A method and device for calculating cerebral hypoperfusion area based on ASL technology

Abstract

The present invention provides a method and apparatus for calculating cerebral hypoperfusion areas based on ASL technology. The method comprises: obtaining a cerebral blood flow image of a region of interest; segmenting the cerebral blood flow image based on the axis of symmetry of the cerebral blood flow image to obtain a first cerebral blood flow image and a second cerebral blood flow image; flipping the first cerebral blood flow image to obtain a reference blood flow image corresponding to the second cerebral blood flow image; and obtaining hypoperfusion areas in the second cerebral blood flow image based on the reference blood flow image and the second cerebral blood flow image. By dividing the cerebral blood flow image into the first and second cerebral blood flow images using the axis of symmetry of the cerebral blood flow image, the present invention takes into account the presence of hyposignal areas on both sides of the entire brain, thereby accurately segmenting hypoperfusion areas in the cerebral blood flow image and avoiding missed detections.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a method and device for calculating cerebral hypoperfusion areas based on ASL technology. Background Art

[0002] Acute ischemic stroke is caused by occlusion of a cerebral artery, leading to brain infarction and associated damage to neurons, astrocytes, and oligodendrocytes. It is the most significant central nervous system vascular event causing death and disability in modern society. Acute ischemic stroke accounts for approximately 80% of all strokes. The key to its treatment lies in the early re-opening of blocked vessels and the rescue of the ischemic penumbra. The ischemic penumbra is crucial for the treatment of acute ischemic stroke, and its assessment and treatment are highly valued by clinicians. Therefore, the ability to clearly identify the ischemic penumbra area is a key concern for clinicians. Currently, in clinical practice, the ischemic penumbra is generally assessed visually.

[0003] Arterial spin labeling (ASL) is an MRI technique that uses water molecules in the blood as endogenous, freely diffusible tracers for cranial perfusion imaging. ASL technology has been around for over 20 years and has undergone several stages of development. With the continuous advancement of ASL technology, especially the recent application of pseudo-continuous ASL (pCASL) sequences, its image quality, imaging range, and imaging speed have greatly improved. It has gradually attracted the attention of imaging and neuroscience researchers and is increasingly being used in scientific research and clinical practice. The basic principle of ASL imaging is to acquire data twice to generate a pair of labeled and control images. The static tissue signals in the labeled and control images are identical; the difference lies in whether the inflow of blood is reversed. The so-called labeling process involves applying an inversion pulse to the neck to mark the area. This causes the water molecules in the labeled blood flowing into the artery to reverse 180 degrees. After a certain period of time, the blood flows into the target layer. Due to the difference between the labeled (inverted) blood and the unlabeled blood signal, the labeled image and the control image are silhouetted. After the static tissue signal is clipped, only the difference between the labeled and unlabeled blood flow signals is displayed. This repeated acquisition and averaging produces a cerebral blood flow map (CBF), which reflects the blood flow and distribution in the brain at a specific moment.

[0004] Currently, the methods of penumbra imaging are generally divided into two types: perfusion with medication and perfusion without medication. Perfusion with medication may cause the following problems: 1. Radiation damage; 2. High cost; 3. Poor comfort; 4. Semi-quantitative problem. Compared with these technologies that require perfusion with medication, the first and most important advantage of ASL technology is that it does not require medication. Secondly, it also has advantages in comfort and cost. At present, ASL technology occupies a very high position in the clinical perfusion application of hospitals. It should often be used for brain examinations in various hospitals. ASL can quantitatively and in vivo measure cerebral blood flow in units of ml·100g -1 min -1 , which represents the total amount of blood flow reaching a certain weight of tissue per unit time.

[0005] ASL provides CBF parameters, which are important parameters reflecting the stability of cerebral hemodynamics and have outstanding advantages in the treatment of cerebrovascular diseases. Imaging examinations for acute stroke patients need to be rapid and efficient, and ASL can provide important hemodynamic information in a relatively short period of time. ASL can detect the presence of an ischemic penumbra. In clinical applications, physicians generally assess the penumbra with the naked eye. The "Chinese Expert Consensus on Clinical Evaluation and Treatment of Ischemic Penumbra in Acute Cerebral Infarction" defines a hypoperfusion zone as an area on ASL where the CBF is less than 40% of the contralateral side. Currently, there is a problem of missed detection in the identification of hypoperfusion zones. Summary of the Invention

[0006] In view of the above problems, the present invention proposes a method and device for calculating cerebral hypoperfusion areas based on ASL technology, which overcome the above problems or at least partially solve the above problems.

[0007] According to a first aspect of the present invention, a method for calculating cerebral hypoperfusion areas based on ASL technology is provided, the method comprising:

[0008] Acquire cerebral blood flow images in the region of interest;

[0009] segmenting the cerebral blood flow image based on a symmetry axis of the cerebral blood flow image to obtain a first cerebral blood flow image and a second cerebral blood flow image;

[0010] flipping the first cerebral blood flow image to obtain a reference blood flow image corresponding to the second cerebral blood flow image;

[0011] Based on the reference blood flow image and the second cerebral blood flow image, a low perfusion area in the second cerebral blood flow image is acquired.

[0012] Optionally, before flipping the first cerebral blood flow image to obtain a reference blood flow image corresponding to the second cerebral blood flow image, the method further includes:

[0013] Acquire a first brain structure image and an ASL image corresponding to the position of the cerebral blood flow image, wherein the first brain structure image includes cerebrospinal fluid;

[0014] performing registration using the ASL image as a fixed image and the first brain structure image as a floating image to obtain a second brain structure image corresponding to the first brain structure image;

[0015] Inputting the second brain structure image into a trained neural network model, and using the neural network model to output the symmetry axis of the cerebral blood flow image, wherein the neural network model is trained with a brain structure image including cerebrospinal fluid and a corresponding symmetry axis labeled image; or

[0016] A cerebrospinal fluid image is segmented from the second brain structure image, and a symmetry axis of the cerebral blood flow image is acquired based on the cerebrospinal fluid image.

[0017] Optionally, before flipping the first cerebral blood flow image to obtain a reference blood flow image corresponding to the second cerebral blood flow image, the method further includes:

[0018] Acquire a first brain structure three-dimensional image and an ASL three-dimensional image of a region of interest, wherein the first brain structure three-dimensional image includes a plurality of slices, and each slice includes cerebrospinal fluid;

[0019] performing registration using the ASL three-dimensional image as a fixed image and the first three-dimensional image of the brain structure as a floating image to obtain a second three-dimensional image of the brain structure;

[0020] Inputting the second three-dimensional image of the brain structure into a trained neural network model, and using the neural network model to output the symmetry axis of the cerebral blood flow image, wherein the neural network model is trained using the three-dimensional image of the brain structure including cerebrospinal fluid and the corresponding symmetry axis labeled image; or

[0021] A cerebrospinal fluid image is segmented from a slice having the largest cerebrospinal fluid area in the second three-dimensional image of the brain structure, and a symmetry axis of the cerebral blood flow image is acquired based on the cerebrospinal fluid image.

[0022] Optionally, acquiring the symmetry axis of the cerebral blood flow image based on the cerebrospinal fluid image includes:

[0023] inverting the cerebrospinal fluid region in the cerebrospinal fluid image to obtain a gap region between the cerebrospinal fluid region and the edge of the frame;

[0024] Identify a first gap region located above the cerebrospinal fluid image and a second gap region located below the cerebrospinal fluid image, obtain a first coordinate point corresponding to a maximum abscissa of the first gap region, and obtain a second coordinate point corresponding to a minimum abscissa of the second gap region;

[0025] Calculating an offset angle based on a straight line passing through the first coordinate point and the second coordinate point;

[0026] An image symmetry axis is determined using the centroid of the cerebrospinal fluid image and the offset angle.

[0027] Optionally, before inverting the cerebrospinal fluid region in the cerebrospinal fluid image to obtain a gap region between the cerebrospinal fluid region and the edge of the frame, the method further includes:

[0028] Find four vertices of the cerebrospinal fluid in the cerebrospinal fluid image, and select the maximum abscissa value of the two upper vertices, the minimum abscissa value of the two lower vertices, the maximum ordinate value of the two left vertices, and the minimum ordinate value of the two right vertices;

[0029] The cerebrospinal fluid region is constructed using the maximum value of the abscissa, the minimum value of the abscissa, the maximum value of the ordinate, and the minimum value of the ordinate.

[0030] Optionally, flipping the first cerebral blood flow image to obtain a reference blood flow image corresponding to the second cerebral blood flow image includes:

[0031] acquiring a brain structure image corresponding to the cerebral blood flow image;

[0032] Based on the symmetry axis of the cerebral blood flow image, acquiring a third brain structure image corresponding to the first cerebral blood flow image and a fourth brain structure image corresponding to the second cerebral blood flow image;

[0033] flipping the third brain structure image, taking the fourth brain structure image as a fixed image and the flipped third brain structure image as a floating image, and obtaining a deformation field of the flipped third brain structure image relative to the fourth brain structure image;

[0034] Based on the deformation field, the flipped first cerebral blood flow image is adjusted to obtain a reference blood flow image.

[0035] Optionally, acquiring a low perfusion area in the second cerebral blood flow image based on the reference blood flow image and the second cerebral blood flow image includes:

[0036] Calculating the cerebral blood flow ratio of corresponding pixels of the reference blood flow image and the second cerebral blood flow image;

[0037] The areas with cerebral blood flow ratio less than the set threshold were marked and morphologically processed to obtain low perfusion areas.

[0038] According to a second aspect of the present invention, there is provided a device for calculating cerebral hypoperfusion areas based on ASL technology, the device comprising:

[0039] A cerebral blood flow image acquisition module is used to acquire cerebral blood flow images of the region of interest;

[0040] a segmentation module, configured to segment the cerebral blood flow image based on a symmetry axis of the cerebral blood flow image to obtain a first cerebral blood flow image and a second cerebral blood flow image;

[0041] a reference blood flow image acquisition module, configured to flip the first cerebral blood flow image to acquire a reference blood flow image corresponding to the second cerebral blood flow image;

[0042] The low perfusion area acquisition module is used to acquire the low perfusion area in the second cerebral blood flow image based on the reference blood flow image and the second cerebral blood flow image.

[0043] According to a third aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium is used to store program code, and the program code is used to execute the method for calculating cerebral hypoperfusion areas based on ASL technology according to any one of the first aspects.

[0044] According to a fourth aspect of the present invention, a computing device is provided, comprising a processor and a memory: the memory is used to store program code and transmit the program code to the processor; the processor is used to execute the method for calculating cerebral hypoperfusion areas based on ASL technology as described in any one of the first aspects according to instructions in the program code.

[0045] The present invention provides a method and device for calculating cerebral hypoperfusion areas based on ASL technology. After acquiring a cerebral blood flow image of a region of interest, the image is divided into a first cerebral blood flow image and a second cerebral blood flow image by using the symmetry axis of the cerebral blood flow image. This allows for accurate segmentation of hypoperfusion areas in the cerebral blood flow image, avoiding missed detections, by taking into account the presence of hypointensity areas on both sides of the brain. Furthermore, through cerebrospinal fluid image segmentation, axis of symmetry acquisition, and ASL hypointensity domain segmentation, hypoperfusion areas can be segmented strictly according to symmetrical region comparison, reducing false positives and simultaneously accommodating bilateral hypoperfusion. Compared to manual and workstation-based delineation results, the method is more objective and reduces labor and time costs.

[0046] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are specifically listed below.

[0047] Based on the following detailed description of specific embodiments of the present invention in conjunction with the accompanying drawings, those skilled in the art will become more aware of the above and other objects, advantages and features of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present invention. The same reference symbols are used throughout the drawings to represent the same components. In the drawings:

[0049] Figure 1a 、 Figure 1b The color schematic diagram and grayscale schematic diagram of cerebral hypoperfusion in the temporal lobe and frontal lobe are shown respectively;

[0050] Figure 2a 、 Figure 2b The color schematic and grayscale schematic diagrams show the presence of low perfusion areas on both the left and right sides, respectively;

[0051] Figure 3 A schematic flow chart of a method for calculating cerebral hypoperfusion areas based on ASL technology according to an embodiment of the present invention is shown;

[0052] Figure 4a 、 Figure 4b A color schematic diagram and a grayscale schematic diagram respectively show evaluation results of determining a low perfusion area based on the ASL technology according to an embodiment of the present invention;

[0053] Figure 5 A schematic diagram of the structure of a method and apparatus for calculating cerebral hypoperfusion areas based on ASL technology according to an embodiment of the present invention is shown;

[0054] Figure 6 A schematic structural diagram of a method and apparatus for calculating cerebral hypoperfusion areas based on ASL technology according to another embodiment of the present invention is shown. DETAILED DESCRIPTION

[0055] Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art. It should be noted that some of the drawings in the specification use color schematic diagrams to more clearly express the nuclear magnetic resonance images and the low perfusion area evaluation effect of the method provided in this embodiment.

[0056] MRI stands for Magnetic Resonance Imaging. Its principle is that the energy released (also called the signal) attenuates differently in different structural environments within a substance. By detecting the emitted electromagnetic waves through an external magnetic field, the position and type of the atomic nuclei that make up the object can be determined, thereby creating an image of the object's internal structure. During the MRI imaging process, by varying the factors influencing the MR signal, different images can be obtained. These different images are called sequences. For example, by weighting according to the T1 value, a T1 sequence is obtained, while by weighting according to the T2 value, a T2 sequence is obtained. A single case can have multiple sequences, each consisting of many slices.

[0057] ASL provides CBF parameters, a crucial indicator of cerebral hemodynamic stability, and has significant advantages in treating cerebrovascular disease. To the inventor's knowledge, the identification of cerebral hypoperfusion areas primarily involves the following methods: 1. Visual identification by clinicians, the most common method; 2. Delineation of hypoperfusion areas using a dedicated workstation; and 3. Using the median grayscale value of the affected side as the normal brain tissue signal value, segmenting areas with grayscale values below 40% of the normal brain tissue signal as hypoperfusion areas.

[0058] First, both the clinician's eye recognition and the workstation's delineation are subjective, and they also consume more energy and require more labor costs. Low repeatability is also one of the disadvantages of the above two methods. Some areas in the ASL-CBF image also appear as low signals. Using the current algorithm to solve this problem will cause multiple detections, and some areas that appear as low signal areas will be detected. Figure 1a 、 Figure 1b As shown in the figure, low signals are present in the temporal lobe and frontal lobe, and the cerebrospinal fluid area also presents a low signal area. The existing algorithm results can easily identify these parts as low perfusion areas. In addition, the existing algorithm only considers the situation of unilateral low signal areas, and does not consider the situation where low signal areas exist on both sides of the brain. The situation where low signal areas exist on both sides of the patient's left and right brain is not considered, which will lead to missed detection. Figure 2a 、 Figure 2b As shown in the CBF image, we can clearly see low perfusion areas on both sides of the patient. The existing algorithm cannot fully detect this situation and may miss detections.

[0059] The embodiment of the present invention provides a method for calculating cerebral hypoperfusion areas based on ASL technology, such as Figure 3 The method for calculating cerebral hypoperfusion areas based on ASL technology in an embodiment of the present invention may at least include the following steps S301 to S304.

[0060] S301, obtaining a cerebral blood flow image of a region of interest. The cerebral blood flow image of the region of interest may be a cerebral blood flow (CBF) image of the brain.

[0061] S302: Segment the cerebral blood flow image based on the symmetry axis of the cerebral blood flow image to obtain a first cerebral blood flow image and a second cerebral blood flow image. The first cerebral blood flow image is a left-brain cerebral blood flow image, and the second cerebral blood flow image is a right-brain cerebral blood flow image; or the first cerebral blood flow image is a right-brain cerebral blood flow image, and the second cerebral blood flow image is a left-brain cerebral blood flow image.

[0062] S303: Flip the first cerebral blood flow image to obtain a reference blood flow image corresponding to the second cerebral blood flow image.

[0063] S304 : Acquire a low perfusion area in the second cerebral blood flow image based on the reference blood flow image and the second cerebral blood flow image.

[0064] The obtained first cerebral blood flow image is flipped to obtain a reference blood flow image corresponding to the second cerebral blood flow image, and then the hypoperfusion area of the second cerebral blood flow image can be obtained based on the reference blood flow image and the second cerebral blood flow image. The method provided in this embodiment divides the cerebral blood flow image into the first cerebral blood flow image and the second cerebral blood flow image using the symmetry axis of the cerebral blood flow image, taking into account the presence of hyposignal areas on both sides of the entire brain, and can thus accurately segment the hypoperfusion area in the cerebral blood flow image to avoid missed detection.

[0065] As mentioned in step S302 above, the cerebral blood flow image is segmented based on the symmetry axis of the cerebral blood flow image to obtain a first cerebral blood flow image and a second cerebral blood flow image. Optionally, the symmetry axis of the cerebral blood flow image can be determined before this. In this embodiment, the symmetry axis of the cerebral blood flow image can be obtained by following steps A1 to A4.

[0066] A1. Acquire a first brain structure image and an ASL image corresponding to the location of the cerebral blood flow image. The first brain structure image includes cerebrospinal fluid. The brain structure image in this embodiment can be understood as a T1 image, which is easy to obtain and provides clear brain structure. The ASL image consists of two images: a control image (an image without blood labeling) and a labeled image. Both control and labeled images can be used in this embodiment of the present invention.

[0067] A2: Register the ASL image as the fixed image and the first brain structural image as the floating image to obtain a second brain structural image corresponding to the first brain structural image. Optionally, during the registration, an affine transformation is performed on the first brain structural image and the ASL image to achieve image coordinate matching between the first and ASL images, thereby obtaining a second brain structural image, denoted as warp T1 in this embodiment.

[0068] A3 inputs the second brain structure image into a trained neural network model, and uses the neural network model to output the symmetry axis of the cerebral blood flow image. The neural network model is trained using a brain structure image including cerebrospinal fluid and a corresponding symmetry axis labeled image. The neural network model can be trained using a convolutional neural network (CNN) or other neural network, which are well known to those skilled in the art and will not be described in detail here.

[0069] A4, segmenting the cerebrospinal fluid image from the second brain structure image, and obtaining the axis of symmetry of the cerebral blood flow image based on the cerebrospinal fluid image. After obtaining the second brain structure image obtained by registering the first brain structure image, the cerebrospinal fluid image F_Mask can be segmented from the second brain structure image. Optionally, a threshold setting method can be used, that is, a signal threshold is set. Based on the signal threshold, the signal exceeding the threshold in the second brain structure image warp T1 can be determined as cerebrospinal fluid, thereby obtaining a cerebrospinal fluid image. The axis of symmetry is then determined based on the cerebrospinal fluid image to segment the first and second cerebral blood flow images.

[0070] This embodiment provides two methods for obtaining the symmetry axis of the cerebral blood flow image. In actual applications, one of the methods or a combination of the two methods can be selected according to different needs, and this embodiment does not limit this.

[0071] In this embodiment, the T1 sequence is a three-dimensional sequence, which is composed of multiple slices. Optionally, in an optional embodiment of the present invention, the symmetry axis of the cerebral blood flow image can also be obtained through the following steps B1 to B4.

[0072] B1. Acquire a first brain structure three-dimensional image and an ASL three-dimensional image of a region of interest. The first brain structure three-dimensional image includes multiple slices, and each slice includes cerebrospinal fluid.

[0073] B2, registering the ASL 3D image as a fixed image and the 3D image of the first brain structure as a floating image to obtain a 3D image of the second brain structure.

[0074] B3, inputting the three-dimensional image of the second brain structure into a trained neural network model, and using the neural network model to output the symmetry axis of the cerebral blood flow image, where the neural network model is trained using the three-dimensional image of the brain structure including cerebrospinal fluid and the corresponding symmetry axis labeled image; or;

[0075] B4, segmenting a cerebrospinal fluid image from the slice with the largest cerebrospinal fluid area in the three-dimensional image of the second brain structure, and obtaining a symmetry axis of the cerebral blood flow image based on the cerebrospinal fluid image.

[0076] The method provided in this embodiment requires segmenting a cerebrospinal fluid image from the slice with the largest cerebrospinal fluid area in the three-dimensional image of the second brain structure, and then determining the symmetry axis of the cerebral blood flow image based on the cerebrospinal fluid image segmented from the slice with the largest cerebrospinal fluid area. Optionally, step B4 of segmenting the cerebrospinal fluid image from the slice with the largest cerebrospinal fluid area in the three-dimensional image of the second brain structure may include:

[0077] B4-1, segmenting the cerebrospinal fluid image from the slice with the largest cerebrospinal fluid area in the three-dimensional image of the second brain structure by a threshold setting method. Specifically, B4-1 may include:

[0078] B4-1-1, obtain the slice with the largest CSF area in the 3D image of the second brain structure. As previously mentioned, the 3D image of the first brain structure includes multiple slices. To determine the slice with the largest CSF area, the sum of the CSF pixel values in each slice can be calculated. The slice with the largest sum of pixel values is the slice with the largest CSF area.

[0079] The symmetry axis of the cerebral blood flow image obtained based on the cerebrospinal fluid image may specifically include:

[0080] B4-2, find the four vertices of the cerebrospinal fluid in the cerebrospinal fluid image, and select the maximum abscissa value of the two upper vertices, the minimum abscissa value of the two lower vertices, the maximum ordinate value of the two left vertices, and the minimum ordinate value of the two right vertices.

[0081] B4-3, construct the cerebrospinal fluid region with the maximum value of the abscissa, the minimum value of the abscissa, the maximum value of the ordinate, and the minimum value of the ordinate.

[0082] Taking the slice with the largest cerebrospinal fluid area in the three-dimensional image of the brain structure as an example, find the four vertices a1, a2, a3, and a4 of the cerebrospinal fluid in the slice, and select the maximum horizontal coordinate x of the two upper vertices. max , the minimum value of the horizontal coordinate x between the two vertices below min , the maximum value y of the ordinate of the two vertices on the left max , the minimum y value of the vertical coordinate of the two vertices on the right min , construct the cerebrospinal fluid area [x min :x max ,y min :y max In this embodiment, by respectively limiting the minimum and maximum values of the horizontal and vertical coordinates, the cerebrospinal fluid area can be narrowed, thereby obtaining the upper and lower gap areas, and effectively determining the upper and lower concave points.

[0083] B4-4, the cerebrospinal fluid region is inverted to obtain the gap region between the cerebrospinal fluid region and the edge of the frame. For the obtained cerebrospinal fluid image, other regions in the image except the cerebrospinal fluid region can be used as the gap region. In this embodiment, the obtained cerebrospinal fluid region is inverted, and the four gap regions C above, below, left, and right of the cerebrospinal fluid region and the edge of the frame are obtained. up 、C down 、C left 、C right .

[0084] B4-5, identify the first gap area located above the cerebrospinal fluid image and the second gap area located below, obtain the first coordinate point corresponding to the maximum horizontal coordinate based on the first gap area, and obtain the second coordinate point corresponding to the minimum horizontal coordinate based on the second gap area.

[0085] Remove the gap areas on both sides, and find the point T1(x1, y1) with the maximum horizontal coordinate in the upper gap area and the point T2(x2, y2) with the minimum horizontal coordinate in the lower gap area, which are the upper and lower convex points T1(x1, y1) and concave points T2(x2, y2) of the cerebrospinal fluid.

[0086] B4-6, calculating the offset angle based on the straight line passing through the first coordinate point and the second coordinate point.

[0087] Determine the straight line passing through T1 (x1, y1) and T2 (x2, y2), and calculate the offset angle. The calculation formula for the offset angle is as follows:

[0088]

[0089] B4-7, determine the image symmetry axis using the centroid and offset angle of the CSF image.

[0090] The offset angle is the angle between the straight line through T1 (x1, y1) and T2 (x2, y2) and the x-axis, and the symmetry axis is the plane formed by the straight line through T1 (x1, y1) and T2 (x2, y2) mapped to each layer of the image in space. i and offset angle The symmetry axis of the cerebral blood flow image can be determined.

[0091] The processing of steps B4-2 to B4-7 can also be applied to the symmetry axis for obtaining the cerebral blood flow map in A4.

[0092] The above step S303 of flipping the first cerebral blood flow image to obtain a reference blood flow image corresponding to the second cerebral blood flow image includes the following steps C1 to C4.

[0093] C1, obtain the brain structure image corresponding to the cerebral blood flow image (denoted as CBF image in this embodiment). For the cerebral blood flow image of the region of interest, it is necessary to obtain the brain structure image corresponding to its position (denoted as T1 image in this embodiment).

[0094] C2, based on the symmetry axis of the cerebral blood flow image, obtain a third brain structure image (L_T1) corresponding to the first cerebral blood flow image (L_CBF) and a fourth brain structure image (R_T1) corresponding to the second cerebral blood flow image (R_CBF).

[0095] The symmetry axis of the acquired cerebral blood flow image can be used to segment the brain structure image to obtain a third brain structure image corresponding to the first cerebral blood flow image and a fourth brain structure image corresponding to the second cerebral blood flow image. In this embodiment, the first image represents the left brain and the second image represents the right brain. In actual applications, the first image can also represent the right brain and the second image can represent the left brain.

[0096] C3, flipping the third brain structure image, taking the fourth brain structure image as the fixed image and the flipped third brain structure image as the floating image, and obtaining the deformation field of the flipped third brain structure image relative to the fourth brain structure image;

[0097] The third brain structure image (L_T1) is flipped to obtain a flipped image (f_T1_L). The fourth brain structure image (R_T1) is used as a fixed image and the flipped image (f_T1_L) is used as a moving image for registration (e.g., non-rigid registration) to obtain the deformation field (Ng_T1_L).

[0098] C4, based on the deformation field obtained in C3, adjusts the flipped first cerebral blood flow image to obtain a reference blood flow image.

[0099] The first cerebral blood flow map (L_CBF) is flipped to obtain the flipped first cerebral blood flow map (fl_CBF_L), and the deformation field (Ng_T1_L) is applied to the flipped first cerebral blood flow map (fl_CBF_L), fl_CBF_L=f Ng_T1_L (fl_CBF_L) to obtain the reference blood flow image.

[0100] The above step S304, based on the reference blood flow image and the second cerebral blood flow image, obtaining the low perfusion area in the second cerebral blood flow image includes:

[0101] D1, calculating the cerebral blood flow ratio of corresponding pixels between the reference blood flow image and the second cerebral blood flow image;

[0102] D2, mark the area where the cerebral blood flow ratio is less than the set threshold and perform morphological processing to obtain the low perfusion area.

[0103] In step C4 above, the deformation field (Ng_T1_L) has been applied to the flipped first cerebral blood flow map (fl_CBF_L). Further, the CBF value ratio of the corresponding pixels of the flipped first cerebral blood flow map (fl_CBF_L) and its corresponding second cerebral blood flow map (R_CBF) can be calculated.

[0104]

[0105] Among them, the (i, j) table has i rows and j columns, indicating the pixel position in the image.

[0106] On the other hand, this embodiment may also include step C5: flipping the fourth brain structure image, and using the third brain structure image as a fixed image and the flipped fourth brain structure image as a floating image, to obtain a deformation field of the flipped fourth brain structure image relative to the third brain structure image; C6, based on the deformation field obtained in step C5, adjusting the flipped first cerebral blood flow map to obtain a reference blood flow image.

[0107] That is, the fourth brain structure image (R_T1) is flipped to obtain a flipped image (f_T1_R), and the third brain structure image (R_T1) is used as a fixed image (fixed image), and the flipped image (f_T1_R) is used as a moving image for registration (such as non-rigid registration) to obtain a deformation field (Ng_T1_R); further, the second cerebral blood flow map (R_CBF) is flipped to obtain a flipped second cerebral blood flow map (fl_CBF_R), and the deformation field (Ng_T1_L) is applied to the flipped second cerebral blood flow map (fl_CBF_R), fl_CBF_R=f Ng_T1_R (fl_CBF_R) to obtain the reference blood flow image.

[0108] Correspondingly, in steps D1 to D2, the CBF value ratio of corresponding pixels of the flipped second cerebral blood flow map (fl_CBF_R) and the corresponding first cerebral blood flow map (L_CBF) can be calculated;

[0109]

[0110] θ L θ RThe area with a ratio less than the set threshold is marked and morphologically processed to obtain the low perfusion area. To achieve the above purpose, we use T1 sequence and ASL sequence for multimodal image registration. After determining the image symmetry axis image, the image symmetry axis image is mapped to the CBF image to obtain the left and right brain images, and non-rigid registration is performed. The purpose of using non-rigid registration is to be able to compare each pixel with the corresponding pixel on the opposite side in turn to obtain the low perfusion area. The example of the result obtained using this method is shown in the figure below. Figure 4a 、 Figure 4b As shown, the left side is the CBF image; the right side pseudo-color image represents the low perfusion area where the CBF ratio is less than the set threshold.

[0111] Generally speaking, the input of this embodiment is the following three parts: brain structure image (T1 image), cerebral blood flow image (CBF image) and ASL image (ASL control image). The process mainly includes the following three parts: multimodal image mutual information registration of brain structure image and ASL image, acquisition of image symmetry axis and deformation field of left and right brain based on brain structure image, acquisition of reference cerebral blood flow images of left and right cerebral blood flow images based on deformation field, and segmentation of CBF low signal domain based on reference cerebral blood flow image. The embodiment of the present invention uses cerebrospinal fluid image to determine the image symmetry axis, and uses T1 image to determine the cerebrospinal fluid image. T1 image presents each area of brain tissue very clearly and intuitively, which allows us to determine the location of cerebrospinal fluid area, and then accurately segment the low perfusion area in CBF image, taking into account the situation where low signal areas exist on both sides of the whole brain, to avoid missed detection.

[0112] The results of the present invention are more objective than those obtained manually or by workstation delineation, and they require less labor and time. The present invention compares corresponding pixels in the left and right brain after non-rigid registration. Compared to existing algorithms, the present invention can segment hypoperfusion areas strictly based on symmetrical regional comparison, reducing false positives and accommodating bilateral hypoperfusion.

[0113] Based on the same inventive concept, an embodiment of the present invention further provides a device for calculating cerebral hypoperfusion areas based on ASL technology, such as Figure 5 As shown, the device for calculating cerebral hypoperfusion areas based on ASL technology in this embodiment includes:

[0114] A cerebral blood flow image acquisition module 510 is used to acquire a cerebral blood flow image of a region of interest;

[0115] a segmentation module 520 for segmenting the cerebral blood flow image based on the symmetry axis of the cerebral blood flow image to obtain a first cerebral blood flow image and a second cerebral blood flow image;

[0116] a reference blood flow image acquisition module 530 for flipping the first cerebral blood flow image to acquire a reference blood flow image corresponding to the second cerebral blood flow image;

[0117] The low perfusion region acquisition module 540 is configured to acquire the low perfusion region in the second cerebral blood flow image based on the reference blood flow image and the second cerebral blood flow image.

[0118] In an optional embodiment of the present invention, Figure 6 As shown, the apparatus for calculating cerebral hypoperfusion areas based on the ASL technology in this embodiment may further include a symmetry axis acquisition module 550;

[0119] A symmetry axis acquisition module 550 is configured to acquire a first brain structure image and an ASL image corresponding to a position of the cerebral blood flow image, wherein the first brain structure image includes cerebrospinal fluid;

[0120] The ASL image is used as a fixed image and the first brain structure image is used as a floating image for registration, thereby obtaining a second brain structure image corresponding to the first brain structure image;

[0121] Inputting the second brain structure image into a trained neural network model, and using the neural network model to output the symmetry axis of the cerebral blood flow image, the neural network model is trained with the brain structure image including cerebrospinal fluid and the corresponding symmetry axis labeled image; or,

[0122] A cerebrospinal fluid image is segmented from the second brain structure image, and a symmetry axis of the cerebral blood flow image is obtained based on the cerebrospinal fluid image.

[0123] In an optional embodiment of the present invention, the symmetry axis acquisition module 550 may also be used to:

[0124] Acquire a first brain structure three-dimensional image and an ASL three-dimensional image of a region of interest, wherein the first brain structure three-dimensional image includes a plurality of slices, and each slice includes cerebrospinal fluid;

[0125] The ASL 3D image is used as a fixed image and the 3D image of the first brain structure is used as a floating image for registration to obtain a 3D image of the second brain structure;

[0126] Inputting the three-dimensional image of the second brain structure into a trained neural network model, and using the neural network model to output the symmetry axis of the cerebral blood flow image, the neural network model being trained with the three-dimensional image of the brain structure including cerebrospinal fluid and the corresponding symmetry axis labeled image; or

[0127] A cerebrospinal fluid image is segmented from a slice having the largest cerebrospinal fluid area in the three-dimensional image of the second brain structure, and a symmetry axis of a cerebral blood flow image is obtained based on the cerebrospinal fluid image.

[0128] In an optional embodiment of the present invention, the symmetry axis acquisition module 550 may also be used to:

[0129] The cerebrospinal fluid region in the cerebrospinal fluid image is inverted to obtain the gap region between the cerebrospinal fluid region and the edge of the frame;

[0130] Identify a first gap region located above the cerebrospinal fluid image and a second gap region located below the cerebrospinal fluid image, obtain a first coordinate point corresponding to a maximum abscissa value of the first gap region, and obtain a second coordinate point corresponding to a minimum abscissa value of the second gap region;

[0131] Calculating an offset angle based on a straight line passing through the first coordinate point and the second coordinate point;

[0132] The image symmetry axis was determined using the centroid of the CSF image and the offset angle.

[0133] In an optional embodiment of the present invention, the symmetry axis acquisition module 550 may also be used to:

[0134] Find the four vertices of the cerebrospinal fluid and select the maximum abscissa value of the two upper vertices, the minimum abscissa value of the two lower vertices, the maximum ordinate value of the two left vertices, and the minimum ordinate value of the two right vertices;

[0135] The cerebrospinal fluid region was constructed using the maximum value of the abscissa, the minimum value of the abscissa, the maximum value of the ordinate, and the minimum value of the ordinate.

[0136] In an optional embodiment of the present invention, the reference blood flow image acquisition module 530 may also be used to:

[0137] obtaining a brain structure image corresponding to the cerebral blood flow image;

[0138] Based on the symmetry axis of the cerebral blood flow image, acquiring a third brain structure image corresponding to the first cerebral blood flow image and a fourth brain structure image corresponding to the second cerebral blood flow image;

[0139] flipping the third brain structure image, taking the fourth brain structure image as a fixed image and the flipped third brain structure image as a floating image, and obtaining a deformation field of the flipped third brain structure image relative to the fourth brain structure image;

[0140] Based on the deformation field, the flipped first cerebral blood flow image is adjusted to obtain a reference blood flow image.

[0141] In an optional embodiment of the present invention, the low perfusion region acquisition module 540 may also be used to:

[0142] Calculating the cerebral blood flow ratio of corresponding pixels between the reference blood flow image and the second cerebral blood flow image;

[0143] The areas with cerebral blood flow ratio less than the set threshold were marked and morphologically processed to obtain low perfusion areas.

[0144] An optional embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium is used to store program code, and the program code is used to execute the method for calculating cerebral hypoperfusion areas based on ASL technology described in the above embodiment.

[0145] An optional embodiment of the present invention further provides a computing device, which includes a processor and a memory: the memory is used to store program code and transmit the program code to the processor; the processor is used to execute the method for calculating brain hypoperfusion areas based on ASL technology described in the above embodiment according to instructions in the program code.

[0146] Those skilled in the art will clearly understand that the specific working processes of the systems, devices, modules and units described above can refer to the corresponding processes in the aforementioned method embodiments, and for the sake of brevity, they will not be further described here.

[0147] In addition, the functional units in various embodiments of the present invention may be physically independent of each other, or two or more functional units may be integrated together, or all functional units may be integrated into a single processing unit. The above-mentioned integrated functional units may be implemented in the form of hardware, software, or firmware.

[0148] Those skilled in the art will understand that if the integrated functional unit is implemented in the form of software and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention can essentially or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, which includes a number of instructions for enabling a computing device (such as a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention when running the instructions. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0149] Alternatively, all or part of the steps of implementing the aforementioned method embodiments may be accomplished by hardware associated with program instructions (such as a computing device such as a personal computer, a server, or a network device), and the program instructions may be stored in a computer-readable storage medium. When the program instructions are executed by a processor of a computing device, the computing device executes all or part of the steps of the method described in each embodiment of the present invention.

[0150] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that within the spirit and principles of the present invention, they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some or all of the technical features therein. However, these modifications or replacements do not deviate from the scope of protection of the present invention.

Claims

1. A method for calculating cerebral hypoperfusion areas based on ASL technology, characterized in that: The method comprises: Acquire cerebral blood flow images in the region of interest; segmenting the cerebral blood flow image based on a symmetry axis of the cerebral blood flow image to obtain a first cerebral blood flow image and a second cerebral blood flow image; flipping the first cerebral blood flow image to obtain a reference blood flow image corresponding to the second cerebral blood flow image; acquiring a low perfusion area in the second cerebral blood flow image based on the reference blood flow image and the second cerebral blood flow image; Before flipping the first cerebral blood flow image to obtain a reference blood flow image corresponding to the second cerebral blood flow image, the method further includes: Acquire a first brain structure image and an ASL image corresponding to the position of the cerebral blood flow image, wherein the first brain structure image includes cerebrospinal fluid; perform registration using the ASL image as a fixed image and the first brain structure image as a floating image to acquire a second brain structure image corresponding to the first brain structure image; The second brain structure image is input into a trained neural network model, and the symmetry axis of the cerebral blood flow image is output using the neural network model, where the neural network model is trained with a brain structure image including cerebrospinal fluid and a corresponding symmetry axis marked image; or, a cerebrospinal fluid image is segmented from the second brain structure image, and the symmetry axis of the cerebral blood flow image is obtained based on the cerebrospinal fluid image.

2. The method according to claim 1, characterized in that Before flipping the first cerebral blood flow image to obtain a reference blood flow image corresponding to the second cerebral blood flow image, the method further includes: Acquire a first brain structure three-dimensional image and an ASL three-dimensional image of a region of interest, wherein the first brain structure three-dimensional image includes a plurality of slices, and each slice includes cerebrospinal fluid; performing registration using the ASL three-dimensional image as a fixed image and the first three-dimensional image of the brain structure as a floating image to obtain a second three-dimensional image of the brain structure; Inputting the second three-dimensional image of the brain structure into a trained neural network model, and using the neural network model to output the symmetry axis of the cerebral blood flow image, wherein the neural network model is trained using the three-dimensional image of the brain structure including cerebrospinal fluid and the corresponding symmetry axis labeled image; or A cerebrospinal fluid image is segmented from a slice having the largest cerebrospinal fluid area in the second three-dimensional image of the brain structure, and a symmetry axis of the cerebral blood flow image is acquired based on the cerebrospinal fluid image.

3. The method according to claim 1 or 2, characterized in that Acquiring a symmetry axis of the cerebral blood flow image based on the cerebrospinal fluid image includes: inverting the cerebrospinal fluid region in the cerebrospinal fluid image to obtain a gap region between the cerebrospinal fluid region and the edge of the frame; Identify a first gap region located above the cerebrospinal fluid image and a second gap region located below the cerebrospinal fluid image, obtain a first coordinate point corresponding to a maximum abscissa value of the first gap region, and obtain a second coordinate point corresponding to a minimum abscissa value of the second gap region; Calculating an offset angle based on a straight line passing through the first coordinate point and the second coordinate point; An image symmetry axis is determined using the centroid of the cerebrospinal fluid image and the offset angle.

4. The method according to claim 3, characterized in that Before inverting the cerebrospinal fluid region in the cerebrospinal fluid image to obtain a gap region between the cerebrospinal fluid region and the edge of the frame, the method further includes: Find four vertices of the cerebrospinal fluid in the cerebrospinal fluid image, and select the maximum abscissa value of the two upper vertices, the minimum abscissa value of the two lower vertices, the maximum ordinate value of the two left vertices, and the minimum ordinate value of the two right vertices; The cerebrospinal fluid region is constructed using the maximum value of the abscissa, the minimum value of the abscissa, the maximum value of the ordinate, and the minimum value of the ordinate.

5. The method according to claim 1, wherein The flipping of the first cerebral blood flow image to obtain a reference blood flow image corresponding to the second cerebral blood flow image includes: acquiring a brain structure image corresponding to the cerebral blood flow image; Based on the symmetry axis of the cerebral blood flow image, acquiring a third brain structure image corresponding to the first cerebral blood flow image and a fourth brain structure image corresponding to the second cerebral blood flow image; flipping the third brain structure image, taking the fourth brain structure image as a fixed image and the flipped third brain structure image as a floating image, and obtaining a deformation field of the flipped third brain structure image relative to the fourth brain structure image; Based on the deformation field, the flipped first cerebral blood flow image is adjusted to obtain a reference blood flow image.

6. The method according to claim 5, characterized in that The acquiring, based on the reference blood flow image and the second cerebral blood flow image, a low perfusion area in the second cerebral blood flow image comprises: Calculating the cerebral blood flow ratio of corresponding pixels of the reference blood flow image and the second cerebral blood flow image; The areas with cerebral blood flow ratio less than the set threshold were marked and morphologically processed to obtain low perfusion areas.

7. A device for calculating cerebral hypoperfusion areas based on ASL technology, characterized in that: The device comprises: A cerebral blood flow image acquisition module is used to acquire cerebral blood flow images of the region of interest; a segmentation module, configured to segment the cerebral blood flow image based on a symmetry axis of the cerebral blood flow image to obtain a first cerebral blood flow image and a second cerebral blood flow image; a reference blood flow image acquisition module, configured to flip the first cerebral blood flow image to acquire a reference blood flow image corresponding to the second cerebral blood flow image; a low perfusion area acquisition module, configured to acquire a low perfusion area in the second cerebral blood flow image based on the reference blood flow image and the second cerebral blood flow image; A symmetry axis acquisition module is configured to acquire a first brain structure image and an ASL image corresponding to the position of the cerebral blood flow image, wherein the first brain structure image includes cerebrospinal fluid; and to perform registration using the ASL image as a fixed image and the first brain structure image as a floating image to acquire a second brain structure image corresponding to the first brain structure image; The second brain structure image is input into a trained neural network model, and the neural network model is used to output the symmetry axis of the cerebral blood flow image. The neural network model is trained with a brain structure image including cerebrospinal fluid and a corresponding symmetry axis marked image; or, the cerebrospinal fluid image is segmented from the second brain structure image, and the symmetry axis of the cerebral blood flow image is obtained based on the cerebrospinal fluid image.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium is used to store program code, and the program code is used to execute the method for calculating the cerebral hypoperfusion area based on the ASL technology according to any one of claims 1 to 6.

9. A computing device, characterized in that The computing device includes a processor and a memory: The memory is used to store program code and transmit the program code to the processor; The processor is configured to execute the method for calculating cerebral hypoperfusion areas based on ASL technology according to any one of claims 1 to 6 according to the instructions in the program code.

Images