Magnetic resonance image channel selection method based on adaptive threshold and PCA (Principal Component Analysis)

Abstract

The invention discloses a magnetic resonance image channel selection method based on an adaptive threshold and PCA, and relates to the technical field of nuclear magnetic resonance medical imaging. Comprising the steps of image preprocessing, adaptive threshold separation, contour fitting, channel pre-screening, contour fitting, channel pre-screening, image data matrix construction, principal component analysis, channel sorting and screening and the like. By dynamically calculating the threshold and segmenting the local characteristics of the image, the accuracy and adaptability of segmentation are improved, the foreground and the background can be more accurately separated under the condition that the contrast ratio is large, and meanwhile, PCA analysis is adopted to calculate the principal component of channel data, fully consider the correlation between channels and reserve the main information of the data. The PCA can effectively reduce dimensions, select a channel with the maximum contribution degree, and ensure that important information is not lost. The combination of the two has better robustness and systematicness when processing complex scenes and multichannel data, redundancy and noise are effectively reduced, and the signal-to-noise ratio of magnetic resonance images is improved.

Description

Technical Field

[0001] The invention relates to the technical field of nuclear magnetic resonance medical imaging, and in particular to a magnetic resonance image channel selection method based on adaptive threshold and PCA. Background Art

[0002] Magnetic resonance imaging (MRI) is a non-invasive imaging technology that is widely used in medical diagnosis and scientific research. The multi-channel characteristics of MRI images provide rich information, but also bring complexity to data processing. Effectively selecting the most useful channels for specific applications can significantly improve data processing efficiency and analysis accuracy, and improve the signal-to-noise ratio of magnetic resonance images.

[0003] At present, the methods for selecting magnetic resonance image channels mainly include methods based on statistical features, frequency domain analysis and machine learning. However, statistical feature methods are difficult to capture the complex correlation information between channels; although frequency domain analysis methods can reduce noise to a certain extent, there is still the problem of information loss; machine learning methods, especially deep learning methods, can automatically learn effective channel selection strategies from data, but their training process requires a large amount of labeled data, high computing resource requirements, and is not easy to explain. To this end, we provide a magnetic resonance image channel selection method based on adaptive threshold and PCA. Summary of the invention

[0004] The purpose of the present invention is to provide a magnetic resonance image channel selection method based on adaptive threshold and PCA. The method combines the advantages of adaptive threshold and principal component analysis, realizes the preliminary screening of useful information through the adaptive threshold method, and further optimizes the channel selection through analysis, which can effectively reduce redundancy and noise and improve the signal-to-noise ratio of magnetic resonance images.

[0005] The technical problems solved by the present invention are:

[0006] (1) The existing threshold method does not work well in images with large contrast changes. It cannot adapt to the local characteristics of the image and easily leads to segmentation errors. In particular, it is difficult to accurately segment the foreground and background in complex images when there is a lot of noise or the contrast between the foreground and background is low.

[0007] (2) Existing feature selection methods ignore channel correlation. Statistical features or frequency domain analysis methods usually process each channel independently without fully considering the correlation between channels, which may lead to the omission of important information. Feature selection methods cause information loss. These methods may not retain the main information of the data when selecting features, especially in multi-channel data, where information loss is more serious.

[0008] (3) Existing technologies have poor robustness when facing complex scenes or data and are easily affected by noise and redundant information, which affects the final processing effect. Existing methods lack systematicity and integrity, and it is difficult to comprehensively consider and select the most useful channels in multi-channel data, resulting in limited processing effects.

[0009] The present invention can be implemented by the following technical solution: A magnetic resonance image channel selection method based on adaptive threshold and PCA, comprising the following steps:

[0010] Step 1: performing image preprocessing on the original magnetic resonance image data;

[0011] Step 2: Use adaptive thresholding to separate the preprocessed image to obtain a binary image of the foreground and background;

[0012] Step 3: extract the foreground contour and calculate the feature parameters of the corresponding contour;

[0013] Step 4: Compare the characteristic parameters of the corresponding contour with the set conditions, preliminarily screen out the channels containing useful information, and generate a channel list;

[0014] Step 5: Construct the image data matrix and fill it in according to the channel list and image correspondence;

[0015] Step 6: Calculate the covariance matrix of the image data matrix, and perform PCA analysis on the eigenvalues ​​and eigenvectors of the covariance matrix to obtain the principal components of each channel and their contribution;

[0016] Step 7: Sort the channels according to their contribution, select channels whose contribution exceeds a certain threshold as valid channels, and obtain the final valid channel list.

[0017] A further technical improvement of the present invention is that the image preprocessing includes noise reduction processing, contrast enhancement and normalization processing.

[0018] A further technical improvement of the present invention is that the characteristic parameters in step three include the length and area of ​​the contour, and when the length or area of ​​the contour meets the set condition threshold, the corresponding channel is added to the channel list.

[0019] A further technical improvement of the present invention is that in the image data matrix, a matrix is ​​constructed with one pixel as one row and one channel as one column.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] 1. The present invention dynamically calculates the threshold value and performs segmentation according to the local characteristics of the image, thereby improving the accuracy and adaptability of the segmentation, and can more accurately separate the foreground and background, especially in the case of large contrast.

[0022] 2. The present invention adopts PCA principal component analysis to calculate the principal components of channel data, fully consider the correlation between channels, and retain the main information of the data. PCA can effectively reduce the dimension, select the channel with the largest contribution, and ensure that important information is not lost.

[0023] 3. The present invention improves the processing effect by combining the adaptive threshold method and PCA. This technical solution has better robustness and systematicity when processing complex scenes and multi-channel data. The adaptive threshold provides high-precision preliminary screening, and PCA further optimizes channel selection, which comprehensively improves the processing effect.

[0024] 4. The present invention adopts a two-stage screening process. First, an adaptive threshold is used to screen out channels containing significant information. Then, PCA is used to further screen out channels with large contributions, thereby reducing the interference of redundant information and improving the efficiency and accuracy of data processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0026] Figure 1 It is a schematic diagram of the method flow of the present invention;

[0027] Figure 2 It is a schematic diagram of the threshold segmentation results of each channel of the cylindrical water model of the present invention;

[0028] Figure 3 Schematic diagram for comparing the merged image of the entire channel of the cylindrical water model (left) and the merged image of the screening channel of the present invention (right);

[0029] Figure 4 Schematic diagram for comparing the in vivo brain full channel merged image (left) and the screening channel merged image of the present invention (right);

[0030] Figure 5 Schematic diagram comparing the in vivo lumbar vertebrae full-channel merged image (left) and the screening channel merged image of the present invention (right). DETAILED DESCRIPTION

[0031] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0032] See also Figure 1 As shown, a magnetic resonance image channel selection method based on adaptive threshold and PCA specifically includes the following steps:

[0033] Step 1: Image preprocessing

[0034] The original magnetic resonance image data, i.e., the multi-channel MRI image data, is subjected to image denoising to remove high-frequency noise; at the same time, the image contrast is increased to highlight important features, and normalization is performed to obtain the pre-processed image data, i.e., I pre (x,y,c)=Preprocess(I(x,y,c));

[0035] Among them, I(x,y,c) represents the original image of the cth channel, I pre (x, y, c) represents the preprocessed image.

[0036] Step 2: Adaptive Threshold Segmentation

[0037] The adaptive threshold method is used to separate the foreground and background of the preprocessed image to obtain a binary image, namely T adaptive (x,y,c)=AdaptiveThreshold(I pre (x,y,c)), where T adaptive (x, y, c) represents the adaptive threshold image of the cth channel;

[0038] Adaptive thresholding is a technique used in image processing that separates the foreground and background in an image by dynamically calculating a threshold. Unlike fixed thresholding methods, adaptive thresholding determines the threshold for each pixel or small area based on the local characteristics of the image. This method is particularly suitable for images with large contrast changes and can extract useful information more accurately.

[0039] Step 3: Fitting the Contour

[0040] Extract the foreground contour from the binary image and calculate the contour's length, area and other characteristic parameters to obtain the contour information and its parameter characteristics, that is, {Ci(c)} = Contours(T adaptive (x,y,c)); where {Ci(c)} represents the set of all extracted contours in the cth channel.

[0041] Step 4: Channel Pre-screening

[0042] Compare the characteristic parameters of the profile with the predetermined screening criteria, thereby screening out channels that may contain useful information and generating a channel list;

[0043] Specifically, channels are filtered based on the length or area of ​​the contour, and channels that meet specific conditions are selected:

[0044] S selected ={c|Length(Ci(c))>Lmin or Area(Ci(c))>Amin}

[0045] Among them, Lmin and Amin represent the thresholds of length and area respectively, S selected Represents a list of channels.

[0046] Step 5: Matrix image data

[0047] Construct a data matrix, with each row representing a pixel and each column representing a channel. Fill the matrix according to the channel list and the corresponding image data to obtain the image data matrix, that is:

[0048]

[0049] Where X is the image data matrix, c1,c2,…,cn∈S selected .

[0050] Step 6: Principal Component Analysis

[0051] Calculate the covariance matrix of the image data matrix, solve the eigenvalues ​​and eigenvectors of the covariance matrix, perform PCA analysis, and obtain the principal components of each channel and their contribution, that is:

[0052] X=U∑V T

[0053] Among them, U is the left singular vector matrix, Σ is the singular value matrix, V is the right singular vector matrix, and the square of the singular value represents the variance contribution of the principal component.

[0054] Step 7: Channel sorting and filtering

[0055] According to the contribution of the principal component, the channels are sorted from large to small, and the channels whose contribution reaches a certain threshold are selected as valid channels. A valid channel list is generated and transmitted to the user or downstream image analysis and application modules through the interface.

[0056] The valid channel list is S final ={ci|Contribution(ci)>τ}, where τ is the contribution threshold.

[0057] The following example illustrates the effect of this method:

[0058] (1) A cylindrical water phantom was selected, and a total of 16 channel images were preprocessed using Gaussian filtering; adaptive threshold segmentation was performed using the Otsu threshold method (an adaptive threshold determination method) to obtain a binary image, and then contour fitting and channel preselection were performed; the fitting of each channel was obtained, and the channels with high noise levels of channels 2, 13, 14, and 15 were removed, such as Figure 2 As shown. After further screening by principal component analysis, the channel composite image is obtained as shown Figure 3,Compared to full channel merging, the signal-to-noise ratio of the merged image after channel screening by this method is improved.

[0059] (2) For the in vivo brain, a total of 16 channel images were preprocessed using Gaussian filtering, and adaptive threshold segmentation was performed using the Otsu threshold method to obtain a binary image. Contour fitting and channel preselection were then performed, and the channel composite image was further screened by principal component analysis. Figure 4 ,Compared to full channel merging, the signal-to-noise ratio of the merged image after channel screening by this method is improved.

[0060] (3) For the in vivo lumbar spine, a total of 16 channel images were preprocessed using Gaussian filtering, and adaptive threshold segmentation was performed using the Otsu threshold method to obtain a binary image. Contour fitting and channel preselection were then performed, and the channel composite image was further screened by principal component analysis. Figure 5 ,Compared to full channel merging, the signal-to-noise ratio of the merged image after channel screening by this method is improved.

[0061] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A magnetic resonance image channel selection method based on adaptive threshold and PCA, characterized in that: The steps include: Step 1: performing image preprocessing on the original magnetic resonance image data; Step 2: Use adaptive thresholding to separate the preprocessed image to obtain a binary image of the foreground and background; Step 3: extract the foreground contour and calculate the feature parameters of the corresponding contour; Step 4: Compare the characteristic parameters of the corresponding contour with the set conditions, preliminarily screen out the channels containing useful information, and generate a channel list; Step 5: Construct the image data matrix and fill it in according to the channel list and image correspondence; Step 6: Calculate the covariance matrix of the image data matrix, and perform PCA analysis on the eigenvalues ​​and eigenvectors of the covariance matrix to obtain the principal components of each channel and their contribution; Step 7: Sort the channels according to their contribution, select channels whose contribution exceeds a certain threshold as valid channels, and obtain the final valid channel list.

2. The magnetic resonance image channel selection method based on adaptive threshold and PCA according to claim 1, characterized in that: The image preprocessing includes noise reduction, contrast enhancement and normalization.

3. The magnetic resonance image channel selection method based on adaptive threshold and PCA according to claim 1, characterized in that: The characteristic parameters described in step three include the length and area of ​​the contour. When the length or area of ​​the contour meets the set condition threshold, the corresponding channel is added to the channel list.

4. The magnetic resonance image channel selection method based on adaptive threshold and PCA according to claim 1, characterized in that: In the image data matrix, a matrix is ​​constructed with one pixel as one row and one channel as one column.

Images