Measuring Size and Shape of the Hippocampus in MR Images Using a Deformable Shape Model (see full paper here)

A method for segmentation and quantification of the shape and size of the hippocampus is proposed, based on an automated image analysis algorithm. The algorithm uses a deformable shape model to locate the hippocampus in magnetic resonance images and to determine a geometric representation of its boundary. The deformable model combines three types of information. First, it employs information about the geometric properties of the hippocampal boundary, from a local and relatively finer scale to a more global and relatively coarser scale. Second, the model includes a statistical characterization of normal shape variation across individuals, serving as prior knowledge to the algorithm. Third, the algorithm utilizes a number of manually defined boundary points, which can help guide the model deformation to the appropriate boundaries, wherever these boundaries are weak or not clearly defined in MR images. Excellent agreement is demonstrated between the algorithm and manual segmentations by well-trained raters, with a correlation coefficient equal to 0.97 and algorithm/rater differences statistically equivalent to inter-rater differences for manual definitions.


• Generating the hippocampal surface model from the manually segmented hippocampi. (a) A 3D rendering of an expert’s manual segmentation of the left hippocampus, (b) its surface representation, (c) the degree of similarity between the attribute vector of the end point (indicated by the red arrow in (b)) and the attribute vectors of all other hippocampal points. The similarity degrees are normalized between 0 and 1, with red and blue colors respectively denoting the similarity degrees 1 and 0, as shown in the color bar. Colors for other similarity degrees are provided in the color bar. This figure demonstrates that the attribute vectors uniquely characterize certain parts of the hippocampal boundary, based on their geometric structure, and therefore help establish anatomically meaningful correspondences based on attribute vector similarity.




• An example of hippocampal segmentation using the AFDM algorithm. Final segmentation (green) is overlaid on (a) the manual segmentation result, and (b) the original images. The slice numbers of the images in (a) and (b) are the same.
  
  
  A 3D rendering of the segmentation results in the above. The model (red) with the carefully manually segmented hippocampi (green), (b) Overlay of the final deformed model with the manually segmented hippocampi.




• Quantitative evaluation of the performance of our algorithm. Volume and overlap errors are computed between our segmentation and the real hippocampus, and also between two well-trained raters.




• Hippocampus segmentation example 1