病理图像分析的深度学习方法研究综述

A Review of Deep Learning Methods for Pathology Image Analysis

数字病理在临床诊断中的普及为人工智能在病理学中的应用创造了条件。由于强大的建模表征能力,深度学习技术在计算病理学中得到广泛应用,在改善疾病诊断方面展现了巨大潜力。本文回顾了深度学习和病理图像分析相结合的应用,并概述了三个关键任务的领域进展,包括分类、分割和检测。针对每一个任务,介绍了相关的临床价值、技术难点、以及主流的算法设计等。目前病理人工智能算法已经取得了一些令人瞩目的成果,但仍未达到临床应用的标准。本文分析了将人工智能技术从研究转化为临床应用时面临的挑战以及未来研究方向。

Recent advances of digital pathology in clinical diagnosis make it possible for the application of artificial intelligence to pathology.Due to the powerful learning ability in dealing with complex patterns,deep learning algorithms are widely being applied to computational pathology and have great potential to improve diagnosis.Here we review the intersection between deep learning and pathology image analysis,and survey the progress of three key tasks:classification,segmentation,and detection.For each task,we cover its clinical value,technical challenges,and state-of-the-art algorithms.Despite the promising results,very few artificial intelligence algorithms have reached clinical deployment.We describe the challenges that still need to be addressed before transforming artificial intelligence technologies from research to clinical practice and suggest directions for future work.