Hierarchical atlas-based segmentation

A method for segmenting an image, comprising:

• receiving a reference image;
• retrieving a stored annotated template image;
• determining a transformation function that maximizes a degree of similarity between the template image and the reference image when applied to the template image, said transformation function being limited to a first number of degrees of freedom;
• first refining the determined transformation function to maximize the degree of similarity between the template image and the reference image when applied to the template image, while the transformation function is limited to a second number of degrees of freedom that is larger than the first number of degrees of freedom;
• and
• using the refined transformation function to map the annotations of the template image to the received reference image.

A computer system comprising:

• a processor;
• and
• a non-transitory, tangible, program storage medium, readable by the computer system, embodying a program of instructions executable by the processor to perform method steps for segmenting an image, the method comprising:
• registering an annotated template image to an acquired reference image using only rigid transformations to define a transformation function relating the annotated template image to the acquired reference image;
• refining the defined transformation function by registering the annotated template image to the acquired reference image using only affine transformations;
• further refining the refined transformation function by registering the annotated template image to the acquired reference image using only multi-affine transformations;
• still further refining the twice refined transformation function by registering the annotated template image to the acquired reference image using deformation transformations;
• and
• using the three-times-refined transformation function to segment the acquired reference image by mapping the annotations of the template image to the acquired reference image and applying the mapped annotations of the template image to the structure of the acquired transformation function.