TW201212891A - Method of two-dimentional Ensemble Empirical Mode Decomposition for analyzing brain disease - Google Patents

Abstract

The present invention provides a method of two-dimentional ensemble empirical mode decomposition (2D EEMD) for analyzing brain disease. It comprises the followed steps: providing a medical image; using the 2D EEMD method to decompose the medical image and to generate several decomposed images; choosing two or more decomposed images as texture images; combining the several texture images to get one target image; using convoluted calculation on the target image to get a first image; normalizing the first image to get a second image; using color-mapping method on the second image to get a color image; fusing the color image with the medical image to get a reference image; using the reference image to diagnose the disease area.

Description

201212891 VI. Description of the Invention: [Technical Field of the Invention] The present invention relates to a method for analyzing brain lesions, and a method for analyzing brain lesions by a two-dimensional empirical mode deconstruction method. [Prior Art] A stroke causes local obstruction or bleeding inside the cerebral blood vessels, such as an ischemic test (emb〇Usm), an ischemic infarction (such as her coffee), or a hemorrhagic rupture (hem_age). The brain tissue will be subjected to females, and hematoma or ischemia will occur, which will lead to neurological symptoms and phenomena, which may cause the patient to have disturbance of consciousness, or have other nervous system defects. The stroke of the brain is the second highest in the country. It is second only to the total of all neoplasms. If compared with the side malignant tumors, the brain towel is much higher than any single malignant tumor. Studies have shown that time has a great impact on the treatment of stroke, especially for acute cerebral apoplexy. If given appropriate treatment during prime time, it can increase the chance of stroke rehabilitation or reduce the level of disability. Among them, the prime time was three hours within the onset of the disease, and the treatment was given by intravenous injection of the lysing agent (Rec〇mbinant plasmin〇gen Activator, rt_PA), and rt_PA was given immediately within three hours. The chance of a good treatment outcome is 1.6^2.8 times that of unadministered, so thrombolytic agents can be considered as a specific drug for the treatment of acute ischemic stroke. However, the current non-developer brain computerized tomography Non-ContractCT (NCCT), which is commonly used to detect the location of stroke in the brain, is not sensitive to the ischemic changes in the ultra-acute ischemic stroke 201212891, and will often miss the gold of treatment. period. In the past, the Nortuki Takahashi team in Japan used the adaptive filter to improve the signal-to-noise ratio of computerized tomographic images to improve the recognition rate of the lesion area. However, the Norituki Takahashi team only increased the signal-to-noise ratio and did not use the film itself. The analysis of the characteristics of the acquired physiological signals, if the operation can be used to emphasize the lesion features of the computed tomography image, will improve the accuracy of the clinical investigations to make the most favorable treatment for patients.

[Summary of the Invention]

The invention provides a method for analyzing a brain lesion by using a two-dimensional empirical mode deconstruction method, comprising the steps of: providing a medical image; destructuring the medical image by using a two-dimensional 镰 mode deconstruction method; Deconstructing the image; manually or self-receiving the image to invert a plurality of suitable deconstructed images as a plurality of feature images; combining the images to obtain a target image; Performing a texture operation on the image to obtain a first image; Lai's first-child butterfly (10) paste-second scale; mapping the second image to a color image to obtain a color image; and merging the color image with the medical image Processing to obtain a reference image; paste the reference image full of a lesion. Reference miscellaneous caps 铋 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 4 4 201212891 The physiological signal of the tomographic image is a non-linear signal, and the information of the color image in the reference image is quite conducive to the interpretation of the lesion characteristics. Therefore, the _二_^_^0 method provided by the present invention can effectively calculate the feature region of the image and provide a faster and more accurate interpretation of the lesion location. When the brain _ disease _ lin, can be considered in the neonatal ischemic stroke, Ba's disease, righteousness, _ after disease, face and a bribery, head traumatic age-related fractures, Gu (four) blood, brain Contusion, stroke, cerebrovascular rupture

Hemorrhage, cerebral aneurysm, congenital arteriovenous malformation, benign brain tumor, intracranial malignant tumor, brain abscess, brain tuberculosis or hydrocephalus. Especially for the case of hyperacute cerebral apoplexy, it is more necessary to quickly judge the disease in prime time, so that the treatment is most beneficial to the patient. The advantages and spirit of the present invention, as well as the more detailed embodiments, can be further understood by the following embodiments and the accompanying drawings. [Embodiment] The present invention provides a two-dimensional empirical mode deconstruction method (two dimentional)

Ensemble Empirical Mode Decomposition^ 2D EEMD) A method for brain lesion analysis 'used to quickly interpret the location of a lesion in a stroke image by using this analysis method, especially for acute ischemic strokes' Treatment. The two-dimensional empirical mode deconstruction method mentioned above is the method of empirical mode decomposition (Empirical Mode Decompositioiij EMD) by Huang et al. After 2000, this technology is applied to image processing. Image processing is the application of two-dimensional empirical mode deconstruction. The empirical mode decomposition method is completed by the screening process through multiple iterations of 201212891. A number of deconstructed nas are extracted, in which different texture round images can be clearly displayed, which can be used for further analysis and calculation of the feature regions. Please refer to the first-graph, which is a method step of analyzing the brain lesions by the two-dimensional empirical mode deconstruction method of the present invention; and the following is the method step of analyzing the brain lesions by the two-dimensional surface mode destructive method:

St. 1: providing a medical image; the medical image has _pixels, and in this embodiment, the medical image is a brain computerized tomographic image of a stroke;

SteP2: the brain computerized tomographic image is removed from the skull signal; the brain computerized tomographic image contains an original signal, and the lower envelope information;

Step3: Performing a plurality of deconstructed images by using a two-dimensional empirical mode deconstruction method to obtain a plurality of deconstructed images; the deconstructed images have pixels, and the upper envelope information and the lower envelope information are averaged. To obtain an average envelope information, the average envelope information is subtracted from the original signal to obtain a first deconstructed image IMF(1) 'where the sum of the average envelope information is zero. The first deconstructed image history (1) includes a first original signal, and the first original signal includes first-up envelope information and first-lower envelope information, and then the first de-constructed shirt image IMF(l) is used as a reference An upper envelope information and a first lower envelope information are averaged to obtain a first average envelope information, and then the first original signal is subtracted from the first envelope information to obtain a second deconstructed image ^^(7); The second deconstructed trace (2) image is used as a reference, and so on, the third deconstructed image_(3), the fourth deconstructed image IMF(4)·.. is obtained; in this embodiment, the deconstructed processing results in 6^ deconstructed images. Best, but not limited to this. 201212891 Tian Ran's first deconstructed image secret (1), second deconstructed image IMF (2), third deconstructed image test (3), and the second image (4) have _pixel and first-average envelope information, second average The envelope information, the third average envelope information, and the fourth average envelope information are all zero. Step4 · Selecting the image from the deconstructed images by manual or automatic selection; the method of human jade is as follows: the first deconstructed image secret (1), the second deconstructed image (2), and the third Deconstruction ~ like IMF (3), 解 four deconstructed image secret (4) ... break, Wei out a number of deconstructed images with _ Jia value as feature images; automatic selection method such as side calculation of multiple pieces of deconstructed image energy, ie The sum of the absolute values of the amplitude, and the energy image is used as a plurality of feature images. Of course, the feature images also have m*n pixels.

St6p5: #_Feature image is merged to make a target image, which also has m*n pixels.

SteP6: using -i*j; the filter performs a texture operation (_〇luti〇n) on the target image to obtain a first image having a pixel; wherein 'the i*j value is smaller than the The value of x*y, and the value of i can be equal to the value of 』, and the value of χ can also be equal to the value of y; in the embodiment, the value of 'i value and 』 is 5, but not limited thereto. Wherein, the I*j filter is a calculated entropy value (entlOpy)_H. The meaning of the 'smoke is the characteristic value used to measure the random gray scale distribution, that is, if there is a group of discrete information, we need to get The average amount of information of the discrete information will be expressed by (10). This quantity represents the uncertainty of the transfer. In this embodiment, the first image is taken as an example, and each i* on the first image is used. j range has - side value, _ net value represents the uncertainty of this range respectively 201212891

It also has a specific elastic range change, and the formula for calculating the entropy (H) value is: H=-I(p*l〇g2(p)), P is the histogram of the first image. The probability of histogram counts appears. Inflammation P normalizes the figure of the first shirt to make one. The first image also has X*y pixels; the purpose of normalization is to make the first every static number have individual drops but疋The 焖 values are relative adjusted by the corresponding standard values.汾Print 8 · Color image mapping (color_mapping) of the second image to obtain a color image, and the smoke value normalized by the second image corresponds to a different color according to the numerical value. . Step9: Combining the color image with the medical image to obtain a reference tree; in this embodiment, the variegated image is merged with the brain computerized tomographic image of the brain stroke, and the fusion processing includes the above two The image is aligned, the color image is translucent, and the two sheets are superimposed to form a reference image.

SteplO uses the reference image to determine a lesion; in this embodiment, the physician can easily determine the lesion of the stroke through the reference image. 〃Please refer to the second figure 'County One-pixel 512*512 Brain Stroke Brain Computer Tomography Image' The brain computerized tomography image of the brain towel is described below as an example of the present invention.凊 Refer to the second figure, which is the image of the skull of the computerized tomographic image of the brain, and the information of the skull is removed by adjusting the image threshold esh〇ld or by selecting it. 201212891 Please refer to the fourth figure, which is to deconstruct the brain tomographic image of the brain stroke by 2DEEMD, and obtain six deconstructed images m〇del~m〇de6; the deconstructed images model~mode6 have 512* 512 pixels. After that, the physician used the experience to deconstruct the deconstruction and selected several deconstructed images with the judgment value as the feature image model~mode5. The feature images model~mode5 also have 512*512 pixels ° 'Please refer to the fifth figure. 'It is the lion image that the feature image m〇del~m() de5 enters and processes; the Muna is similar to 512*512 pixels. The target image is textured using a 5*5 filter to obtain a first image having 507*507 pixels. The first image is normalized to obtain a second image, and the second image also has 507*507 pixels.凊 Refer to the sixth® 'single image for color image mapping to a color image. Please refer to the seventh image and fuse the color image with the brain computerized tomographic image of the brain stroke to obtain a reference image. The doctor can easily determine the lesion at the lesion of the stroke through the reference image. The lesion at the stroke usually occurs on the reference image and the color is more symmetrical. Taking this actual remuneration as an example, the position of the lesion occurs at a position where the fine value is small, and the color is indicated. The reference image contained in the reference image is equivalent to the image of the phantom, and the brain computed tomography side of the brain stroke included in the reference image is a structural image, so that the combination of functional images and structural images enables It is easier for the physician to determine the location of the lesion. In addition, because the physiological signal of the computed tomography image is a non-linear signal, and the reference image contains color image information, this part of the information can be quite beneficial for the interpretation of the lesion 9 201212891. Therefore, the data segmentation method using the two-dimensional empirical mode deconstruction method provided by the present invention has the H test area and the 槪 细 细 细 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 In particular, the purchase of acute ischemic stroke requires the need to judge the disease at prime time, so that it is most beneficial to the patient. The present invention is based on the above, and the lining is not limited to the spirit and inventive entity of the present invention only in the above embodiments. Those skilled in the art will be able to easily understand and utilize other components or means to produce the same effect. Therefore, the advantages of the invention should be included in the following description. The following detailed description combines the advantages of the invention with the accompanying drawings, in which: The first figure: the method step diagram of the brain lesion analysis using 2DEEMD; the second picture: the brain computerized tomography image of the brain stroke of the present invention; the second picture: the brain computer of the brain stroke of the present invention The image after the removal of the information of the faulty shadow bone; the fourth picture: the paste of the invention 2DEEMD·the brain towel wind __ image is deconstructed, and six deconstructed images ^^^^ are obtained; fifth figure: godlike Depending on the image of the target image; the sixth image: the color image obtained by mapping the second image to the color image on this side; 201212891

Seventh figure: The reference image of the present invention is obtained by fusing a color image with a brain computerized tomographic image of a brain stroke. [Main component symbol description] None % 11

Claims (1)

201212891 VII. Patent application scope · The method for brain lesion analysis includes the following steps: a. - medical image 'the pixel of the medical image; b. providing a two-dimensional _ state destructuring method to deconstruct the medical image Processing, and obtaining a plurality of deconstructed images; c selecting a plurality of suitable deconstructed images as a plurality of characteristic images by means of artificial or self-funding 鼬 彡 ; ; ; ; ; ; ; ; ; ; ; ; ; d d d d d d d Go to the target image; e. texture-calculate the target image with the -i*i filter to create a first image having x*y pixels; f· coloring the first image Mapping, to produce a color image; & combining the color image with the medical image to create a reference image; and /h. using the reference image to determine a lesion. 2. A method for correcting brain lesions by using a two-dimensional modal modal method as described in the patent application, wherein the medical image is a computed tomography image. 3. The method for performing a brain lesion according to the two-dimensional empirical mode deconstruction method described in claim i of the patent scope, wherein after the first image is obtained in step e, the first image is first normalized. In order to obtain a second image, and then the second image is subjected to a color 6,1 * ' Γ Γ 0 4 4 patent scope 1st item ^ marriage two turn state deconstruction method for brain disease 12 201212891 variation analysis method The deconstructed image, the feature image, and the target image have pixels. White tool 5 · A method for performing brain lesion analysis using a two-dimensional empirical mode deconstruction method as described in item i of the patent application scope, wherein the second image has x*y pixels; The method of performing brain lesion analysis using the two-dimensional empirical mode deconstruction method according to item j, wherein the n value is smaller than the m value. " 7. For the method of brain disease change analysis by the two-dimensional fiber mode deconstruction method described in item J of the patent scope of the application of the month, wherein the value of i is smaller than the value of m. A method for cutting a brain lesion by a two-dimensional fiber mode deconstruction method, wherein the step a further comprises the step of removing the skull signal. 9. The "two-dimensional experience" as described in claim 1 The modal destructive method is a method for analyzing a brain lesion, wherein the medical image further comprises an original signal, and the original signal L 3 is enveloped with information and the following envelope information. 10. The method for finding a brain lesion analysis using the two-dimensional weaving mode deconstruction method as described in claim 9 of the patent application scope, wherein the deconstructing process of step b includes the upper envelope information and the lower envelope t-information Go to - Pingchi Jobs, re-submit the original signal minus the average envelope information to correspondingly get a first deconstructed image. A method for analyzing brain lesions by using a two-mode modal deconstruction method as described in claim 1 of the claim, wherein the sum of the average envelope information is zero. 12. The filter of the i*j敝m estimator value of the step e in the case of the brain lesion analysis by the two-dimensional weaving mode deconstruction method as claimed in the first claim. 13 201212891 13. A method for brain lesion analysis using a two-dimensional empirical mode deconstruction method as described in claim 1, wherein the lesion is a lesion of a brain lesion. 14. For the method of brain lesion analysis using the two-dimensional empirical mode deconstruction method described in Item 13 of the patent application, the lesions of the genus of the genus are classified as stagnation, financial dysfunction, Parkinson's disease, epilepsy, brain tumor. Aftereffects, meningoencephalitis and Alzheimer's disease, head trauma caused by intracranial hemorrhage, brain contusion, stroke, cerebrovascular rupture, cerebral aneurysm congenital arteriovenous malformation, benign brain tumor, intracranial malignancy Tumor, brain pus, brain tuberculosis or hydrocephalus.