深度学习和影像组学在膀胱癌精准诊疗中的研究进展

Advances in deep learning and Radiomics for precision diagnosis and treatment of bladder cancer

近年来,膀胱癌(bladder cancer,BCa)的发病率逐年上升,已经成为威胁中老年人健康的重要因素之一,BCa的早期发现和预后监测日益成为研究的热点。影像组学是一种高通量的定量特征提取方法,可以挖掘多模态医学图像中包含的信息,然后对这些海量图像进行综合分析以提取表型特征,并探索患者预后与这些提取的特征之间的关系。深度学习是一种表示学习方法,其中复杂的多层神经网络架构通过将输入信息转换为多层次抽象自动学习数据表示。本文从泌尿外科临床医生的角度,综述了影像组学和深度学习在BCa病理分级和分期预测、淋巴结转移预测和疗效评估等精准诊疗的研究进展,并对未来研究方向进行了展望。

In recent years,the incidence of bladder cancer(BCa)has been increasing year by year and has become one of the important factors threatening the health of middle-aged and elderly people,and the early detection and prognosis monitoring of BCa has increasingly become a hot spot of current research.Radiomics is a high-throughput quantitative feature extraction method that mines the information contained in multimodal medical images,then synthesises these massive images to extract phenotypic features and explores the relationship between patient prognosis and these extracted features.Deep learning is a representation learning approach in which complex multilayer neural network architectures automatically learn data representations by transforming input information into multi-level abstractions.This paper reviews the research progress of radiomics and deep learning in precision diagnosis and treatment of BCa from the perspective of urological clinicians,including pathological grading and staging prediction,tumour lymph node metastasis prediction and efficacy assessment,and provides an outlook on future research directions.