Development of a universal organ segmentation method based on similar image retrieval and machine-learning by using a large database of medical images

本研究的目的（一种方法解决医学图像上不同的器官分割问题）基本实现。我们的研究策略（基于机器学习和数据驱动的方法，而不是直接将人类经验传递给计算机程序）被证明是有效的解决医学图像上的器官分割问题。我们得出的结论是（在庞大的数据库和充分的准备下，器官分割等图像处理程序可以大大简化）。提出了在几十秒内完成器官自动分割的算法。我们对算法的性能进行了评估，确认了在1分钟内可以从躯干CT图像中自动分割出肾脏、脾脏等主要器官。

The aim of this research (one procedure solves different organ segmentation problems on medical images) was almost realized. Our research policy for procedure design by using (machine learning and data driven based approach instead of transferring the human experience to computer program directly) was proved to be efficient to solve organ segmentation problems on medical images. We reached a conclusion that (image processing procedures such as organ segmentation can be greatly simplified under the supporting of a large database and well preparations). The algorithm that proposed to (accomplish the automatic organ segmentation within several tens of seconds) was implemented. We evaluated the performance of the algorithm and confirmed that the major organ such as kidney, spleen, and so on can be segmented automatically within one minute from the torso CT images.