3D Medical Image Processing Algorithm Competition in Japan
Akinobu Shimizu, Takayuki Kitasaka, Shigeru Nawano, K. Shinozaki, Yuki Tateno · 2011
This paper reports on a medical image processing algorithm competition that has been held annually in Japan from 2002 to 2010. It is the world’s oldest competition in the field of 3D medical image processing, and covers various targets: the liver, pancreas, hepatocellular carcinomas, and hepatic vascular and metastatic liver tumors. This paper presents the algorithm entries and results of the competition. The paper also discusses the benefits of the competition, among which the main benefit is that such a competition can help to rank existing algorithms by using an unknown image database. In addition, there are several secondary benefits. For example, an image database that is distributed by organizers helps researchers who suffer from a shortage of images to conduct training and carry out validation. Another example is the progress achieved in performance evaluation methodologies used for comparing algorithms. The Japanese competition employs a different evaluation scheme than other international competitions; the necessity and advantages of using this different scheme is explained in this paper. Another important benefit of the competition is the aggregation of the algorithms registered in the competition. Therefore, in this paper, we also present the results of the algorithm aggregation and its superiority over a winner’s algorithm.