Cerebrospinal fluid automatic segmentation and quantification method and system based on artificial intelligence

Abstract

The invention discloses a cerebrospinal fluid automatic segmentation and quantification method and system based on artificial intelligence, and belongs to the technical field of medical image processing, and the method comprises the steps: obtaining image data, carrying out the preprocessing of the image data, and obtaining a preprocessing data set; constructing an image three-dimensional reconstruction model, and processing the preprocessed data set based on the image three-dimensional reconstruction model to obtain a three-dimensional reconstruction data set; constructing a cerebrospinal fluid automatic segmentation model, and inputting the three-dimensional reconstruction data set into the cerebrospinal fluid automatic segmentation model for segmentation to obtain a cerebrospinal fluid segmentation result; original voxel pitch information is obtained, the cerebrospinal fluid segmentation result and the original voxel pitch information are calculated, and a quantitative result is obtained. The technical effect is remarkable, full-process automation from data reading to automatic calculation of the volume of the cerebrospinal fluid can be achieved, and the efficiency and accuracy of cerebrospinal fluid analysis are greatly improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical image processing, and in particular relates to an artificial intelligence-based automatic segmentation and quantification method and system for cerebrospinal fluid. Background Art

[0002] With the development of medical imaging technology, the diagnosis of brain diseases increasingly relies on image analysis. Among them, cerebrospinal fluid (CSF), as an important component of the brain, plays a protective and nutritional role around the brain and spinal cord, and is essential for maintaining a stable brain environment. Under normal circumstances, the production and absorption of cerebrospinal fluid are balanced. However, when there is an imbalance in the production and absorption of cerebrospinal fluid, such as when the cerebrospinal fluid exceeds the normal threshold, it will lead to increased intracranial pressure, affect brain blood circulation, cause brain tissue compression, and even induce serious complications such as brain herniation. On the contrary, an abnormal reduction in cerebrospinal fluid will also affect the protection and nutrition of the brain, and have a negative impact on the patient's health.

[0003] Quantitative analysis of cerebrospinal fluid, that is, the precise measurement of the volume or pressure of cerebrospinal fluid, is extremely important for diagnosing brain diseases (such as hydrocephalus, cerebral hemorrhage, etc.), evaluating the effect of treatment, and judging the prognosis of the disease. Accurate monitoring of the amount of cerebrospinal fluid can not only help doctors identify the development of the disease in a timely manner, but also is an important basis for evaluating the effect of treatment and predicting the prognosis of the disease. The reduction in the amount of effusion is often regarded as a sign of good disease prognosis and an important reference for judging the patient's prognosis and formulating subsequent treatment plans.

[0004] Traditional quantitative analysis methods of cerebrospinal fluid mainly rely on manual operations by radiologists or technicians, including selecting specific anatomical structures as reference points on brain images, and then manually measuring morphological indicators such as the distance and depth between reference points to semi-quantitatively evaluate the amount and severity of effusion. This method is not only time-consuming, but the results are also affected by the operator's skill level and subjective judgment, resulting in inconsistency and inaccuracy in the analysis, and it also takes a lot of time. The main pain points of traditional quantitative analysis methods are as follows:

[0005] 1. Limitations of manual segmentation: It relies on doctors or technicians to manually identify and demarcate the cerebrospinal fluid area on the image, which is not only inefficient, but also has poor accuracy and repeatability of the results and is easily affected by personal experience and judgment.

[0006] 2. Impact of image quality: Traditional methods are highly dependent on image quality. Image noise, insufficient resolution or low contrast will affect the accuracy of segmentation and quantification.

[0007] 3. Lack of standardized processes: Different institutions may use different techniques and standards for CSF segmentation and quantification, which makes it difficult to compare results across institutions and limits data sharing and comprehensive analysis.

[0008] Therefore, there is an urgent need for a fast, automated, highly precise and standardized method to conduct quantitative analysis of cerebrospinal fluid to significantly improve the efficiency and accuracy of the analysis. Summary of the invention

[0009] In order to solve the above technical problems, the present invention proposes an artificial intelligence-based cerebrospinal fluid automatic segmentation and quantification method and system to solve the problems existing in the above-mentioned prior art.

[0010] To achieve the above objectives, the present invention provides an artificial intelligence-based automatic segmentation and quantification method for cerebrospinal fluid, comprising:

[0011] Acquire brain image data, wherein the image data is CT or MRI data, and preprocess the image data to obtain a preprocessed data set;

[0012] Constructing a three-dimensional image reconstruction model, and processing the preprocessed data set based on the three-dimensional image reconstruction model to obtain a three-dimensional reconstruction data set;

[0013] Constructing a cerebrospinal fluid automatic segmentation model, inputting the three-dimensional reconstructed data set into the cerebrospinal fluid automatic segmentation model for segmentation, and obtaining a cerebrospinal fluid segmentation result;

[0014] The original voxel spacing information is obtained, and the cerebrospinal fluid segmentation result and the original voxel spacing information are calculated to obtain a quantitative result.

[0015] Preferably, the process of preprocessing the brain image data to obtain a preprocessed data set includes:

[0016] performing contrast processing on the brain image data by a contrast enhancement method to obtain a contrast enhanced image;

[0017] Removing noise from the contrast enhanced image based on smoothing, blurring and filtering techniques to obtain a noise-removed image;

[0018] Adjusting the pixels of the noise-removed image to a range of 0-1 to generate a pixel-adjusted image;

[0019] The cerebrospinal fluid region in the pixel-adjusted image is labeled to obtain the preprocessed data set.

[0020] Preferably, the process of obtaining the three-dimensional reconstruction data set includes:

[0021] Constructing the three-dimensional reconstruction model of the image based on a convolutional neural network;

[0022] The preprocessed data set is input into the image three-dimensional reconstruction model for three-dimensional reconstruction to obtain the three-dimensional reconstruction data set.

[0023] Preferably, the process of obtaining the cerebrospinal fluid segmentation result includes:

[0024] Build a cerebrospinal fluid automatic segmentation model based on the U-Net network model;

[0025] The three-dimensional reconstructed data set is input into the cerebrospinal fluid automatic segmentation model for calculation to obtain the cerebrospinal fluid segmentation result.

[0026] Preferably, the process of constructing a cerebrospinal fluid automatic segmentation model based on the U-Net network model includes:

[0027] Constructing a U-Net network model, and training the U-Net network model based on a training set to obtain a training model;

[0028] The training model is verified based on the verification set, and the performance is evaluated based on the training model through the test set and then output to obtain the cerebrospinal fluid automatic segmentation model,

[0029] Preferably, the process of obtaining the quantitative result comprises:

[0030] Calculating the volume of the cerebrospinal fluid region in the cerebrospinal fluid segmentation result to obtain a volume calculation result;

[0031] The original voxel spacing information is acquired, and calculation is performed based on the original voxel spacing information and the volume calculation result to obtain the quantitative result.

[0032] Preferably, the expression for obtaining the volume calculation result is: cerebrospinal fluid volume = cerebrospinal fluid prime number × single voxel volume.

[0033] Preferably, the expression of the single voxel volume is: voxel volume=voxel spacing X×voxel spacing Y×voxel spacing Z.

[0034] To achieve the above objectives, the present invention also provides an artificial intelligence-based cerebrospinal fluid automatic segmentation and quantification system, comprising:

[0035] An image acquisition module is used to acquire image data, preprocess the image data, and obtain a preprocessed data set;

[0036] An image reconstruction module, used for constructing a three-dimensional image reconstruction model, and processing the preprocessed data set based on the three-dimensional image reconstruction model to obtain a three-dimensional reconstruction data set;

[0037] A cerebrospinal fluid segmentation module is used to construct a cerebrospinal fluid automatic segmentation model, input the three-dimensional reconstructed data set into the cerebrospinal fluid automatic segmentation model for segmentation, and obtain a cerebrospinal fluid segmentation result;

[0038] A cerebrospinal fluid quantitative module is used to obtain original voxel spacing information, calculate the cerebrospinal fluid segmentation result and the original voxel spacing information, and obtain a quantitative result;

[0039] The result saving module is used to display and save the cerebrospinal fluid segmentation result and the quantitative result.

[0040] Compared with the prior art, the present invention has the following advantages and technical effects:

[0041] The technical effect of the present invention is remarkable, and it can realize the automation of the whole process from data reading to automatic calculation of cerebrospinal fluid volume, greatly improving the efficiency and accuracy of cerebrospinal fluid analysis. In addition, the method and system provide accurate data support, significantly saving doctors' film reading time and reducing their workload. Through the application of deep learning models, the present invention also has good adaptability and scalability, can cope with various complex medical image processing scenarios, and has broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0043] Figure 1 A schematic diagram of the system structure of an embodiment of the present invention;

[0044] Figure 2 A schematic diagram of a structure of a segmentation model according to an embodiment of the present invention;

[0045] Figure 3 A schematic diagram of a segmentation model according to an embodiment of the present invention;

[0046] Figure 4 This is a flow chart of cerebrospinal fluid quantification according to an embodiment of the present invention. DETAILED DESCRIPTION

[0047] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0048] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0049] Embodiment 1

[0050] like Figure 1-4 As shown, this embodiment provides an artificial intelligence-based cerebrospinal fluid automatic segmentation and quantification method, including:

[0051] S1, acquiring brain image data, and preprocessing the brain image data to obtain a preprocessed data set;

[0052] S2, constructing an image three-dimensional reconstruction model, and processing the preprocessed data set based on the image three-dimensional reconstruction model to obtain a three-dimensional reconstruction data set;

[0053] The specific implementation steps of S2 are: constructing a three-dimensional image reconstruction model to perform three-dimensional reconstruction on the read two-dimensional image;

[0054] S3, constructing a cerebrospinal fluid automatic segmentation model, inputting the three-dimensional reconstructed data set into the cerebrospinal fluid automatic segmentation model for segmentation, and obtaining a cerebrospinal fluid segmentation result;

[0055] The specific implementation steps of S3 are: constructing a cerebrospinal fluid automatic segmentation model, inputting the three-dimensional brain image into the cerebrospinal fluid automatic segmentation model, and realizing automatic recognition and segmentation of the cerebrospinal fluid;

[0056] S4, obtaining original voxel spacing information, calculating the cerebrospinal fluid segmentation result and the original voxel spacing information to obtain a quantitative result.

[0057] The specific implementation steps of S4 are: according to the automatic segmentation result of cerebrospinal fluid, using the distance information between the original voxels and the original voxels to calculate and generate quantitative results;

[0058] S5, displaying the cerebrospinal fluid segmentation results and quantitative data in various forms, and saving the segmentation results and quantitative information;

[0059] Further optimizing the scheme, preprocessing the brain image data, and obtaining the preprocessed data set includes:

[0060] Performing contrast processing on the brain image data by using a contrast enhancement method to obtain a contrast enhanced image, the specific implementation steps are: S11, using the contrast enhancement method to adjust the contrast of the image through histogram equalization;

[0061] Based on smoothing, blurring and filtering techniques, the noise of the contrast enhanced image is removed to obtain a noise-removed image; the specific implementation steps are: S12, using a noise reduction method, using smoothing, blurring and filtering techniques to remove unnecessary noise in the image;

[0062] The pixels of the noise-removed image are adjusted to be within the range of 0-1 to generate a pixel-adjusted image; the specific implementation steps are: S13, adjusting the pixels of the brain image to be within the range of 0-1, that is, normalization processing;

[0063] The cerebrospinal fluid region in the pixel-adjusted image is labeled to obtain the preprocessed data set.

[0064] Further optimizing the scheme, the process of obtaining the three-dimensional reconstruction data set includes:

[0065] Constructing the three-dimensional reconstruction model of the image based on a convolutional neural network;

[0066] The preprocessed data set is input into the image 3D reconstruction model to perform 3D reconstruction to obtain the 3D reconstruction data set, specifically:

[0067] S21, inputting the preprocessed brain image into a three-dimensional reconstruction module, performing three-dimensional reconstruction on the two-dimensional brain image to obtain a three-dimensional brain image;

[0068] Further optimizing the scheme, the process of obtaining the cerebrospinal fluid segmentation result includes:

[0069] Based on the U-Net network model, a cerebrospinal fluid automatic segmentation model is constructed; specifically:

[0070] The three-dimensional reconstructed data set is input into the cerebrospinal fluid automatic segmentation model for calculation to obtain the cerebrospinal fluid segmentation result.

[0071] To further optimize the solution, the process of building a cerebrospinal fluid automatic segmentation model based on the U-Net network model includes:

[0072] Constructing a U-Net network model, and training the U-Net network model based on a training set to obtain a training model;

[0073] The training model is verified based on the validation set, and the performance is evaluated based on the training model through the test set and then output to obtain the cerebrospinal fluid automatic segmentation model, specifically:

[0074] S31, inputting the three-dimensional brain image into the cerebrospinal fluid intelligent segmentation module to automatically identify and segment the cerebrospinal fluid;

[0075] S32, dividing the pre-labeled brain image set into a training set, a validation set, and a test set;

[0076] S33, performing image processing on the training set to expand the database of the training set, wherein the image processing method includes at least one of flipping, rotating, and adjusting contrast;

[0077] S34, using the training set and the verification set after image processing to train and verify the convolutional neural network to obtain a cerebrospinal fluid intelligent segmentation model, and using the test set to test and evaluate the effect of the recognition model.

[0078] Further optimizing the scheme, the process of obtaining the quantitative results includes:

[0079] Calculating the volume of the cerebrospinal fluid region in the cerebrospinal fluid segmentation result to obtain a volume calculation result;

[0080] The original voxel spacing information is acquired, and calculation is performed based on the original voxel spacing information and the volume calculation result to obtain the quantitative result.

[0081] Further optimizing the scheme, the expression for obtaining the volume calculation result is: cerebrospinal fluid volume = cerebrospinal fluid prime number × single voxel volume. The specific steps are:

[0082] S41, calculating the total volume of cerebrospinal fluid in the segmented area by acquiring the cerebrospinal fluid segmentation result and integrating the voxel spacing information;

[0083] S411, the cerebrospinal fluid calculation formula is: cerebrospinal fluid volume = cerebrospinal fluid prime number × single voxel volume;

[0084] Further optimizing the solution, the expression of the volume of a single voxel is: voxel volume = voxel spacing X × voxel spacing Y × voxel spacing Z, and the specific steps are:

[0085] S4111, the single voxel volume calculation formula is: single voxel volume = voxel spacing X×voxel spacing Y×voxel spacing Z;

[0086] Furthermore, the details of S5 are as follows:

[0087] S51, the segmentation results and quantitative data of cerebrospinal fluid are displayed in various forms, including but not limited to the following forms:

[0088] S511, 3D visualization model, through 3D reconstruction technology, the segmentation results of cerebrospinal fluid are presented as a 3D model, which can view the distribution of cerebrospinal fluid in the brain from various angles. This form intuitively displays the spatial structure of cerebrospinal fluid, which helps doctors understand the specific location of cerebrospinal fluid accumulation and its possible impact on surrounding tissues;

[0089] S512, slice display, the 3D reconstructed brain model is sliced ​​and displayed according to different planes (such as coronal plane, sagittal plane and cross section), and the segmentation result of cerebrospinal fluid is displayed on each slice, which is convenient for doctors to conduct detailed analysis from different perspectives;

[0090] S513, tabular report, summarizes quantitative information of CSF (such as total volume, CSF volume in each brain region, etc.) in a tabular form. The report can include detailed data such as the trend of CSF volume changes and comparison with normal values, providing quantitative reference information for doctors;

[0091] S514, standardized data format, uses the standard format of medical imaging data (such as DICOM) to save segmented cerebrospinal fluid images and quantitative information, ensuring data interoperability and long-term compatibility, and supporting viewing and analysis in different medical imaging systems.

[0092] Embodiment 2

[0093] This embodiment provides an artificial intelligence-based cerebrospinal fluid automatic segmentation and quantification system, including:

[0094] The image acquisition module is used to input the patient's brain image and pre-process the image through a standardized process to ensure that the image quality meets the requirements of subsequent analysis;

[0095] The image reconstruction module uses a deep learning algorithm to perform 3D reconstruction on the collected 2D images to construct a 3D model of the brain. The key to this module is the high-precision image reconstruction technology, which ensures the accuracy of the model and the details of the image;

[0096] The cerebrospinal fluid segmentation module uses an image segmentation network model to identify and segment the cerebrospinal fluid of the three-dimensional reconstructed brain model;

[0097] The cerebrospinal fluid quantification module calculates the volume of cerebrospinal fluid based on the cerebrospinal fluid segmentation results and the voxel spacing information of the original image. This module not only provides accurate calculation of the total volume of cerebrospinal fluid, but also provides regional volume analysis according to clinical needs, providing detailed data support for clinical diagnosis.

[0098] The result storage module is based on the segmentation and quantification results of cerebrospinal fluid. The module displays the results in multiple ways and stores relevant information.

[0099] In a further optimized solution, the medical image is a brain image;

[0100] To further optimize the scheme, the specific method of image preprocessing is as follows:

[0101] The preprocessing method includes contrast enhancement, noise reduction, normalization, etc.

[0102] The contrast of the image is also adjusted through histogram equalization based on the contrast enhancement method;

[0103] Based on the noise reduction method, smoothing, blurring and filtering techniques are used to remove unnecessary noise from the image;

[0104] Based on adjusting the brain image pixels to a range of 0-1;

[0105] To further optimize the solution, the cerebrospinal fluid area in the preprocessed brain image is annotated to complete the preprocessing

[0106] After further optimizing the scheme, the preprocessed brain image samples are then input into the first convolutional neural network for 3D reconstruction;

[0107] To further optimize the scheme, the three-dimensional reconstructed brain images were input into the convolutional neural network, and the model was trained to achieve the localization and automatic segmentation of cerebrospinal fluid.

[0108] According to a further optimization scheme, the image segmentation network model used for image segmentation is trained using a U-Net network model, which uses a large amount of preprocessed training data to learn the characteristics of cerebrospinal fluid to achieve high-precision segmentation of the cerebrospinal fluid area.

[0109] Based on the 3D reconstructed brain data, it is divided into training set, validation set, and test set;

[0110] Based on the training set and the validation set, the U-Net network model is trained and validated to obtain a cerebrospinal fluid intelligent segmentation model, and the segmentation model is tested and evaluated using the test set;

[0111] The result saving module is used to display and save the cerebrospinal fluid segmentation results in a variety of ways;

[0112] To further optimize the scheme, the construction method of the cerebrospinal fluid quantitative module is:

[0113] Based on the cerebrospinal fluid segmentation results, the volume of the cerebrospinal fluid area is calculated;

[0114] The quantitative results are obtained based on the distance information between the cerebrospinal fluid segmentation results and the original voxels.

[0115] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. An artificial intelligence-based automatic segmentation and quantification method for cerebrospinal fluid, characterized in that: The following steps are involved: Acquire image data, wherein the image data is brain CT or MRI data, and preprocess the image data to obtain a preprocessed data set; Constructing a three-dimensional image reconstruction model, and processing the preprocessed data set based on the three-dimensional image reconstruction model to obtain a three-dimensional reconstruction data set; Constructing a cerebrospinal fluid automatic segmentation model, inputting the three-dimensional reconstructed data set into the cerebrospinal fluid automatic segmentation model for segmentation, and obtaining a cerebrospinal fluid segmentation result; The original voxel spacing information is obtained, and the cerebrospinal fluid segmentation result and the original voxel spacing information are calculated to obtain a quantitative result.

2. The method for automatic segmentation and quantification of cerebrospinal fluid based on artificial intelligence according to claim 1, characterized in that: The process of preprocessing the image data to obtain a preprocessed data set includes: Performing contrast processing on the image data by a contrast enhancement method to obtain a contrast enhanced image; Removing noise from the contrast enhanced image based on smoothing, blurring and filtering techniques to obtain a noise-removed image; Adjusting the pixels of the noise-removed image to a range of 0-1 to generate a pixel-adjusted image; The cerebrospinal fluid region in the pixel-adjusted image is labeled to obtain the preprocessed data set.

3. The method for automatic segmentation and quantification of cerebrospinal fluid based on artificial intelligence according to claim 1, characterized in that: The process of obtaining the three-dimensional reconstruction data set includes: Constructing the three-dimensional reconstruction model of the image based on a convolutional neural network; The preprocessed data set is input into the image three-dimensional reconstruction model for three-dimensional reconstruction to obtain the three-dimensional reconstruction data set.

4. The method for automatic segmentation and quantification of cerebrospinal fluid based on artificial intelligence according to claim 1, characterized in that: The process of obtaining the cerebrospinal fluid segmentation result comprises: Build a cerebrospinal fluid automatic segmentation model based on the U-Net network model; The three-dimensional reconstructed data set is input into the cerebrospinal fluid automatic segmentation model for calculation to obtain the cerebrospinal fluid segmentation result.

5. The method for automatic segmentation and quantification of cerebrospinal fluid based on artificial intelligence according to claim 4, characterized in that: The process of building a cerebrospinal fluid automatic segmentation model based on the U-Net network model includes: Constructing a U-Net network model, and training the U-Net network model based on a training set to obtain a training model; The training model is verified based on the verification set, and the performance is evaluated based on the training model through the test set and then output to obtain the cerebrospinal fluid automatic segmentation model.

6. The method for automatic segmentation and quantification of cerebrospinal fluid based on artificial intelligence according to claim 1, characterized in that: The process of obtaining quantitative results includes: Calculating the volume of the cerebrospinal fluid region in the cerebrospinal fluid segmentation result to obtain a volume calculation result; The original voxel spacing information is acquired, and calculation is performed based on the original voxel spacing information and the volume calculation result to obtain the quantitative result.

7. The method for automatic segmentation and quantification of cerebrospinal fluid based on artificial intelligence according to claim 1, characterized in that: The expression for obtaining the volume calculation result is: cerebrospinal fluid volume = cerebrospinal fluid prime number × single voxel volume.

8. The method for automatic segmentation and quantification of cerebrospinal fluid based on artificial intelligence according to claim 7, characterized in that: The expression of the volume of a single voxel is: volume of a single voxel=voxel spacing×voxel spacing×voxel spacing.

9. An artificial intelligence-based automatic segmentation and quantification system for cerebrospinal fluid, characterized in that: include: An image acquisition module, used to acquire brain CT or MRI image data, and preprocess the image data to obtain a preprocessed data set; An image reconstruction module, used for constructing a three-dimensional image reconstruction model, and processing the preprocessed data set based on the three-dimensional image reconstruction model to obtain a three-dimensional reconstruction data set; A cerebrospinal fluid segmentation module is used to construct a cerebrospinal fluid automatic segmentation model, input the three-dimensional reconstructed data set into the cerebrospinal fluid automatic segmentation model for segmentation, and obtain a cerebrospinal fluid segmentation result; A cerebrospinal fluid quantitative module is used to obtain original voxel spacing information, calculate the cerebrospinal fluid segmentation result and the original voxel spacing information, and obtain a quantitative result; The result saving module is used to display and save the cerebrospinal fluid segmentation result and the quantitative result.

Images