基于Spark的并行医学图像处理研究

Parallel processing researches of Medical image based on Spark

摘要:

随着医学图像规模的不断增长，为了快速且有效的处理医学图像并使各类图像处理算法得到应用。文章将传统的医学图像处理方法与Spark整合起来，提出了基于Spark的并行医学图像处理方法.首先，采用基于二进制的图像预处理转换方法，存储图像到分布式文件系统HDFS中；其次，应用传递函数的方法，避免了图像处理算法进行MapReduce转化，实现了快速的通用图像并行处理；最后，以肺叶DR图像分割算法为实例证明了基于Spark医学图像并行处理有较好的适应性和较高的效率，并适应大规模图像的并行处理。

Abstract:

With the growing number of medical images, in order to effectively and quickly process medical images and make all kinds of medical images processing method get quickly transplant and apply. A medical image parallel processing method based on Spark is proposed, which integrate traditional medical image processing method with the Spark. Firstly, preprocess images base on binary conversion method, store images in the HDFS distributed file system. Secondly, implements a transfer function way, which realized image parallel processing quickly, avoid convert the image processing algorithms to the MapReduce algorithms. Thirdly, according to the experiment, medical image parallel processing method based on Spark has preferable adaptability and efficiency, which can adapt to the large-scale image processing.