Image enhancement technology-based method for processing alzheimer's disease brain image

Abstract

Disclosed is an image enhancement technology-based method for processing an Alzheimer's disease brain image, the method comprising: acquiring brain image data of an Alzheimer's disease patient and obtaining a region-of-interest image; performing contrast enhancement on the region-of-interest image, determining whether noise and artifacts are present, and obtaining a contrast-enhanced lesion region image; constructing a gradient amplitude map, performing edge detection on the basis of the magnitude of the gradient amplitude, performing segmentation using a region growing method, and obtaining a precise lesion region via a preset growth criterion; analyzing a grayscale histogram of the precise lesion region to obtain a grayscale distribution feature, and using an enhancement strategy to perform processing to obtain a lesion region image subjected to fine structure enhancement processing; and fusing the contrast-enhanced lesion region image, the lesion region image subjected to fine structure enhancement processing, and a raw image to obtain an enhanced Alzheimer's disease brain image. The present invention can effectively improve the quality of Alzheimer's disease brain images.

Description

A brain imaging method for Alzheimer's disease based on image enhancement technology Technical Field

[0001] This invention relates to the field of information processing technology, and in particular to a method for processing brain images of Alzheimer's disease based on image enhancement technology. Background Technology

[0002] In the processing of brain images for Alzheimer's disease, the low contrast and blurred boundaries between lesion areas and normal areas make accurate identification and localization of lesions difficult. While contrast enhancement factors in image enhancement techniques can improve the contrast between lesion and normal areas, excessive contrast enhancement may introduce noise and artifacts, interfering with subsequent image analysis. Furthermore, because Alzheimer's lesions vary in shape and size, a single contrast enhancement method cannot adequately address the enhancement needs of different lesion areas. In addition, the internal structure of Alzheimer's lesions is complex, containing various tissue types, and simple contrast enhancement may not effectively highlight these fine structures. Therefore, how to improve the contrast between lesion and normal areas while simultaneously enhancing the fine structures within the lesion area and avoiding the introduction of excessive noise and artifacts is a significant technical challenge in the processing of Alzheimer's disease brain images.

[0003] Summary of the Invention

[0004] To address the technical problems existing in the prior art, this invention proposes a brain image processing method for Alzheimer's disease based on image enhancement technology, providing more accurate and reliable data support for subsequent image analysis.

[0005] To achieve the above objectives, the present invention provides a method for processing Alzheimer's disease brain images based on image enhancement technology, comprising:

[0006] Obtain brain imaging data from Alzheimer's patients, segment the brain images using image segmentation methods, and obtain images of regions of interest;

[0007] An adaptive histogram equalization method is used to enhance the contrast of the region of interest image, and the presence of noise and artifacts is determined to obtain a contrast-enhanced lesion region image.

[0008] A gradient magnitude map is constructed, edge detection is performed based on the magnitude of the gradient magnitude, and a region growing method is used for segmentation. By using preset growth criteria, the precise lesion area is obtained.

[0009] The gray-level histogram of the precise lesion region is analyzed to obtain the gray-level distribution characteristics, and an enhancement strategy is applied to process it to obtain a lesion region image after fine structure enhancement.

[0010] The contrast-enhanced lesion area image, the lesion area image after fine structure enhancement, and the original image are fused together to obtain the enhanced brain image of Alzheimer's disease.

[0011] Compared with the prior art, the present invention has the following advantages and technical effects:

[0012] This invention first segments brain images to obtain regions of interest. Then, addressing the low contrast and blurred boundaries of lesion regions, an adaptive histogram equalization algorithm is used to enhance contrast, combined with gradient analysis and region growing algorithms to achieve accurate segmentation. Considering the diverse morphologies of lesion regions, this invention selects appropriate enhancement strategies based on grayscale distribution characteristics and employs multi-scale analysis methods such as wavelet transform to enhance internal fine structures. Finally, the enhanced lesion regions are fused with the original images, and satisfactory enhancement results are obtained through iterative optimization. This invention can effectively improve the quality of brain images in Alzheimer's disease, providing more accurate and reliable data support for subsequent analysis, and is of great significance for disease diagnosis and research. Attached Figure Description

[0013] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0014] Figure 1 is a flowchart of an Alzheimer's disease brain image processing method based on image enhancement technology according to an embodiment of the present invention. Detailed Implementation

[0015] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0016] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0017] This invention proposes a brain image processing method for Alzheimer's disease based on image enhancement technology, as shown in Figure 1, including:

[0018] Obtain brain imaging data from Alzheimer's patients, segment the brain images using image segmentation methods, and obtain images of regions of interest;

[0019] Specifically, brain imaging data of Alzheimer's patients is acquired, and the brain imaging data is preprocessed to obtain preprocessed brain imaging data; a prior knowledge model of the region of interest is constructed; a deep learning-based image segmentation method, combined with the prior knowledge model, is used to segment the preprocessed brain imaging data to obtain segmented region of interest images; features of the segmented region of interest are extracted to construct feature vectors; a machine learning algorithm is used to classify the feature vectors to obtain classification results; based on the classification results, it is determined whether the patient has Alzheimer's disease.

[0020] An adaptive histogram equalization method is used to enhance the contrast of the region of interest image, and it is determined whether there is noise and artifacts introduced by over-enhancement, resulting in an improved image of the lesion region.

[0021] Specifically, the region of interest (ROI) image is acquired; for the ROI image, an adaptive histogram equalization algorithm is used to adaptively adjust the contrast of the local region to obtain a contrast-enhanced ROI image; for the contrast-enhanced ROI image, it is determined whether there is noise and artifacts introduced by over-enhancement. If so, noise suppression and artifact removal processing are performed to obtain a lesion region image with improved image quality; feature parameters of the lesion region image are extracted.

[0022] A gradient magnitude map is constructed, edge detection is performed based on the magnitude of the gradient magnitude, and a region growing method is used for segmentation. The boundary of the lesion region is obtained by using a preset growth criterion.

[0023] Specifically, based on the input original medical image, the Sobel operator is used to calculate the gradient magnitude of the image in the horizontal and vertical directions, obtaining the gradient magnitude map. According to a preset gradient magnitude threshold, the gradient magnitude map is binarized, setting pixels greater than the threshold to 1 and pixels less than the threshold to 0, resulting in a binarized edge image. Morphological closing operations are used to process the binarized edge image, filling small holes and connecting broken edges to obtain a complete edge contour. The coordinates of edge pixels are extracted from the edge contour, and the least squares method is used to fit the edge pixels, obtaining a coarse boundary curve of the lesion region. The curvature of each point on the coarse boundary curve is calculated, and the inflection point position on the boundary curve is determined based on the curvature magnitude, dividing the boundary curve into multiple line segments. For each line segment, a Hough transform is used to detect straight lines, and the endpoint coordinates of the line segment are determined based on the slope and intercept of the line, obtaining the polygonal boundary of the lesion region.

[0024] Based on the obtained polygonal boundary of the lesion region, pixels inside the boundary are extracted as initial seed points for the region growing algorithm. The growth criterion for the region growing algorithm is set as the gray-level difference between the seed point and its adjacent pixels being less than a threshold T. Based on the gray-level distribution characteristics of the lesion region, the threshold T is initially set to 20. Pixels satisfying the growth criterion are searched for in all directions around the seed point and marked as lesion regions. This process is repeated until no new pixels are added to the lesion region. During the region growing process, the threshold T is dynamically adjusted based on the morphological characteristics of the lesion region, such as its area, perimeter, and average gray-level value. When the area of ​​the lesion region is less than 100 square millimeters, the threshold T is reduced to 15; when the contrast between the average gray-level value of the lesion region and the surrounding normal tissue is less than 30%, the threshold T is increased to 25 to improve segmentation accuracy and obtain precise lesion region boundaries.

[0025] The gray-level histogram of the precise lesion area is analyzed to obtain the gray-level distribution characteristics, and an enhancement strategy is applied to process the image to obtain an enhanced image of the lesion area.

[0026] Specifically, the U-Net model, based on deep learning, automatically identifies and extracts lesion regions from images. Based on the extracted lesion region images, its gray-level histogram is calculated to analyze the distribution of gray values. In this embodiment, if the gray-level histogram shows that most pixel gray values ​​are concentrated in the low gray-level range (e.g., 0-50), it indicates that the gray-level distribution of the lesion region is concentrated in the low gray-level range. A non-linear stretching algorithm is used to enhance image contrast by adjusting parameters, resulting in a significant increase in brightness in originally darker areas, thereby enhancing the visibility of the lesion region. Conversely, if the gray-level histogram shows that gray values ​​are mainly concentrated in the high gray-level range (e.g., 200-255), a linear stretching algorithm is used to adjust the gray-level range to the entire usable gray-level space, thereby improving the overall contrast of the image and obtaining the processing result.

[0027] The image is decomposed into multiple scales using wavelet transform to obtain wavelet coefficients at different scales. Based on the statistical characteristics of the wavelet coefficients, texture features reflecting the fine structure within the lesion area are extracted, and texture feature vectors are constructed. Machine learning algorithms such as support vector machines are used to classify the texture feature vectors and determine the tissue type within the lesion area. Based on the classification results of the tissue types, morphological processing methods are used to enhance regions of different tissue types, highlighting the differences between different tissues. The enhanced lesion area image is then reconstructed using wavelet transform to obtain the lesion area image after fine structure enhancement.

[0028] The contrast-enhanced lesion area image, the lesion area image after fine structure enhancement, and the original image are fused together to obtain the enhanced brain image of Alzheimer's disease.

[0029] Specifically, including:

[0030] Step 1: Obtain contrast-enhanced images of the lesion area and precise lesion area image data after fine structure enhancement processing.

[0031] Step 2: Obtain raw brain imaging data for Alzheimer's disease.

[0032] Step 3: The contrast-enhanced lesion area image data, the precise lesion area image data after fine structure enhancement processing, and the original Alzheimer's disease brain image data are fused together using a weighted average method according to preset weight parameters to obtain the fused brain image.

[0033] Step 4: Evaluate the quality of the fused brain images using metrics such as peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) to determine whether the quality of the fused images meets the preset threshold requirements.

[0034] Step 5: If the quality assessment result of the fused image is lower than the preset threshold, return to step 3, adjust the weight parameters of the fusion algorithm, and perform image fusion again until the quality assessment result meets the requirements.

[0035] Step 6: Store the fused brain image data that meets the quality requirements as input data for subsequent Alzheimer's disease image analysis.

[0036] Step 7: Based on the fused brain imaging data, machine learning algorithms such as convolutional neural networks are used to analyze and predict the severity of Alzheimer's disease and the extent of brain atrophy, providing data support for disease diagnosis and treatment.

[0037] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An Alzheimer's disease brain image processing method based on image enhancement technology, characterized in that, include: Obtain brain imaging data from Alzheimer's patients, segment the brain images using image segmentation methods, and obtain images of regions of interest; An adaptive histogram equalization method is used to enhance the contrast of the region of interest image, and the presence of noise and artifacts is determined to obtain a contrast-enhanced lesion region image. A gradient magnitude map is constructed, edge detection is performed based on the magnitude of the gradient magnitude, and a region growing method is used for segmentation. By using preset growth criteria, the precise lesion area is obtained. The gray-level histogram of the precise lesion region is analyzed to obtain the gray-level distribution characteristics, and an enhancement strategy is applied to process it to obtain a lesion region image after fine structure enhancement. The contrast-enhanced lesion region image, the lesion region image after fine structure enhancement, and the original image are fused together to obtain an enhanced Alzheimer's disease brain image.

2. The Alzheimer's brain image processing method based on image enhancement technology according to claim 1, characterized in that, The brain image is segmented using an image segmentation method to obtain the region of interest image, including: The brain image data is preprocessed to obtain preprocessed brain image data; a prior knowledge model of the region of interest is constructed; and a deep learning-based image segmentation algorithm is used, combined with the prior knowledge model, to segment the preprocessed brain image data to obtain segmented region of interest images.

3. The image enhancement technique based Alzheimer's brain image processing method according to claim 1, wherein, Obtain contrast-enhanced images of the lesion area, including: For the region of interest image, an adaptive histogram equalization algorithm is used to adjust the local region. The contrast is adjusted to obtain a contrast-enhanced region of interest image. For the contrast-enhanced region of interest image, it is determined whether there is noise and artifacts introduced by over-enhancement. If so, noise suppression and artifact removal are performed to obtain a contrast-enhanced lesion area image.

4. The image enhancement technique based Alzheimer's brain image processing method according to claim 1, wherein, Constructing the gradient magnitude map and performing edge detection based on the gradient magnitude includes: Based on the input original medical image, the Sobel operator is used to calculate the gradient magnitude of the original medical image in the horizontal and vertical directions, and the gradient magnitude map of the image is obtained. Based on the preset gradient magnitude threshold, the gradient magnitude map is binarized, with pixels greater than the threshold set to 1 and pixels less than the threshold set to 0, resulting in a binarized edge image. Morphological closing operations are used to process the binarized edge image, fill small holes in the edge image, connect broken edges to obtain a complete edge contour, extract the coordinates of edge pixels based on the edge contour, and fit the edge pixels using the least squares method to obtain a rough boundary curve of the lesion area. Calculate the curvature of each point on the rough boundary curve, determine the inflection point position on the boundary curve based on the curvature, divide the boundary curve into several line segments, use Hough transform to detect straight lines on each line segment, and determine the endpoint coordinates of the line segment based on the slope and intercept of the line to obtain the polygonal boundary of the lesion area.

5. The Alzheimer's brain image processing method based on image enhancement technology according to claim 4, characterized in that, Obtaining the precise lesion area includes: Based on the vertex coordinates of the polygon boundary, the boundary of the lesion area is drawn on the original medical image, the lesion area is separated from the normal area, and the lesion area is extracted. A region growing algorithm is used to finely segment the lesion region. Based on the morphological characteristics of the lesion region, the growth criterion parameters of the region growing algorithm are dynamically adjusted to control the region growing process and obtain accurate segmentation. The affected area.

6. The image enhancement technique based Alzheimer's brain image processing method according to claim 1, wherein, Obtain images of the lesion area after fine-structure enhancement processing, including: For the precise lesion area, a grayscale histogram of the precise lesion area is obtained to obtain grayscale distribution characteristics; based on the grayscale distribution characteristics, the grayscale distribution type of the lesion area is determined; if the grayscale distribution type is a concentration of low grayscale ranges, a nonlinear stretching algorithm is used for contrast enhancement; if the grayscale distribution type is a concentration of high grayscale ranges, a linear stretching algorithm is used for contrast enhancement, and the contrast enhancement algorithm is applied to the precise lesion area to obtain the processing result.

7. The image enhancement technique-based Alzheimer's disease brain image processing method according to claim 6, characterized in that, Obtaining images of lesion areas after fine-structure enhancement also includes: Based on the processing results, wavelet transform is used for multi-scale decomposition to obtain wavelet coefficients at different scales. According to the statistical characteristics of the wavelet coefficients, texture features reflecting the fine structure inside the lesion area are extracted, texture feature vectors are constructed, and the texture feature vectors are classified to determine the tissue type inside the lesion area. Based on the classification results of tissue types, morphological processing methods are used to enhance regions of different tissue types. The enhanced lesion area image is then reconstructed using wavelet transform to obtain the lesion area image after fine structure enhancement.

Images