GPU Histogram Computation for 3D Image Registration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for registering 3D image volumes, such as those from CT and MRI, are inefficient in computing transformations that maximize mutual information, requiring extensive computation time and resources.
Innovation Solution
The use of a graphics processing unit (GPU) to accelerate histogram computation through a two-pass approach, enabling faster data scattering and probability determination for pixel value coincidences, allowing for efficient registration of 3D image volumes by maximizing mutual information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If conventional CPU-based methods are used for histogram computation in mutual information registration, then implementation simplicity is maintained, but computation time becomes excessively long
Solution Approach 1:
The patent replaces the conventional CPU-based sequential processing mechanism with a GPU-based parallel processing mechanism. Specifically, it uses vertex shaders to perform parallel histogram computation across multiple pixel locations simultaneously, and fragment shaders to accumulate histogram values in parallel. This substitution of the computational mechanism reduces computation time from minutes to seconds while accepting increased implementation complexity through shader programming.
2Productivity
If standard GPU programming approaches are used, then hardware utilization is limited, but achieving faster processing requires novel accelerated methods
Solution Approach 1:
The patent segments the histogram computation task into distinct shader stages: vertex shaders handle per-vertex probability calculations and coordinate transformations, while fragment shaders handle the accumulation of histogram values. This segmentation allows each shader stage to be optimized independently and充分利用 the GPU's parallel processing capabilities, achieving significant speedup over conventional methods.
Solution Approach 2:
The patent transforms the traditional 1D or 2D histogram computation into a 3D histogram computation by incorporating the transformation parameters as an additional dimension. This allows the GPU to compute the entire transformation space simultaneously through parallel processing, rather than sequentially evaluating each transformation parameter combination.
3Measurement precision
If accurate mutual information calculation is performed using complete histogram data, then registration accuracy is maximized, but memory requirements and computation time increase
Solution Approach 1:
The patent extracts only the essential histogram data needed for mutual information calculation from the complete 3D image volumes. By computing histograms over sampled pixel locations rather than all possible pixel combinations, and by using the GPU's parallel processing to compute probabilities directly from the histogram bins, it achieves accurate registration with reduced memory requirements compared to storing and processing complete image data.
Data Source
AI summary
The invention provides, in some aspects, methods and apparatus for registering 3D image volumes by utilizing a graphics processing unit (GPU) to facilitate maximizing the “mutual information” of the two image volumes being registered. Particularly, the GPU is used to compute a histogram representing a probability that a pixel value in one of the images being registered coincides with a pixel value in a transform of the other image. In one aspect, this histogram is generated by a two-step approach including (i) “drawing” a polygon covering a two-component render target having as many pixels as there are samples, where the value of each pixel in the target includes the intensities of the respective pixels of the first and second images, (ii) interpreting that render target as a buffer of vertex positions and, for each sample, rendering into a new render target, a single vertex using coordinates from that buffer.