| 主权项 |
1. A method of analyzing a medical image, the medical image comprising one or more than one region of interest, the method comprising:
a) providing the medical image comprising a set of actual image values; b) rescaling the actual image values to produce corresponding rescaled image values and to produce a rescaled image from the rescaled image values; c) deriving a histogram of the rescaled image values; d) using the histogram to derive an adaptive segmentation threshold that can be used to split the rescaled image into two sub-images, a first sub-image with intensities at or below the adaptive segmentation threshold and a second sub-image with intensities above the adaptive segmentation threshold, or a first sub-image with intensities below the adaptive segmentation threshold and a second sub-image with intensities at or above the adaptive segmentation threshold, or a first sub-image with intensities below the adaptive segmentation threshold and a second sub-image with intensities above the adaptive segmentation threshold; e) using the adaptive segmentation threshold to recursively split the rescaled image to generate a Hierarchical Region Splitting Tree of sub(sub) images based on consistency of the rescaled image values of the rescaled image; f) terminating the recursive splitting of the sub(sub) images using one or more than one predetermined criteria thereby completing the Hierarchical Region Splitting Tree; and g) identifying one sub(sub) image in the terminated Hierarchical Region Splitting Tree which comprises the region of interest; the method further comprising performing a secondary rescaling of the rescaled image values of every rescaled sub(sub) image in the Hierarchical Region Splitting Tree back to the actual image values present in the medical image to create a secondary rescaled medical image, thereby producing a secondarily rescaled sub(sub) image comprising the region of interest; where the rescaled image values fit in [0,255] unsigned 8-bit integer range; where the predetermined criteria is selected from the group consisting of area threshold=50 pixels and (standard deviation threshold=10 rscVals (StdDevTh=10 rscVals) and kurtosis threshold=1.5); where the region of interest is a representation of an abnormality in the living human tissue, and where the method further comprises quantifying the abnormality in the living human tissue; where the method further comprises performing a secondary resealing of the resealed image values in every resealed sub(sub) image in the Hierarchical Region Splitting Tree back to the actual image values present in the medical image to create a secondary resealed medical image, and determining an image value or a set of image values of actual image values in the medical image after the secondary resealing, where the image value or a set of image values of actual image values determined identifies the abnormality represented in the medical image for the modality being used to generate the medical image provided; and where the method further comprises preparing a mask of the sub(sub) image containing the representation of the abnormality, and cleaning the mask to remove small outlier regions to generate a cleaned mask of the sub(sub) image containing the representation of the abnormality. |
