Nuclear medicine imaging method and system

Abstract

The invention relates to the technical field of medical imaging, in particular to a nuclear medicine imaging method and system, and the method comprises the steps: patient preparation, tracer agent distribution detection, image enhancement, multi-target dynamic imaging, image quantitative analysis and data fusion, and result output and diagnosis suggestion. A preliminary image of tracer agent distribution in the body of a patient is obtained through a multi-modal imaging device, a high-resolution image is generated through a convolutional neural network noise suppression algorithm, a contrast ratio self-adaptive enhancement algorithm and a super-resolution reconstruction algorithm based on a generative adversarial network, and accurate dynamic analysis is carried out. In combination with a random forest algorithm and a time sequence analysis model, performing quantitative analysis on the radioactive intensity, the volume and the metabolic rate of the target spot, and generating personalized diagnosis suggestions; according to the invention, the precision and diagnosis effect of nuclear medicine imaging are greatly improved, and the method is suitable for early diagnosis, curative effect evaluation and disease progress monitoring of clinical diseases.

Description

Technical Field

[0001] The present invention relates to the field of medical imaging technology, and in particular to a nuclear medicine imaging method and system. Background Art

[0002] Nuclear medicine imaging technology plays a vital role in modern medicine and is widely used in early diagnosis of diseases, efficacy evaluation and disease progression monitoring. This technology generates imaging data of target tissues or organs by introducing radioactive isotope-labeled tracers and utilizing their distribution characteristics in the human body.

[0003] Although existing nuclear medicine imaging technology has made progress in image resolution and multimodal imaging, there are still some technical bottlenecks. First, when dealing with noise, contrast and resolution, traditional image enhancement algorithms often have difficulty in balancing the preservation of image details and the suppression of noise, resulting in imaging quality that is difficult to meet the needs of accurate diagnosis. Second, when dynamically monitoring and analyzing lesion areas, existing systems lack the ability to accurately track the dynamic changes of multiple targets, making it easy to miss subtle lesions or fail to accurately assess lesion progression. Summary of the invention

[0004] The invention provides a nuclear medicine imaging method and system.

[0005] A nuclear medicine imaging method comprising the following steps:

[0006] S1, patient preparation: according to the patient's specific condition and imaging requirements, select a suitable radioactive isotope-labeled tracer, and introduce it into the patient's body through intravenous injection or oral administration. After entering the human body, the tracer is distributed in the target tissue or organ according to the preset time;

[0007] S2, preliminary detection of tracer distribution: using a multimodal imaging device to detect the preliminary distribution of the tracer in the patient's body and obtain a preliminary image of the tracer distribution, the detection adopts a multimodal imaging technology combining PET (positron emission tomography) and SPECT (single photon emission computed tomography), and the preliminary image is preprocessed by a dual-modal fusion algorithm based on a Gaussian mixture model to obtain a preliminary three-dimensional image of the tracer distribution;

[0008] S3, image enhancement processing: the preliminary three-dimensional image obtained by detection is processed by an image enhancement algorithm based on a convolutional neural network, the processing includes using a noise suppression algorithm to filter out noise, using a contrast adaptive enhancement algorithm to improve contrast, and using a super-resolution reconstruction algorithm to improve image resolution, and using a trained multi-layer convolutional neural network model to process the image layer by layer to improve image details and clarity;

[0009] S4, multi-target dynamic imaging: Based on the preliminary image enhancement, combined with the time series analysis algorithm based on long short-term memory network, the distribution of radioactive tracers at multiple targets is dynamically tracked, the concentration changes of tracers at different targets are monitored in real time, and dynamic imaging sequences are generated to evaluate the lesion characteristics of multiple targets;

[0010] S5, image quantitative analysis and data fusion: for dynamic imaging sequences, a quantitative analysis model based on the random forest algorithm is used to fuse and analyze image data at different time points, calculate the radioactivity intensity, volume and metabolic rate of each target, generate a quantitative analysis report, and compare and verify the analysis results with the preliminary images based on the Mahalanobis distance;

[0011] S6, Result output and diagnostic recommendations: Based on the results of quantitative analysis, a diagnostic report is automatically generated, including the radioactivity concentration of each target, lesion size and development trend. Combined with the pathological characteristics and clinical data analysis based on Bayesian reasoning, personalized diagnostic recommendations are provided to help doctors develop treatment plans.

[0012] Optionally, the S1 specifically includes:

[0013] S11, Patient assessment: Conduct a comprehensive assessment of the patient, including medical history, allergy history, medication history, and current medical status, to ensure that the patient is suitable for receiving the radioisotope-labeled tracer;

[0014] S12, Tracer selection: Based on the patient's assessment results, select an appropriate radioactive isotope-labeled tracer. The selection process includes selecting a tracer with the best affinity and half-life based on the specificity of the target tissue or organ.

[0015] S13, dose calculation: calculate the injection dosage of the tracer;

[0016] S14, tracer injection: the tracer is introduced into the patient’s body either intravenously or orally;

[0017] S15, target tissue distribution time: After the tracer is injected, wait for a preset time to ensure that the tracer can be fully distributed to the target tissue or organ. The preset time is determined by previous clinical trials or literature data.

[0018] Optionally, the S2 specifically includes:

[0019] S21, imaging equipment configuration: selecting and configuring a multimodal imaging equipment to ensure that the equipment can perform PET and SPECT imaging simultaneously, wherein the equipment includes a detector for image acquisition and a computer system for data processing;

[0020] S22, preliminary image acquisition: using the configured imaging device, perform a preliminary scan on the patient to obtain a preliminary distribution image of the tracer in the body;

[0021] S23, image fusion: PET images and SPECT images are fused using a dual-modality fusion algorithm based on a Gaussian mixture model;

[0022] S24, 3D image reconstruction: using the fused image data, a voxel-based 3D reconstruction algorithm is used to generate a preliminary 3D image of the tracer distribution.

[0023] Optionally, the S3 specifically includes:

[0024] S31, noise filtering: using a noise suppression algorithm based on a convolutional neural network to process the preliminary three-dimensional image;

[0025] S32, contrast enhancement: using contrast adaptive enhancement algorithm, improve the contrast of the image through local histogram equalization technology, and enhance the detail display;

[0026] S33, resolution improvement: a super-resolution reconstruction algorithm based on a generative adversarial network is used to generate high-resolution images by learning the mapping relationship between low-resolution and high-resolution images;

[0027] S34, image detail enhancement: Use a pre-trained deep convolutional neural network to process the image layer by layer to enhance the details and clarity of the image.

[0028] Optionally, the S4 specifically includes:

[0029] S41, time series data acquisition: using an imaging device, imaging the patient multiple times at predetermined time intervals, and acquiring tracer distribution images at multiple time points;

[0030] S42, time series analysis: using a time series analysis algorithm based on long short-term memory networks to analyze the tracer distribution at different time points and predict the dynamic changes of the tracer in the target tissue;

[0031] S43, multi-target analysis: Based on the time series analysis results, dynamic analysis is performed on multiple targets to evaluate the tracer concentration changes of each target and generate dynamic imaging sequences of multiple targets.

[0032] Optionally, the S5 specifically includes:

[0033] S51, image data preprocessing: preprocessing the dynamic imaging sequence, including image registration and standardization operations;

[0034] S52, quantitative analysis model establishment: A quantitative analysis model was established based on the random forest algorithm to calculate the radioactivity intensity, volume and metabolic rate of the target;

[0035] S53, data fusion analysis: Data fusion analysis was performed using Mahalanobis distance to integrate image data at different time points to calculate the radioactivity intensity, volume, and metabolic rate of the target;

[0036] S54, quantitative analysis report generation: Generate a quantitative analysis report based on the calculation results and compare it with the preliminary image results to verify the accuracy and reliability of the analysis.

[0037] Optionally, the S6 specifically includes:

[0038] S61, diagnostic report generation: based on the quantitative analysis results, a diagnostic report including the radioactivity concentration of each target point, the size of the lesion and the development trend is automatically generated;

[0039] S62, pathological feature analysis: Based on the Bayesian inference model, the quantitative analysis results are integrated with the pathological feature data to establish the association between the target lesions and clinical features;

[0040] S63, diagnostic recommendation generation: Combine the diagnostic report and pathological feature analysis results to provide personalized diagnostic recommendations, including treatment plans and follow-up monitoring recommendations.

[0041] Optionally, the step S31 uses a noise suppression algorithm based on a convolutional neural network to process the preliminary three-dimensional image, specifically including:

[0042] Input layer: receives preliminary 3D image data;

[0043] Convolution layer: Use multiple three-dimensional convolution kernels (3D convolution) to perform convolution operations on the input three-dimensional image data to extract spatial and depth features;

[0044] Activation layer: Apply the ReLU activation function to the output of the convolutional layer;

[0045] Pooling layer: downsamples the feature map output by the activation layer to reduce data dimension and computational complexity;

[0046] Deconvolution layer: After pooling, deconvolution operation is applied to gradually reconstruct the spatial resolution of the image;

[0047] Output layer: Outputs the three-dimensional image after noise suppression processing.

[0048] A nuclear medicine imaging system, used to implement the above-mentioned nuclear medicine imaging method, comprises the following modules:

[0049] Patient preparation module: used to select and introduce radioactive isotope-labeled tracers according to the patient's condition and imaging requirements, and monitor the distribution of the tracers in the target tissue or organ;

[0050] Tracer distribution detection module: including a multimodal imaging device for detecting the preliminary distribution of the tracer in the patient's body and generating a preliminary three-dimensional image through a dual-modality fusion algorithm combining positron emission tomography and single photon emission computed tomography;

[0051] Image enhancement module: including a noise suppression algorithm based on a convolutional neural network, a contrast adaptive enhancement algorithm, and a super-resolution reconstruction algorithm based on a generative adversarial network, which are used to filter noise, enhance contrast, and improve resolution of preliminary three-dimensional images to generate high-quality enhanced images;

[0052] Multi-target dynamic imaging module: used to dynamically track the distribution of radioactive tracers at multiple targets and generate dynamic imaging sequences in combination with time series analysis algorithms;

[0053] Image quantitative analysis and data fusion module: including the quantitative analysis model and data fusion algorithm based on the random forest algorithm, which is used to perform quantitative analysis on dynamic imaging sequences, calculate the radioactivity intensity, volume and metabolic rate of the target, and generate quantitative analysis reports;

[0054] Diagnostic output module: used to generate diagnostic reports based on quantitative analysis results, and combine pathological characteristics and clinical data to provide personalized diagnostic suggestions to help doctors formulate treatment plans.

[0055] Beneficial effects of the present invention:

[0056] The present invention, through the coordinated work of the patient preparation module, the tracer distribution detection module and the image enhancement module, the system can effectively select the radioactive isotope-labeled tracer suitable for the patient, and use multimodal imaging technology to generate high-resolution preliminary three-dimensional images. In particular, the noise suppression algorithm based on the convolutional neural network and the super-resolution reconstruction algorithm based on the generative adversarial network significantly improve the clarity and detail of the image. This multi-level image enhancement processing can ensure the high quality of the imaging results and provide a solid foundation for accurate diagnosis.

[0057] The present invention combines the multi-target dynamic imaging module and the image quantitative analysis and data fusion module in the system with time series analysis and random forest algorithm, which can track the dynamic distribution of tracers at different targets in real time and conduct quantitative analysis of their radioactivity intensity, volume and metabolic rate. This function enables doctors to dynamically monitor the progression of the disease, especially in the evaluation of multi-target lesions, providing a more accurate diagnostic basis and facilitating early detection and treatment of the disease.

[0058] In this invention, the diagnostic output module generates a diagnostic report containing radioactive concentration, lesion size and development trend, and combines pathological characteristics and clinical data to provide personalized diagnostic suggestions. The system uses advanced machine learning models and big data analysis to provide doctors with treatment plan suggestions, ensure the accuracy of diagnosis and the effectiveness of treatment, and improve the overall quality of medical services and the treatment effect of patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0060] Figure 1 is a schematic diagram of an embodiment of the present invention;

[0061] Figure 2 Schematic diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0062] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0063] It should be noted that the references to "one embodiment", "an embodiment", "an exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0064] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0065] like Figure 1As shown, a nuclear medicine imaging method comprises the following steps:

[0066] S1, patient preparation: according to the patient's specific condition and imaging requirements, select the appropriate radioactive isotope-labeled tracer and introduce it into the patient's body through intravenous injection or oral administration. After entering the human body, the tracer is distributed in the target tissue or organ according to the preset time;

[0067] S2, preliminary detection of tracer distribution: using a multimodal imaging device to detect the preliminary distribution of the tracer in the patient's body and obtain a preliminary image of the tracer distribution. The detection adopts a multimodal imaging technology combining PET (positron emission tomography) and SPECT (single photon emission computed tomography). The preliminary image is preprocessed by a dual-modal fusion algorithm based on a Gaussian mixture model to obtain a preliminary three-dimensional image of the tracer distribution.

[0068] S3, image enhancement processing: the preliminary three-dimensional image obtained by detection is processed by an image enhancement algorithm based on a convolutional neural network. The processing includes using a noise suppression algorithm to filter out noise, using a contrast adaptive enhancement algorithm to improve contrast, and using a super-resolution reconstruction algorithm to improve image resolution. The trained multi-layer convolutional neural network model is used to process the image layer by layer to improve the details and clarity of the image.

[0069] S4, multi-target dynamic imaging: Based on the preliminary image enhancement, combined with the time series analysis algorithm based on long short-term memory network, the distribution of radioactive tracers at multiple targets is dynamically tracked, the concentration changes of tracers at different targets are monitored in real time, and dynamic imaging sequences are generated to evaluate the lesion characteristics of multiple targets;

[0070] S5, image quantitative analysis and data fusion: for dynamic imaging sequences, a quantitative analysis model based on the random forest algorithm is used to fuse and analyze image data at different time points, calculate the radioactivity intensity, volume and metabolic rate of each target, generate a quantitative analysis report, and compare and verify the analysis results with the preliminary images based on the Mahalanobis distance;

[0071] S6, Result output and diagnostic recommendations: Based on the results of quantitative analysis, a diagnostic report is automatically generated, including the radioactivity concentration of each target, lesion size and development trend. Combined with the pathological characteristics and clinical data analysis based on Bayesian reasoning, personalized diagnostic recommendations are provided to help doctors develop treatment plans.

[0072] S1 specifically includes:

[0073] S11, Patient assessment: Conduct a comprehensive assessment of the patient, including medical history, allergy history, medication history, and current medical status, to ensure that the patient is suitable for receiving the radioisotope-labeled tracer;

[0074] S12, Tracer selection: Based on the patient's assessment results, select an appropriate radioactive isotope-labeled tracer. The selection process includes selecting a tracer with the best affinity and half-life based on the specificity of the target tissue or organ.

[0075] S13, Dose calculation: Calculate the injection dosage of the tracer, calculated as:

[0076]

[0077] Where D is the tracer injection dose (unit: MBq), A is the absorption coefficient of the target tissue or organ, W is the patient's weight (unit: kg), C is the tracer concentration (unit: MBq / mL), t 1 / 2 is the half-life of the tracer (unit: hours);

[0078] S14, tracer injection: introduce the tracer into the patient's body by intravenous injection or oral administration, ensure the sterility of the injection process and monitor the patient's physiological response;

[0079] S15, target tissue distribution time: After the tracer is injected, wait for a preset time to ensure that the tracer can be fully distributed to the target tissue or organ. The preset time is determined by previous clinical trials or literature data.

[0080] S2 specifically includes:

[0081] S21, Imaging equipment configuration: Select and configure multimodal imaging equipment to ensure that the equipment can perform PET and SPECT imaging simultaneously. The equipment includes detectors for image acquisition and a computer system for data processing.

[0082] S22, preliminary image acquisition: using the configured imaging device, perform a preliminary scan on the patient to obtain a preliminary distribution image of the tracer in the body;

[0083] S23, Image fusion: A dual-modality fusion algorithm based on a Gaussian mixture model is used to fuse the PET image and the SPECT image. The fusion process includes the following calculations:

[0084]

[0085] Among them, x is the image data to be fused, θ is the model parameter set, π k is the weight of the kth Gaussian component, is the kth Gaussian component, μ k and Σ k are the mean vector and covariance matrix respectively;

[0086] The Gaussian mixture model is trained based on the expectation maximization algorithm. The steps are as follows:

[0087] (1) Parameter π of the Gaussian mixture model k , μ k and∑ k Initialize, obtain the initial cluster center through the k-means algorithm, and set the initial covariance matrix to the unit matrix;

[0088] (2) Calculate the probability that each sample belongs to each Gaussian component. The formula is as follows:

[0089]

[0090] (3) Update the parameters π of the Gaussian mixture model k , μ k and∑ k , the formula is as follows:

[0091]

[0092] Among them, γ(z ik ) is the sample x i The posterior probability of belonging to the kth Gaussian component, N k is the effective sample number of the kth Gaussian component, and N is the total number of samples;

[0093] S24, 3D image reconstruction: using the fused image data, a voxel-based 3D reconstruction algorithm is used to generate a preliminary 3D image of the tracer distribution, ensuring that the spatial resolution of the image meets the clinical diagnosis requirements;

[0094] A dual-modal fusion algorithm based on a Gaussian mixture model was used for the preliminary detection of tracer distribution, which improved the spatial resolution and accuracy of the tracer distribution image.

[0095] S3 specifically includes:

[0096] S31, noise filtering: The preliminary three-dimensional image is processed using a noise suppression algorithm based on a convolutional neural network. The algorithm achieves noise filtering through the following convolution operations:

[0097]

[0098] Among them, Y i,j,k,l is the value of the lth convolution kernel at position (i, j, k) in the output feature map, X i+m,j+n,k+p,c is the pixel value of the input image at channel c position (i+m,j+n,k+p), W m,n,p,c,l is the weight of the lth convolution kernel at position (m,n,p) and channel c, b l is the bias term of the lth convolution kernel, M×N×P is the size of the convolution kernel, and C is the number of channels of the input image;

[0099] S32, contrast enhancement: Use the contrast adaptive enhancement algorithm to improve the contrast of the image and enhance the detail display through the local histogram equalization technology. The specific steps are as follows:

[0100] Input image: Each slice of the preliminary 3D image or the 2D image is taken as input;

[0101] Block processing: Each two-dimensional image is processed in blocks and the local histogram of each block is calculated;

[0102] Contrast limitation and equalization: Perform local histogram equalization according to the preset contrast limitation threshold;

[0103] Interpolation reconstruction: reconstruct all processed image blocks into a complete two-dimensional image;

[0104] Image output: Output the image after adaptive enhancement processing, its contrast is effectively enhanced and the details are clearer;

[0105] S33, resolution improvement: A super-resolution reconstruction algorithm based on a generative adversarial network is used to generate a high-resolution image by learning the mapping relationship between low-resolution and high-resolution images. The specific steps are as follows:

[0106] (1) The basic principle of generative adversarial networks: Generative adversarial networks consist of two main parts: the generator and the discriminator. In super-resolution reconstruction, the purpose of the generative adversarial network is to generate realistic high-resolution images, so that the generated images are difficult to distinguish from the real high-resolution images under the detection of the discriminator;

[0107] Generator: Input a low-resolution image and generate the corresponding high-resolution image through learning;

[0108] Discriminator: distinguishes whether the input image is a real high-resolution image or a fake high-resolution image generated by the generator;

[0109] The training process of the generative adversarial network is a game between the generator and the discriminator: the generator continuously generates more realistic high-resolution images, while the discriminator continuously improves its ability to distinguish between true and false;

[0110] (2) Construction of super-resolution generative adversarial network

[0111] a. Generator: The task of the generator is to convert the input low-resolution image into a high-resolution image. The generator consists of the following modules:

[0112] Convolutional layer: used to extract local features of the image;

[0113] Residual block: Introduces residual connections between convolutional layers to help the network better capture high-frequency details and accelerate convergence;

[0114] Upsampling layer: enlarge the low-resolution feature map to a high-resolution size through transposed convolution or sub-pixel convolution;

[0115] Output convolution layer: generates the final high-resolution image;

[0116] The goal of the generator is to make the generated high-resolution image as close to the real image as possible by minimizing the reconstruction loss (such as L1 or L2 loss) and the adversarial loss;

[0117] The training objective of the generator is expressed as:

[0118] P G =P content +λP adversarial ;

[0119] Among them, P content is the content loss, usually using pixel-level differences, such as MSE (mean square error), P adversarial is the adversarial loss, used to encourage the realism of generated images, and λ is the weight coefficient, balancing the content loss and the adversarial loss;

[0120] b. Discriminator: The discriminator is a binary classification network used to distinguish whether the input image is a real high-resolution image or an image generated by the generator. The structure of the discriminator is usually a multi-layer convolutional neural network, and finally a Sigmoid activation function is used to output the probability value;

[0121] The training objective of the discriminator is expressed as:

[0122]

[0123] Among them, x HR is a real high-resolution image, x LR is the corresponding low-resolution image, G(x LR ) is the high-resolution image generated by the generator;

[0124] (3) The steps of super-resolution reconstruction are as follows:

[0125] a. Low-resolution image input: Input low-resolution image x LR to the generator;

[0126] b. High-resolution image generation: The generator generates a high-resolution image G(x LR );

[0127] C. Discriminator judgment: The generated high-resolution image G(x LR ) and the real high-resolution image x HR The discriminator outputs the probability that it thinks the image is a real image.

[0128] d. Loss calculation:

[0129] For the generator: calculate the content loss between the generated image and the real image, and calculate the adversarial loss at the same time;

[0130] For the discriminator: calculate the loss based on its classification results for real images and generated images;

[0131] e. Network update: Update the generator G to minimize the generator's loss function, and update the discriminator D to maximize the discriminator's accuracy;

[0132] f. Generate high-resolution images: After training, the generator converts low-resolution images into high-resolution images;

[0133] S34, image detail enhancement: using a pre-trained deep convolutional neural network, the image is processed layer by layer to enhance the details and clarity of the image;

[0134] Through multiple image processing technology, the quality of imaging images is significantly improved, providing high-quality image data for subsequent analysis and diagnosis.

[0135] S4 specifically includes:

[0136] S41, time series data acquisition: using an imaging device, imaging the patient multiple times at predetermined time intervals, and acquiring tracer distribution images at multiple time points;

[0137] S42, time series analysis: A time series analysis algorithm based on long short-term memory network was used to analyze the tracer distribution at different time points and predict the dynamic changes of the tracer in the target tissue. The expression of the long short-term memory network model is as follows:

[0138] h t =σ(W h ·[h t-1 ,x t ]+b h );

[0139] Among them, h t is the hidden state at time step t, x t is the input at time step t, W h is the weight matrix, b h is the bias term, σ is the activation function;

[0140] S43, multi-target analysis: Based on the time series analysis results, dynamic analysis is performed on multiple targets to evaluate the tracer concentration changes of each target and generate dynamic imaging sequences of multiple targets;

[0141] Through time series analysis technology, accurate tracking of the dynamic distribution of tracers is achieved, improving the detection capability of multi-target lesion characteristics.

[0142] S5 specifically includes:

[0143] S51, image data preprocessing: preprocess the dynamic imaging sequence, including image registration and standardization operations, to ensure the spatial alignment and grayscale consistency of images at different time points;

[0144] S52, quantitative analysis model establishment: A quantitative analysis model was established based on the random forest algorithm to calculate the radioactivity intensity, volume and metabolic rate of the target. The expression of the random forest model is:

[0145]

[0146] Among them, f(x) is the prediction result, T i is the i-th decision tree, and N is the total number of decision trees;

[0147] S53, data fusion analysis: Data fusion analysis is performed by using the Mahalanobis distance to integrate the image data at different time points to calculate the radioactivity intensity, volume and metabolic rate of the target. The Mahalanobis distance calculation formula is:

[0148]

[0149] Where x and y are data point vectors, and S is the covariance matrix;

[0150] S54, quantitative analysis report generation: generating a quantitative analysis report based on the calculation results and comparing it with the preliminary image results to verify the accuracy and reliability of the analysis;

[0151] Through multi-dimensional quantitative analysis and data fusion, the accuracy of radioactivity intensity and lesion characteristic assessment is improved.

[0152] S6 specifically includes:

[0153] S61, diagnostic report generation: based on the quantitative analysis results, a diagnostic report including the radioactivity concentration of each target point, the size of the lesion and the development trend is automatically generated;

[0154] S62, Pathological feature analysis: Based on the Bayesian reasoning model, the quantitative analysis results are integrated with the pathological feature data to establish the association between the target lesions and clinical features. The model expression of Bayesian reasoning is:

[0155]

[0156] Where P(H|E) is the probability that hypothesis H is true given evidence E, P(E|H) is the probability that evidence E appears when hypothesis H is true, P(H) is the prior probability of hypothesis H, and P(E) is the marginal probability of evidence E.

[0157] S63, diagnostic recommendation generation: combining the diagnostic report and pathological feature analysis results to provide personalized diagnostic recommendations, including treatment options and follow-up monitoring recommendations;

[0158] By comprehensively analyzing quantitative results and clinical characteristics, it provides comprehensive and accurate diagnostic information and suggestions, helping doctors to develop more effective treatment plans.

[0159] S31 uses a noise suppression algorithm based on a convolutional neural network to process the preliminary three-dimensional image, including:

[0160] Input layer: Receives preliminary 3D image data. The size of the input image is W×H×D×C, where W is the image width, H is the height, D is the depth (i.e., the number of voxel layers of the 3D image), and C is the number of channels (usually 1 for grayscale images).

[0161] Convolution layer: Use multiple three-dimensional convolution kernels (3D convolution) to perform convolution operations on the input three-dimensional image data to extract spatial and depth features. The expression of the convolution operation is the same as in S31;

[0162] Activation layer: Apply the ReLU activation function to the output of the convolutional layer. The ReLU activation function expression is:

[0163] f(x)=max(0,x), the ReLU activation function suppresses the influence of noise while retaining effective features;

[0164] Pooling layer: downsamples the feature map output by the activation layer to reduce data dimension and computational complexity, using maximum pooling or average pooling;

[0165] Deconvolution layer: After pooling, in order to restore to the same size as the input image, a deconvolution operation (also called upsampling or transposed convolution) is applied to gradually reconstruct the spatial resolution of the image;

[0166] Output layer: Outputs the 3D image after noise suppression processing, ensuring that the processed image retains sufficient details while significantly reducing noise;

[0167] The preliminary three-dimensional image is processed using a noise suppression algorithm based on a convolutional neural network, which effectively removes the noise in the image, retains the details and structure of the original image, and improves the image quality.

[0168] like Figure 2As shown, a nuclear medicine imaging system is used to implement the above-mentioned nuclear medicine imaging method, including the following modules:

[0169] Patient preparation module: used to select and introduce radioactive isotope-labeled tracers according to the patient's specific condition and imaging needs, and monitor the distribution of tracers in target tissues or organs;

[0170] Tracer distribution detection module: including a multimodal imaging device for detecting the preliminary distribution of the tracer in the patient's body and generating a preliminary three-dimensional image through a dual-modality fusion algorithm combining positron emission tomography and single photon emission computed tomography;

[0171] Image enhancement module: including a noise suppression algorithm based on a convolutional neural network, a contrast adaptive enhancement algorithm, and a super-resolution reconstruction algorithm based on a generative adversarial network, which are used to filter noise, enhance contrast, and improve resolution of preliminary three-dimensional images to generate high-quality enhanced images;

[0172] Multi-target dynamic imaging module: used to dynamically track the distribution of radioactive tracers at multiple targets and generate dynamic imaging sequences in combination with time series analysis algorithms;

[0173] Image quantitative analysis and data fusion module: including the quantitative analysis model and data fusion algorithm based on the random forest algorithm, which is used to perform quantitative analysis on dynamic imaging sequences, calculate the radioactivity intensity, volume and metabolic rate of the target, and generate quantitative analysis reports;

[0174] Diagnostic output module: used to generate diagnostic reports based on quantitative analysis results, and combine pathological characteristics and clinical data to provide personalized diagnostic suggestions to help doctors formulate treatment plans.

[0175] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0176] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A nuclear medicine imaging method, characterized in that: The following steps are involved: S1, patient preparation: according to the patient's specific condition and imaging requirements, select a suitable radioactive isotope-labeled tracer, and introduce it into the patient's body through intravenous injection or oral administration. After entering the human body, the tracer is distributed in the target tissue or organ according to the preset time; S2, preliminary detection of tracer distribution: using a multimodal imaging device to detect the preliminary distribution of the tracer in the patient's body and obtain a preliminary image of the tracer distribution. The detection adopts a multimodal imaging technology combining PET and SPECT, and the preliminary image is preprocessed by a dual-modal fusion algorithm based on a Gaussian mixture model to obtain a preliminary three-dimensional image of the tracer distribution; S3, image enhancement processing: the preliminary three-dimensional image obtained by detection is processed by an image enhancement algorithm based on a convolutional neural network, the processing includes using a noise suppression algorithm to filter out noise, using a contrast adaptive enhancement algorithm to improve contrast, and using a super-resolution reconstruction algorithm to improve image resolution, and using a trained multi-layer convolutional neural network model to process the image layer by layer to improve image details and clarity; S4, multi-target dynamic imaging: Based on the preliminary image enhancement, combined with the time series analysis algorithm based on long short-term memory network, the distribution of radioactive tracers at multiple targets is dynamically tracked, the concentration changes of tracers at different targets are monitored in real time, and dynamic imaging sequences are generated to evaluate the lesion characteristics of multiple targets; S5, image quantitative analysis and data fusion: for dynamic imaging sequences, a quantitative analysis model based on the random forest algorithm is used to fuse and analyze image data at different time points, calculate the radioactivity intensity, volume and metabolic rate of each target, generate a quantitative analysis report, and compare and verify the analysis results with the preliminary images based on the Mahalanobis distance; S6, Result output and diagnostic recommendations: Based on the results of quantitative analysis, a diagnostic report is automatically generated, including the radioactivity concentration of each target, lesion size and development trend. Combined with the pathological characteristics and clinical data analysis based on Bayesian reasoning, personalized diagnostic recommendations are provided to help doctors develop treatment plans.

2. A nuclear medicine imaging method according to claim 1, characterized in that: The S1 specifically includes: S11, Patient assessment: Conduct a comprehensive assessment of the patient, including medical history, allergy history, medication history, and current condition; S12, Tracer selection: Select a radioactive isotope-labeled tracer based on the patient's assessment results; S13, dose calculation: calculate the injection dosage of the tracer; S14, tracer injection: the tracer is introduced into the patient’s body either intravenously or orally; S15, target tissue distribution time: After the tracer is injected, wait for a preset time to ensure that the tracer can be fully distributed to the target tissue or organ.

3. A nuclear medicine imaging method according to claim 2, characterized in that: The S2 specifically includes: S21, imaging equipment configuration: selecting and configuring a multimodal imaging equipment to ensure that the equipment can perform PET and SPECT imaging simultaneously, wherein the equipment includes a detector for image acquisition and a computer system for data processing; S22, preliminary image acquisition: using the configured imaging device, perform a preliminary scan on the patient to obtain a preliminary distribution image of the tracer in the body; S23, image fusion: PET images and SPECT images are fused using a dual-modality fusion algorithm based on a Gaussian mixture model; S24, 3D image reconstruction: using the fused image data, a voxel-based 3D reconstruction algorithm is used to generate a preliminary 3D image of the tracer distribution.

4. A nuclear medicine imaging method according to claim 3, characterized in that: The S3 specifically includes: S31, noise filtering: using a noise suppression algorithm based on a convolutional neural network to process the preliminary three-dimensional image; S32, contrast enhancement: using contrast adaptive enhancement algorithm, improve the contrast of the image through local histogram equalization technology, and enhance the detail display; S33, resolution improvement: a super-resolution reconstruction algorithm based on a generative adversarial network is used to generate a high-resolution image by learning the mapping relationship between low-resolution and high-resolution images; S34, image detail enhancement: Use a pre-trained deep convolutional neural network to process the image layer by layer to enhance the details and clarity of the image.

5. A nuclear medicine imaging method according to claim 4, characterized in that: The S4 specifically includes: S41, time series data acquisition: using an imaging device, imaging the patient multiple times at predetermined time intervals, and acquiring tracer distribution images at multiple time points; S42, time series analysis: using a time series analysis algorithm based on long short-term memory networks to analyze the tracer distribution at different time points and predict the dynamic changes of the tracer in the target tissue; S43, multi-target analysis: Based on the time series analysis results, dynamic analysis is performed on multiple targets to evaluate the tracer concentration changes of each target and generate dynamic imaging sequences of multiple targets.

6. A nuclear medicine imaging method according to claim 5, characterized in that: The S5 specifically includes: S51, image data preprocessing: preprocessing the dynamic imaging sequence, including image registration and standardization operations; S52, quantitative analysis model establishment: A quantitative analysis model was established based on the random forest algorithm to calculate the radioactivity intensity, volume and metabolic rate of the target; S53, data fusion analysis: Data fusion analysis was performed using Mahalanobis distance to integrate image data at different time points to calculate the radioactivity intensity, volume, and metabolic rate of the target; S54, quantitative analysis report generation: Generate a quantitative analysis report based on the calculation results and compare it with the preliminary image results to verify the accuracy and reliability of the analysis.

7. A nuclear medicine imaging method according to claim 6, characterized in that: The S6 specifically includes: S61, diagnostic report generation: based on the quantitative analysis results, a diagnostic report including the radioactivity concentration of each target point, the size of the lesion and the development trend is automatically generated; S62, pathological feature analysis: Based on the Bayesian inference model, the quantitative analysis results are integrated with the pathological feature data to establish the association between the target lesions and clinical features; S63, diagnostic recommendation generation: Combine the diagnostic report and pathological feature analysis results to provide personalized diagnostic recommendations, including treatment plans and follow-up monitoring recommendations.

8. A nuclear medicine imaging method according to claim 4, characterized in that: In S31, a noise suppression algorithm based on a convolutional neural network is used to process the preliminary three-dimensional image, which specifically includes: Input layer: receives preliminary 3D image data; Convolution layer: Use multiple 3D convolution kernels to perform convolution operations on the input 3D image data to extract spatial and depth features; Activation layer: Apply the ReLU activation function to the output of the convolutional layer; Pooling layer: downsamples the feature map output by the activation layer to reduce data dimension and computational complexity; Deconvolution layer: After pooling, deconvolution operation is applied to gradually reconstruct the spatial resolution of the image; Output layer: Outputs the three-dimensional image after noise suppression processing.

9. A nuclear medicine imaging system, used to implement a nuclear medicine imaging method according to any one of claims 1 to 8, characterized in that: Includes the following modules: Patient preparation module: used to select and introduce radioactive isotope-labeled tracers according to the patient's condition and imaging requirements, and monitor the distribution of the tracers in the target tissue or organ; Tracer distribution detection module: including a multimodal imaging device for detecting the preliminary distribution of the tracer in the patient's body and generating a preliminary three-dimensional image through a dual-modality fusion algorithm combining positron emission tomography and single photon emission computed tomography; Image enhancement module: including a noise suppression algorithm based on a convolutional neural network, a contrast adaptive enhancement algorithm, and a super-resolution reconstruction algorithm based on a generative adversarial network, which are used to filter noise, enhance contrast, and improve resolution of preliminary three-dimensional images to generate high-quality enhanced images; Multi-target dynamic imaging module: used to dynamically track the distribution of radioactive tracers at multiple targets and generate dynamic imaging sequences in combination with time series analysis algorithms; Image quantitative analysis and data fusion module: including the quantitative analysis model and data fusion algorithm based on the random forest algorithm, which is used to perform quantitative analysis on dynamic imaging sequences, calculate the radioactivity intensity, volume and metabolic rate of the target, and generate quantitative analysis reports; Diagnostic output module: used to generate diagnostic reports based on quantitative analysis results, and combine pathological characteristics and clinical data to provide personalized diagnostic suggestions to help doctors formulate treatment plans.

Images