US 5,568,384 A · Biomedical imaging and analysis

Abstract

A method for synthesizing three dimensional multimodality image sets into a single composite image with accurate registration and congruence. Surfaces are initially extracted from two or more different images to be matched using semi-automatic segmentation techniques. These surfaces are represented as contours with common features to be matched. A distance transformation is performed for one surface image, and a cost function for the matching process is developed using the distance image. The geometric transformation includes three-dimensional translation, rotation and scaling to accommodate images of different position, orientation and size. The matching process involves efficiently searching this multi-parameter space and adjusting a surface or surfaces to find the best fit among them which minimizes the cost function. The local minima problem is addressed by using a large number of starting points. A pyramid multi-resolution approach is employed to speed up both the distance transformation computation and the multi-parameter minimization processes. Robustness in noise handling is accomplished using multiple thresholds embedded in the multi-resolution search. The method can register both partially overlapped and fragmented surfaces.

A method for synthesizing a base image data set and a match image data set into a single fused composite image data set with accurate registration and congruence using a programmable data processing system, comprising data input means, display means, and data storage means, the method comprising the steps of:

(a) using the data input means to provide the programmable data processing system with the base image data set and the match image data set, the base image data set and the match image data set each comprising a biomedical image generated by a biomedical image generation means;
(b) storing the base image data set and the match image data set in the data storage means;
and
(c) using the programmable data processing system to perform the steps of:
(1) if the voxels of the base image data set are non-cubic, applying shape-based surface interpolation to interpolate intermediate contours between every consecutive pair of actual contours of the base image data set to obtain a smoothly varying surface;
(2) uniformly sampling a plurality of registration points from the surface of the match image data set;
(3) creating a distance image data set by performing a distance transformation to convert the surface of the base image data set into a grey-level image in which all voxels have a value corresponding to the distance to the nearest surface voxel;
(4) transforming geometrically the registration points on the surface of the match image data set to become more closely aligned in orientation with corresponding points on the surface of the distance image data set;
(5) calculating the value of a cost function between registration points on the surface of the match image data set and the corresponding points on the surface of the distance image data set;
(6) if the value of the cost function is below a predetermined multi-level threshold value, reducing the threshold by at least one level and returning to step (4);
(7) storing the value of the cost function with the corresponding values for the geometric transformation parameters for later comparison;
(8) repeating steps (4) through (7) for each of a plurality of predetermined uniformly distributed starting points within the match image data set in order to generate a plurality of cost function values and corresponding geometric transformation parameters;
(9) if the resolution of the base image data set and the match image data set is below a predetermined multi-level maximum resolution level, scaling the base image data set and the match image data set upward at least one level and returning to step (4);
(10) selecting the minimum of the plurality of stored cost function values and using the corresponding stored geometric transformation parameters to map the match image data set onto the base image data set to create a single fused composite multi-valued image with two or more values associated with each 3-D spatial (x,y,z) coordinate uniquely representing and corresponding to the 3-D distribution of values of the base image data set and the match image data set;
and
(11) displaying, on the display means, the composite image values associated with each 3-D spatial (x,y,z) coordinate.

13 more elements

Biomedical imaging and analysis

Life of this patent

Biomedical imaging and analysis

Life of this patent

Abstract

Claims3