摘要 |
<P>PROBLEM TO BE SOLVED: To more improve an efficiency for detecting an anatomical object by using an MSL in the detection of the 3D anatomical object. <P>SOLUTION: As for an inputted medical image, a restricted search range is determined in accordance with a training volume. A position candidate is detected within the restricted search range by using a first classification member to be trained. In the training volume, a direction example is used to form a position-direction assumption from the position candidate. A position-direction candidate is detected from the position-direction assumption by using a second classification member to be trained. In the training volume, a similarity converting assumption is formed from the position-direction candidate in accordance with a scale example. Similarity converting candidates are detected from the similarity converting assumption by using a third classification member to be trained. In accordance with at least one of the similarity converting candidates, the 3D anatomical object is detected in a medical image volume. <P>COPYRIGHT: (C)2010,JPO&INPIT |