospectively analyzed,and the signal intensity-time curve of the region of interest was drawn.Three parameters were defined according to the curve.The lesion was standardized by regional growth algorithm.Features of the three DCE-MRI parameter maps were extracted.The random forest algorithm is used to select the features that are most relevant to identify the disease for the construction of a classifier for diagnosis.The study includes 2 deep learning algorithms,and 3 DCE-MRI parameter maps are used as convolutions.The input of the neural network(CNN)regards the image set of each layer of DCE-MRI as a time series,and the 12-layer DCE image is used as the input of the convolutional long and short time memory(CLSTM)neural network.Results The diagnostic accuracy of radiomics was 0.71,and the average diagnostic accuracy of CNN and CLSTM was 0.71 and 0.81,respectively.Conclusion DCE-MRI-based radiomics and deep learning are feasible in differential diagnosis of spinal metastases in lung cancer,and can provide valuable information for clinical diagnosis.