Method and system for extracting parotid and prostate maximum standard uptake value analysis based on ct and mri images

Abstract

The application discloses a kind of based on CT and MRI image extraction parotid and prostate maximum standard value uptake analysis method, comprising: obtaining the maximum standard uptake value of parotid, the maximum standard uptake value of prostate;Maximum standard uptake value of prostate is compared with the maximum standard uptake value of parotid, whether the metabolic level of prostate is abnormal is judged;If the ratio of the maximum standard uptake value of prostate and the maximum standard uptake value of parotid is greater than preset value, risk prompt is obtained;If the ratio of the maximum standard uptake value of prostate and the maximum standard uptake value of parotid is less than preset value, state prompt is obtained.The application discloses a kind of based on CT and MRI image extraction parotid and prostate maximum standard value uptake analysis method, system;Constant spatial position in PET-CT image is used to construct anatomical navigation, and interference signal of bladder and brain is effectively eliminated.

Description

Technical Field

[0001] This invention relates to the field of image processing technology in smart healthcare, specifically to a method and system for extracting maximum standard values ​​of parotid gland and prostate gland from CT and MRI images. Background Technology

[0002] Current metabolic assessments for prostate cancer rely on doctors manually searching frame by frame for the maximum standard uptake value (SUVmax) of the parotid gland and prostate. This method is time-consuming (3-5 minutes per case), has significant subjective variability (easily affected by experience or fatigue), and is susceptible to interference from bladder artifacts / metallic foreign bodies, which severely restricts the efficiency of accurate diagnosis.

[0003] In existing technologies, PET / CT (Positron Emission Tomography / Computed Tomography) has become the "gold standard" imaging tool for modern oncology diagnosis, staging, treatment monitoring, and prognostic assessment. In the clinical application of PET / CT, the Standardized Uptake Value (SUV), especially the maximum Standardized Uptake Value (SUVmax), is the most critical semi-quantitative indicator for quantifying the uptake of radiotracers (such as 18F-FDG or 68Ga-PSMA) in lesions or specific tissues. In the routine whole-body PET / CT image interpretation process, clinicians typically focus on two specific anatomical regions and measure their SUVmax: Head – Parotid Gland: As an important physiological uptake tissue in the human body, the parotid gland's radioactive uptake level is usually relatively stable and representative. Clinically, the maximum radioactive uptake value (SUVmax) of the parotid gland is often used as an internal reference standard to reflect a patient's "overall metabolic level" or "basal metabolic level." By monitoring the SUV value of the parotid gland, doctors can correct for whole-body imaging errors caused by differences in injection dosage and metabolic time, which is crucial for assessing image quality and calibrating the metabolic extent of other lesions.

[0004] Pelvic cavity – Prostate: For prostate cancer patients, accurately measuring the highest metabolic point within the prostate gland is the core basis for determining the location and invasiveness of the primary tumor and guiding biopsy.

[0005] However, despite the rapid development of PET / CT hardware technology, the current methods for obtaining SUVmax values ​​for these two sites on clinical workstations still face significant technical bottlenecks and operational limitations. Existing technical solutions mainly fall into two categories, both of which have obvious limitations: The first type: traditional manual interactive measurement. This is currently the most commonly used method in clinical practice. Doctors need to browse through hundreds of cross-sectional slices layer by layer on the fused image using the mouse wheel.

[0006] For example, regarding the parotid gland: because the parotid gland is divided into left and right sides and is relatively small, doctors need to visually identify the outline of the parotid gland in the complex anatomical structure of the head and manually avoid interference from surrounding lymph nodes or blood vessels to delineate the region of interest (ROI).

[0007] For example, regarding the prostate: doctors need to locate the prostate within a narrow space deep in the pelvis. This method relies heavily on the doctor's anatomical experience and is tedious and time-consuming. For nuclear medicine departments that review a huge number of slides daily, repetitive manual measurements severely reduce work efficiency, and differences in the range drawn by different doctors can lead to fluctuations in the internal reference value (parotid gland SUV), thus affecting the accuracy of the calibration of systemic diseases.

[0008] The second category comprises segmentation methods based on atlas registration or pure deep learning. Some high-end workstations attempt to use standard brain atlases for registration or to use convolutional neural networks (CNNs) to segment organs directly on PET images. While these methods address automation issues to some extent, they still have the following limitations: High computational cost: 3D non-rigid registration or fully convolutional neural network (FCN / U-Net) inference usually requires expensive GPU hardware support, making it difficult to run in real time on ordinary PCs for reading videos.

[0009] Insufficient robustness: Pure PET deep learning models are highly susceptible to the type of tracer (such as PSMA) and the time after injection. When the physiological uptake patterns of the parotid gland or prostate change (such as after inflammation or radiation therapy), the model is prone to failure.

[0010] In summary, there is an urgent need to provide a localization and extraction method that can effectively and accurately locate the parotid gland while eliminating bladder interference to accurately locate the prostate. Summary of the Invention

[0011] This invention overcomes the shortcomings of the prior art and provides a method and system for extracting maximum standard value uptake analysis of the parotid gland and prostate based on CT and MRI images.

[0012] To achieve the above objectives, the technical solution adopted by this invention is as follows: a method for extracting maximum standard value uptake analysis of the parotid gland and prostate gland based on CT and MRI images, comprising the following methods: Obtain test data; Based on the detection data, the maximum standard uptake values ​​of the parotid gland and the prostate were extracted to obtain the maximum standard uptake values ​​of the parotid gland and the prostate. By comparing the maximum standard uptake value of the prostate gland with that of the parotid gland, it can be determined whether the metabolic level of the prostate gland is abnormal. When the ratio of the maximum standard uptake value of the prostate to the maximum standard uptake value of the parotid gland is greater than the preset value, the metabolic level of the prostate is greater than the average metabolic level of the whole body, and a risk warning is issued; when the ratio of the maximum standard uptake value of the prostate to the maximum standard uptake value of the parotid gland is less than the preset value, the metabolic level of the prostate is less than the average metabolic level of the whole body, and a status warning is issued. Based on the corresponding risk warnings or status alerts, the analysis results are output.

[0013] In a preferred embodiment of the present invention, the detection data includes paired axial pelvic region computed tomography (CT) image sequences and axial pelvic region magnetic resonance imaging (MRI) image sequences of the subject, as well as axial head region computed tomography (CT) image sequences.

[0014] In a preferred embodiment of the present invention, the acquisition of the maximum standard uptake value of the parotid gland, utilizing the high spatial resolution of CT images to calculate the maximum standard uptake value of the parotid gland in the head, includes the following steps: Obtain the head length and the width between CT images, calculate the head length / width between CT images, and obtain the number of CT image frames of the head; For each frame of the acquired head CT image, calculate the maximum standard access value of the CT image frame by frame, and store the maximum standard access value obtained in each frame in an array as the maximum standard access value array. Based on the obtained maximum standard uptake value array, if the patient's head contains metallic radioactive material, a repair algorithm is used to eliminate the interference of the maximum standard uptake value of metallic radioactive material on the maximum standard uptake value of the parotid gland. The local maximum value of the array after excluding the maximum standard uptake value of radioactive metals is taken as the target tissue; the maximum value of the target tissue is taken as the maximum standard uptake value of the parotid gland.

[0015] In a preferred embodiment of the present invention, the repair algorithm is based on Dynamic Statistical Thresholding and Linear Interpolation to repair abnormal mutations in SUVmax in CT images. The abnormal mutations in SUVmax include: high-density stripe artifacts; and artifact removal based on statistical outlier detection. Includes the following steps: Extract each frame of CT image of the head I z The maximum standard uptake value is used to construct a one-dimensional original signal sequence. Where N is the total number of slices, v i SUV in frame i maxThe maximum standard uptake value (SUVmax) is a semi-quantitative indicator calculated based on the highest radioactivity concentration of pixels within the region of interest (ROI) in a PET image. For any voxel i in a frame of an image, its standard ingestion value SUV i The calculation formula is as follows: ; Among them, C i W represents the radioactivity concentration measured in the i-th voxel (unit: kBq / mL); t The patient's weight (in kg) is used to standardize the effect of volume distribution caused by differences in patient size; D represents the total radioactive dose of the injected tracer (in MBq), and the dose of D has been corrected to the start of the scan according to the half-life formula; SUVs targeting the parotid gland or prostate area max That is, all voxels SUVs within the anatomical structure mask area. i The maximum value in the set of numbers: It eliminates differences in patient weight and injection dosage, making metabolic levels between different patients objectively comparable; When calculating the effective injection dose D in the SUV formula, the radionuclide decay law is used to correct the drug activity measured at the calibration time to the imaging start time. The correction formula includes: ; Where D represents the actual in vivo tracer activity (unit: MBq) corrected to the scan start time (ScanStartTime); D cal T represents the initial radioactivity measured on a dosimeter (unit: MBq); 1 / 2 Indicates the physical half-life of the radioactive isotope used (e.g., 18 The half-life of F is approximately 109.77 minutes; t scan Indicates the time point at which the PET scan begins; t cal Indicates the time point at which dose measurement (calibration) is performed; t scan -t cal This indicates the time interval from measurement to scanning; it is important to ensure that the units are consistent with the half-life units. Correction is used to eliminate dosage calculation errors caused by the time difference between injection and scanning, ensuring the SUV (Volume-to-Volume Scale) is accurate. max The physical reality; Adaptive threshold calculation includes: Calculate the median (Median) and standard deviation (STD) of the sequence, and define the artifact detection threshold (T). thresh : Where Median is the median function, sigma is the standard deviation of the sequence, and k is the sensitivity coefficient. The original signal sequence, The standard deviation of the original sequence; Anomaly detection and labeling: Traverse the sequence, when a certain frame v i >T thresh If the corresponding frame is affected by metal artifacts, its index i is marked as an anomaly. Interpolation repair: For frames v marked as anomalous i Using adjacent normal frames v prev and v next Perform linear interpolation repair: ; and They are and The corresponding frame index; Final output: ,in The sequence is the repaired sequence; after eliminating metal artifact interference, the accurate parotid gland metabolic peak index is obtained.

[0016] In a preferred embodiment of the present invention, obtaining the maximum standard uptake value of the prostate includes the following steps: The prostate mask obtained from MRI image segmentation is mapped to CT-PET space using corresponding bone alignment information to construct a CT-PET mask, thereby obtaining the prostate region. Each frame of prostate image is obtained based on the prostate region, and the maximum standard uptake value within the prostate mask is calculated frame by frame. The value obtained in each frame is stored in an array, namely the maximum standard uptake value array of the prostate. Based on the obtained array of maximum standard uptake values ​​of the prostate, and using a truncation algorithm to eliminate the interference of the maximum standard uptake value of the bladder on the maximum standard uptake value of the prostate, the local maximum value in the array of maximum standard uptake values ​​of the prostate without interference sources is the target tissue; the maximum value of the target tissue is taken as the maximum standard uptake value of the prostate.

[0017] In a preferred embodiment of the present invention, the truncation algorithm is a high-level bladder signal truncation algorithm based on gradient mutation detection (for the prostate), including: using an automatic truncation algorithm based on one-dimensional sequential gradient mutation; when the mask contains the bladder, the maximum standard uptake value of the first few frames will be approximately equal and abnormally large compared to the subsequent values, so the maximum standard uptake value of the mask containing the bladder is set to zero.

[0018] In a preferred embodiment of the present invention, an automatic truncation algorithm based on one-dimensional sequence gradient mutation is used to address the interference of bladder hypermetabolic signals that often occur in the head-side of prostate PET image sequences. According to physiological laws, bladder signals exhibit a "continuous high-value plateau" at the beginning of the sequence, while prostate signals are located behind it and have lower intensity.

[0019] In a preferred embodiment of the present invention, the truncation algorithm includes the following steps; Within the registered prostate mask region, the maximum SUV value of each frame is extracted along the Z-axis (from head to foot) to construct a one-dimensional metabolic signal sequence. Where N is the total number of slices; to eliminate the small fluctuations caused by imaging noise, the sequence is first processed by moving average smoothing: ; where w is the sliding window radius, which enables sequence construction and preprocessing.

[0020] The first-order forward difference of the smoothed sequence was calculated using forward difference gradient to quantify metabolic variability between adjacent slices. ;when Furthermore, a larger value indicates a drop from a "high metabolism zone" to a "low metabolism zone"; The algorithm identifies the "cliff-like drop point" at the beginning of the sequence to achieve high-level platform boundary detection. This "cliff-like drop point" must simultaneously satisfy two constraints: "gradient maximization" and "high-level platform verification." The constraints for maximizing the gradient include finding the peak gradient within the first M frames of the sequence. ; The constraints for high-level platform verification include: verification k peak Previous sequence mean Is it significantly higher than the subsequent sequence mean? To conform to the characteristic of "larger first few frames": ; in, The average signal value before the peak. The average signal value after the peak value. is the index of the peak point in the sequence, and N is the total number of frames in the entire sequence; This represents the preprocessed signal value of the j-th frame. When satisfied Where α is the preset drop factor, then determine k peak k represents the true physical boundary between the bladder and prostate.* ; Based on the locked boundary k * Construct a binary weighted mask vector W: ; Multiplying the original sequence by the mask yields the cleaned sequence S. clean And extract the final result from it: ,in, S represents the maximum standardized uptake value of the prostate region, S is the original signal sequence, and W is the weight matrix; this achieves interference zeroing and target extraction.

[0021] In a preferred embodiment of the present invention, a system for extracting maximum standard values ​​of the parotid gland and prostate based on CT and MRI images is provided, which is achieved by extracting and analyzing the maximum standard values ​​of the parotid gland and prostate based on CT and MRI images, including: The input module is used to acquire detection data, which includes paired axial pelvic region computed tomography (CT) image sequences and axial pelvic region magnetic resonance imaging (MRI) image sequences of the subject, as well as axial head region computed tomography (CT) image sequences. The diagnostic analysis module obtains the maximum standard uptake value of the parotid gland and the maximum standard uptake value of the prostate gland based on the test data, compares the maximum standard uptake value of the prostate gland with the maximum standard uptake value of the parotid gland, determines whether the metabolic level of the prostate gland is abnormal, and obtains corresponding risk warnings or status warnings. Output module; used for analyzing results.

[0022] This invention addresses the deficiencies in the technical background, and the beneficial technical effects of this invention are: This invention provides a topologically constrained SUVmax calculation method for key PET-CT sites based on adaptive spatial positioning. This invention utilizes the constant spatial location within PET-CT images to construct anatomical navigation, accurately removing interfering signals from the bladder and brain, and achieving precise extraction of the maximum standard uptake value (SUVmax) of the parotid gland and prostate. This invention provides a cross-modal collaborative algorithm for "separately extracting the maximum standard uptake values ​​of the parotid gland and prostate," transforming the complex problem of searching for systemic lesions into two local extremum optimization problems with strict anatomical boundary constraints, thus improving the convenience, accuracy, and stability of lesion identification. Attached Figure Description

[0023] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0024] Figure 1This is a schematic diagram of the process for extracting maximum standard value uptake analysis of the parotid gland and prostate based on CT and MRI images according to a preferred embodiment of the present invention; Figure 2 This is a comparison diagram of image recognition according to a preferred embodiment of the present invention; Figure 3 It is a segmented prostate mask image; Figure 4 These are the extracted SUV sequence values ​​from the head and prostate of the test subject (which have been desensitized). Detailed Implementation

[0025] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0026] It should be noted that if directional indicators (such as up, down, bottom, top, etc.) are involved in the embodiments of the present invention, these directional indicators are only used to explain the relative positional relationship and movement of the components in a specific posture. If the specific posture changes, the directional indicators will also change accordingly. The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, features defined with "first" and "second" may explicitly or implicitly include one or more of that feature. Unless otherwise explicitly specified and limited, the terms "set," "connected," and "linked" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal connection of two components. For those skilled in the art, the specific meaning of the above terms in the present invention can be understood according to the specific circumstances.

[0027] Example 1, as Figure 1 As shown, a method for extracting maximum standard uptake values ​​of the parotid gland and prostate gland based on CT and MRI images includes the following: Step one: Acquire the detection data; specifically, the detection data includes paired axial pelvic region computed tomography (CT) image sequences and axial pelvic region magnetic resonance imaging (MRI) image sequences, as well as axial head region computed tomography (CT) image sequences of the subject. Furthermore, the data format is the standard DICOM format.

[0028] Step 2: Extract the maximum standard uptake values ​​of the parotid gland and prostate gland based on the detection data to obtain the maximum standard uptake values ​​of the parotid gland and prostate gland.

[0029] The acquisition of the maximum standard uptake value of the parotid gland utilizes the high spatial resolution of CT images to calculate the maximum standard uptake value of the parotid gland in the head, and includes the following steps: The head length and the width between CT images are obtained (the detection data also includes: the head length, i.e., the total three-dimensional length of the head (total length of PET / CT Z-axis scan), and the width between CT images containing the head, i.e., the slice thickness of the CT tomographic image). The head length / the width between CT images is calculated to obtain the number of CT image frames of the head. For each frame of the acquired head CT image, calculate the maximum standard access value of the CT image frame by frame, and store the maximum standard access value obtained in each frame in an array as the maximum standard access value array. Based on the obtained maximum standard uptake value array, if the patient's head contains metallic radioactive substances, a repair algorithm is used to eliminate the interference of the maximum standard uptake value of the metallic radioactive substances on the maximum standard uptake value of the parotid gland.

[0030] The local maximum value of the array after excluding the maximum standard uptake value of radioactive metals is taken as the target tissue; the maximum value of the target tissue is taken as the maximum standard uptake value of the parotid gland.

[0031] Specifically, when a patient's head contains radioactive metallic substances, certain parameters in the maximum standard uptake value array will be abnormally prominent; the maximum standard uptake value of this part will be reset to zero.

[0032] Specifically, by predicting the overall uptake value of the head, the maximum standard uptake value of the parotid gland in the head is locked, and the overall metabolic level of the human body is estimated.

[0033] Furthermore, the repair algorithm employs dynamic statistical thresholding and linear interpolation to repair SUVmax aberrations in CT images. SUVmax aberrations include: high-density stripe artifacts; and metal artifact removal based on statistical outlier detection.

[0034] The determination of the prostate's maximum standard uptake value includes the following steps: The prostate mask image obtained from MRI image segmentation is mapped to CT-PET space using corresponding bone alignment information to construct a CT-PET mask image, thereby obtaining the prostate region. Each frame of prostate image is obtained based on the prostate region, and the maximum standard uptake value within the prostate mask is calculated frame by frame. The value obtained in each frame is stored in an array, namely the maximum standard uptake value array of the prostate. Based on the obtained array of maximum standard uptake values ​​of the prostate, and using a truncation algorithm to eliminate the interference of the maximum standard uptake value of the bladder on the maximum standard uptake value of the prostate, the local maximum value in the array of maximum standard uptake values ​​of the prostate without interference sources is the target tissue; the maximum value of the target tissue is taken as the maximum standard uptake value of the prostate.

[0035] Specifically, the truncation algorithm is a high-level bladder signal truncation algorithm based on gradient mutation detection (for the prostate), which includes: adopting an automatic truncation algorithm based on one-dimensional sequential gradient mutation; when the mask contains the bladder, the maximum standard uptake value of the first few frames will be roughly equal and abnormally large compared to the subsequent values, so the maximum standard uptake value of the mask containing the bladder is set to zero.

[0036] Furthermore, to address the interference from high metabolic signals in the bladder that often appear in the head-side of prostate PET image sequences, based on physiological principles, the bladder signal exhibits a "continuous high-value plateau" at the beginning of the sequence, while the prostate signal is located behind it and has a lower intensity, an automatic truncation algorithm based on one-dimensional sequence gradient mutation is adopted.

[0037] Step 3: Compare the maximum standard uptake value of the prostate gland with the maximum standard uptake value of the parotid gland to determine whether the metabolic level of the prostate gland is abnormal.

[0038] When the ratio of the maximum standard uptake value of the prostate to the maximum standard uptake value of the parotid gland is greater than the preset value, the metabolic level of the prostate is greater than the average metabolic level of the whole body, and a risk warning is issued; when the ratio of the maximum standard uptake value of the prostate to the maximum standard uptake value of the parotid gland is less than the preset value, the metabolic level of the prostate is less than the average metabolic level of the whole body, and a status warning is issued. Step 4: Output the analysis results based on the corresponding risk warnings or status prompts.

[0039] Working principle: The most critical data processing aspect of this invention is transforming a high-dimensional (three-dimensional voxel space) extraction problem into a very low-dimensional (one-dimensional signal) extremum problem through feature extraction. This not only greatly reduces computational complexity but also enhances the algorithm's robustness to single-slice detection errors by utilizing the "trend" of change throughout the sequence.

[0040] The method of this invention is computationally efficient and can be executed quickly on existing central processing units (CPUs). This invention requires no graphics processing unit (GPU) or any other dedicated hardware accelerator. Therefore, this invention can be easily deployed on any standard personal computer, medical imaging workstation, or server. The method of this invention directly processes medical images in the DICOM standard format and can seamlessly interface with any modern CT, MRI equipment, and hospital image archiving and communication systems (PACS), exhibiting excellent compatibility and applicability.

[0041] Example 2, as Figure 1 As shown, in Example 1, as Figure 1 As shown, a method for extracting maximum standard uptake values ​​of the parotid gland and prostate gland based on CT and MRI images includes the following: In step one, the preprocessing of the detection data includes the following features: feature extraction replaces traditional preprocessing: This invention bypasses the traditional method of manually searching for the maximum standard uptake value of the parotid gland and prostate gland frame by frame. Instead, it calculates the maximum standard uptake value of the relevant glands for each frame as the core preprocessing step for ratio calculation. Furthermore, spatial matching is performed using CT scans, and corresponding PET scans at the same location are read simultaneously. The maximum SUV of a single frame is calculated for the PET scans and stored in an array; CT scans are used for anatomical localization and metal artifact recognition.

[0042] In step two, the repair algorithm for obtaining the maximum standard uptake value of the parotid gland uses a combination of dynamic statistical thresholding and linear interpolation to repair abnormal mutations in SUVmax in CT images. Abnormal mutations in SUVmax include: high-density stripe artifacts; and metal artifact removal based on statistical outlier detection. Includes the following steps: Extract each frame of CT image of the head I z The maximum standard uptake value is used to construct a one-dimensional original signal sequence. Where N is the total number of slices (i.e., the total number of slices in the head CT image), v i SUV in frame i max The maximum standard uptake value (SUVmax) is a semi-quantitative indicator calculated based on the highest radioactivity concentration of pixels within the region of interest (ROI) in a PET image. For any voxel i in a frame of an image, its standard ingestion value SUV i The calculation formula is as follows: ; Among them, C iW represents the radioactivity concentration measured in the i-th voxel (unit: kBq / mL); t The patient's weight (in kg) is used to standardize the effect of volume distribution caused by differences in patient size; D represents the total radioactive dose of the injected tracer (in MBq), and the dose of D has been corrected to the start of the scan according to the half-life formula; SUVs targeting the parotid gland or prostate area max That is, all voxels SUVs within the anatomical structure mask area. i The maximum value in the set of numbers: This eliminates differences in patient weight and injection dosage, making metabolic levels objectively comparable between different patients. Here, max{⋅} represents the operation of finding the maximum value in the set. It is a binary mask obtained through image segmentation that marks the three-dimensional spatial region of the target organ; When calculating the effective injection dose D in the SUV formula, the radionuclide decay law is used to correct the drug activity measured at the calibration time to the imaging start time. The correction formula includes: ; Where D represents the total radioactive dose of the injected tracer, i.e., the actual in vivo tracer activity (unit: MBq) corrected to the scan start time (ScanStartTime); D cal T represents the initial radioactivity measured on a dosimeter (unit: MBq); 1 / 2 Indicates the physical half-life of the radioactive isotope used (e.g., 18 The half-life of F is approximately 109.77 minutes; t scan Indicates the time point at which the PET scan begins; t cal Indicates the time point at which dose measurement (calibration) is performed; t scan -t cal This indicates the time interval from measurement to scanning; it is important to ensure that the units are consistent with the half-life units. Through calibration, the dosage calculation error caused by the time difference between injection and scanning is eliminated, ensuring the accuracy of the calculated SUV. max The physical reality; Adaptive threshold calculation includes: Calculate the median (Median) and standard deviation (STD) of the sequence, and define the artifact detection threshold (T). thresh : Where Median is the median function, sigma is the standard deviation of the sequence, and k is the sensitivity coefficient (in this embodiment, k=3, which conforms to the 3sigma principle and covers 99.7% of normal physiological fluctuations). The original signal sequence, The standard deviation of the original sequence; Anomaly detection and labeling: Traverse the sequence, when a certain frame v i >T thresh If the frame is affected by metal artifacts, its index i is marked as an anomaly. Interpolation repair: For frames v marked as anomalous i Using its adjacent normal frames v prev and v next Perform linear interpolation repair. ; in, and They are and The corresponding frame index; Positional scale: indicates that target point i is at the previous point i prev and the next point i next The relative positions between them; the positional ratio applied to the numerical difference On, you need to get the v prev The amount added to the base result yields the linear interpolation result v at point i. i ′.

[0043] Final output: ,in The image sequence is the restored image; after eliminating metal artifacts, the accurate peak metabolic index of the parotid gland is obtained.

[0044] In step two, the truncation algorithm used to obtain the maximum standard uptake value of the prostate includes the following steps; Within the registered prostate mask region, the maximum SUV value of each frame is extracted along the Z-axis (from head to foot) to construct a one-dimensional metabolic signal sequence. Where N is the total number of slices (i.e., the total number of slices in the prostate image); to eliminate the small fluctuations caused by imaging noise, the sequence is first processed by moving average smoothing: ; Where w is the radius of the sliding window (w=1 in this embodiment), which enables sequence construction and preprocessing.

[0045] The first-order forward difference of the smoothed sequence was calculated using forward difference gradient to quantify metabolic variability between adjacent slices. ;when Furthermore, a larger value indicates a drop from a "high metabolism zone" to a "low metabolism zone"; The algorithm identifies the "cliff-like drop point" at the beginning of the sequence to achieve high-level platform boundary detection. This "cliff-like drop point" must simultaneously satisfy two constraints: "gradient maximization" and "high-level platform verification." The constraints for maximizing the gradient include: finding the peak gradient within the first M frames of the sequence (e.g., the first 1 / 3 of the sequence, covering the possible range of bladder occurrence). , The peak value is the gradient peak. The constraints for high-level platform verification include: verification k peak Previous sequence mean Is it significantly higher than k? peak Subsequent sequence mean To conform to the characteristic of "larger first few frames": ; in, The average signal value before the peak. The average signal value after the peak value. is the index of the peak point in the sequence, and N is the total number of frames in the entire sequence; This represents the preprocessed signal value of the j-th frame. When satisfied Where α is a preset drop factor (α=1.5 in this embodiment), then k is determined. peak k represents the true physical boundary between the bladder and prostate. * ; Based on the locked boundary k * Construct a binary weighted mask vector W: ; Multiplying the original sequence by the mask yields the cleaned sequence S. clean And extract the final result from it: ;in, S represents the maximum standardized uptake value of the prostate region, S is the original signal sequence, and W is the weight matrix; this achieves interference zeroing and target extraction.

[0046] Working principle: This invention addresses three core pain points in the prior art: low efficiency of manual delineation, high metabolic signals from the bladder easily leading to prostate location errors, and interference from radioactive materials in the brain and adjacent areas with parotid gland background measurements. This invention provides a topologically constrained PET-CT method for calculating the SUVmax of key sites based on adaptive spatial positioning. This invention can construct anatomical navigation using constant spatial locations in PET-CT images with minimal manual intervention, accurately removing interference signals from the bladder and brain, and achieving precise extraction of the maximum standard uptake value (SUVmax) of the parotid gland and prostate. This invention provides a cross-modal collaborative algorithm for "separately extracting the maximum standard uptake values ​​of the parotid gland and prostate," transforming the complex problem of searching for systemic lesions into two local extremum optimization problems with strict anatomical boundary constraints.

[0047] This invention completely solves the aforementioned problems by extracting the SUVmax of each frame and converting it into a one-dimensional sequence (dimensionality reduction), employing bladder avoidance logic (such as cluster analysis to remove high-metabolic bladder signals), and handling metal artifact robustness (quantile statistics or interpolation). The algorithm has extremely high computational efficiency (<5 seconds / case on a standard i7 CPU), the results are 100% reproducible, the error compared to manual annotation by doctors is less than 5%, and it does not require dedicated hardware such as GPUs. This solution represents a "paradigm shift" in metabolic indicator extraction, providing clinicians with an efficient and objective auxiliary tool that directly supports the early diagnosis, efficacy evaluation, and radiotherapy planning of prostate cancer, demonstrating significant clinical value and the necessity for patent protection.

[0048] This invention can be widely applied in prostate cancer diagnosis and treatment scenarios in urology, nuclear medicine, oncology, and radiotherapy departments. It uses the SUVprostate / SUVparotid ratio to assist in identifying abnormal prostate metabolism (e.g., a ratio >1 indicates high cancer risk), meeting the clinical need for "rapid, accurate, and standardized" metabolic assessment. Its low hardware dependency (it can run on a regular PC) and seamless integration with DICOM / PACS allow for easy deployment in hospitals at all levels (including primary healthcare institutions) without additional hardware investment. In the future, it can be integrated into existing medical imaging workstations or electronic medical record systems to optimize clinical workflows; it can also be combined with deep learning models to improve the predictive efficacy of the correlation between metabolic indicators and pathological results. With the global increase in prostate cancer incidence (approximately 1.4 million new cases annually), the market urgently needs efficient auxiliary diagnostic tools. This invention has the potential to become a "standard tool" for prostate cancer metabolic assessment, covering the entire process from screening to follow-up, improving the efficiency of medical resource utilization, and improving patient prognosis, demonstrating broad commercial application prospects and social value.

[0049] Example 3: A method for extracting maximum standard uptake values ​​of the parotid gland and prostate gland from CT and MRI images.

[0050] Step S1, Data Acquisition: Pelvic scans were performed on several prostate cancer patients, and image sequence groups were acquired for each patient. The image sequence groups included OCT image sequences and OMRIT2 image sequences.

[0051] The OCT sequence contains 496 slices with a slice spacing of 3 mm.

[0052] The OMRIT2 sequence contains 24 slices, with a slice spacing of 3 mm. The physician performs dose calculations on the CT images, and since the prostate's outline is clearer on MRI, the MRI image is aligned to the CT.

[0053] Specifically, the alignment method from MRI to CT includes: using a segmentation model to segment the prostate region frame by frame in the MRIT2 sequence, with the segmentation model being implemented using existing technology; locating the center point based on the femoral head region of the two modal OCT image sequences and the OMRIT2 image sequence, mapping the mask onto the CT sequence, and thus obtaining the CT mask; MRI alignment to CT.

[0054] Step S2: Based on the image sequence group obtained in step S1 and the data for aligning MRI to CT, the alignment is initiated using the method of Example 1 or Example 2.

[0055] Step S2.1, the parotid gland SUVmax automatic extraction (head processing) system first processes the head data from CT and PET, including: Based on a slice thickness of 3mm and a standard head length of 200mm, the system automatically extracts the first 66 frames (Frame 0-66) of the CT sequence as the head search area to lock the range; The system calculates the global SUVmax for each frame to obtain the original sequence S_head. Dynamic statistical thresholding and linear interpolation are used for repair; this achieves sequence calculation and denoising. A sudden spike in the SUV value to over 500 was observed in frames 15-18 (normal physiological uptake is typically <50), and high-density streaks were present in the corresponding CT slice. The system identified this as "metal artifact caused by gold teeth" and detected an anomaly; the values ​​in frames 15-18 were removed from the sequence. The maximum value was found in the remaining sequence, and the parotid gland SUVmax was measured to be 12.80. The test prediction was then output. In this embodiment, the brightest area of ​​the parotid gland, marked manually by the doctor, was measured, and the prostate SUVmax was measured to be 12.77, verifying the stability and accuracy of the test results.

[0056] Step S2.2, Automatic Extraction of Prostate SUVmax (Pelvic Processing) This step involves cross-modal spatial mapping from MRI to PET / CT: Sub-step 2.1.1: Feature-based fast alignment CT feature extraction; Sub-step 2.1.2 involves precisely overlaying the MRI prostate mask onto the corresponding slice of the PET image; Sub-step 2.1.3, Bladder avoidance and calculation: Extract the PET signal sequence S_prostate within the mask area.

[0057] Interference identification: The sequence is shown in the top 3 slices (head side), where the SUV value is as high as 35.0, while the value drops sharply to around 16.0 in subsequent slices.

[0058] System logic judgment: The highest value at the top indicates signal overflow at the bottom of the bladder.

[0059] Data cleaning: Automatically discard the top 3 frames of data.

[0060] Results: Taking the maximum value from the remaining slices, the prostate SUVmax was measured to be 16.50. This is almost identical to the SUVmax of 16.48 obtained by manually marking the brightest area of ​​the prostate by the doctor.

[0061] Interference identification and system logic judgment: High bladder signal truncation algorithm based on gradient mutation detection (for prostate).

[0062] Step S3: The intelligent diagnostic suggestion system automatically calculates the metabolic ratio: Ratio = prostate SUVmax / parotid gland SUVmax = 16.50 / 12.80 = 1.289; Diagnostic output: Since 1.289>1, the system automatically marked in the report: "The local metabolism of the prostate is significantly higher than the physiological background. Further puncture biopsy is recommended in conjunction with clinical findings." The entire process of implementation and effect analysis was run on a standard PC (Intel i7 CPU) without GPU involvement. Time taken: Approximately 4.5 seconds. Accuracy: Successfully avoided metal artifacts and bladder interference; the measured SUV value had an error of less than 5% compared to manual measurements by senior nuclear medicine physicians.

[0063] Working principle: This invention simplifies the complex three-dimensional voxel search problem into a local extremum problem based on the maximum standard uptake value, avoiding repetitive manual slide reading. The algorithm directly outputs structured data to the reporting system, significantly optimizing the workflow in nuclear medicine departments. Manually locating and delineating the ROI (Region of Interest) of the head and prostate in hundreds of slides takes an average of 3-5 minutes per case and requires a high degree of concentration from the physician. This algorithm, however, processes the entire process in less than 5 seconds per case (on a standard i7 CPU) without any manual interaction; greatly improving efficiency and reducing time costs.

[0064] The method of this invention is a deterministic algorithm, ensuring that the output is 100% unique and repeatable for the same set of input data. It completely eliminates the inconsistency in registration results caused by differences in operator experience, fatigue, and subjective judgment in manual or semi-automatic methods. Each step of the algorithm is based on objective mathematical calculations, guaranteeing standardization and high-quality alignment.

[0065] The algorithm of this invention can run smoothly on a single-core CPU of a regular PC, reducing algorithm complexity and hardware dependence. This invention does not rely on complex deep learning models that require massive amounts of data for training, and also avoids iterative optimizations that require powerful parallel computing capabilities. This makes the deployment cost of this invention extremely low, easy to popularize in various clinical environments, and eliminates the need for expensive hardware upgrades; the algorithm of this invention can run smoothly on a single-core CPU of a regular PC, reducing algorithm complexity and hardware dependence.

[0066] For example, in this invention, such as Figure 2 The image shown is a comparison diagram of image recognition according to a preferred embodiment of the present invention; Figure 3 It is a segmented prostate mask image; Figure 4 These are the extracted SUV sequence values ​​from the head and prostate of the test subject (which have been desensitized).

[0067] Example 4: A system for extracting maximum standard values ​​of the parotid gland and prostate based on CT and MRI images, implemented using any one of Examples 1 to 3, including: The input module is used to acquire detection data, which includes paired axial pelvic region computed tomography (CT) image sequences and axial pelvic region magnetic resonance imaging (MRI) image sequences of the subject. The diagnostic analysis module obtains the maximum standard uptake value of the parotid gland and the maximum standard uptake value of the prostate gland based on the test data, compares the maximum standard uptake value of the prostate gland with the maximum standard uptake value of the parotid gland, determines whether the metabolic level of the prostate gland is abnormal, and obtains corresponding risk warnings or status warnings. Output module; used for analyzing results.

[0068] The above specific embodiments are specific support for the concept proposed in this invention, and should not be used to limit the scope of protection of this invention. Any equivalent changes or modifications made on the basis of this technical solution in accordance with the technical concept proposed in this invention shall still fall within the scope of protection of this invention.

Claims

1. A method for extracting maximum standard uptake values ​​of the parotid gland and prostate gland based on CT and MRI images, characterized in that, Including the following methods: Obtain test data; Based on the detection data, the maximum standard uptake values ​​of the parotid gland and the prostate were extracted to obtain the maximum standard uptake values ​​of the parotid gland and the prostate. By comparing the maximum standard uptake value of the prostate gland with that of the parotid gland, it can be determined whether the metabolic level of the prostate gland is abnormal. When the ratio of the maximum standard uptake value of the prostate to the maximum standard uptake value of the parotid gland is greater than the preset value, the metabolic level of the prostate is greater than the average metabolic level of the whole body, and a risk warning is issued; when the ratio of the maximum standard uptake value of the prostate to the maximum standard uptake value of the parotid gland is less than the preset value, the metabolic level of the prostate is less than the average metabolic level of the whole body, and a status warning is issued. Based on the corresponding risk warnings or status alerts, the analysis results are output.

2. The method for extracting maximum standard value uptake analysis of the parotid gland and prostate gland based on CT and MRI images according to claim 1, characterized in that: The detection data includes paired axial pelvic region computed tomography (CT) image sequences and axial pelvic region magnetic resonance imaging (MRI) image sequences of the subjects, as well as axial head region computed tomography (CT) image sequences.

3. The method for extracting maximum standard value uptake analysis of the parotid gland and prostate gland based on CT and MRI images according to claim 1, characterized in that: Obtaining the maximum standard uptake value of the parotid gland involves calculating the maximum standard uptake value of the parotid gland in the head using the high spatial resolution of CT images, including the following steps: Obtain the head length and the width between CT images, calculate the head length / width between CT images, and obtain the number of CT image frames of the head; For each frame of the acquired head CT image, calculate the maximum standard access value of the CT image frame by frame, and store the maximum standard access value obtained in each frame in an array as the maximum standard access value array. Based on the obtained maximum standard uptake value array, if the patient's head contains metallic radioactive material, a repair algorithm is used to eliminate the interference of the maximum standard uptake value of metallic radioactive material on the maximum standard uptake value of the parotid gland. The local maximum value of the array after excluding the maximum standard uptake value of radioactive metals is taken as the target tissue; the maximum value of the target tissue is taken as the maximum standard uptake value of the parotid gland.

4. The method for extracting maximum standard value uptake analysis of the parotid gland and prostate gland based on CT and MRI images according to claim 3, characterized in that: The repair algorithm uses dynamic statistical thresholding and linear interpolation to repair abnormal mutations in SUVmax in CT images. These abnormal mutations include high-density stripe artifacts. The algorithm also removes artifacts based on statistical outlier detection. Includes the following steps: Extract each frame of CT image of the head I z The maximum standard uptake value is used to construct a one-dimensional original signal sequence. Where N is the total number of slices, v i SUV in frame i max The maximum standard uptake value (SUVmax) is a semi-quantitative indicator calculated based on the highest radioactivity concentration of pixels within the region of interest (ROI) in a PET image. For any voxel i in a frame of an image, its standard ingestion value SUV i The calculation formula is as follows: ； Among them, C i W represents the measured radioactivity concentration in the i-th voxel; t The test subject's weight is used to standardize the effect of volume distribution caused by differences in patient size; D represents the total radioactive dose of the injected tracer, and the dose of D has been corrected to the start of the scan according to the half-life formula. SUVs targeting the parotid gland or prostate area max That is, all voxels SUVs within the anatomical structure mask area. i The maximum value in the set of numbers: It eliminates differences in patient weight and injection dosage, making metabolic levels between different patients objectively comparable; When calculating the effective injection dose D in the SUV formula, the radionuclide decay law is used to correct the drug activity measured at the calibration time to the imaging start time. The correction formula includes: ； Where D represents the actual in vivo tracer activity corrected to the start of the scan; D cal This indicates the initial radioactivity measured on a dosimeter; T 1 / 2 Indicates the physical half-life of the radioactive isotope used; t scan Indicates the time point at which the PET scan begins; t cal Indicates the time point at which dose measurement was performed; t scan -t cal This indicates the time interval from measurement to scanning; by correcting for and eliminating dosage calculation errors caused by the time difference between injection and scanning, it ensures SUV accuracy. max The physical reality; Adaptive threshold calculation includes: Calculate the median (Median) and standard deviation (STD) of the sequence, and define the artifact detection threshold (T). thresh : Where Median is the median function, sigma is the standard deviation of the sequence, and k is the sensitivity coefficient. The original signal sequence, The standard deviation of the original sequence; Anomaly detection and labeling: Traverse the sequence, when a certain frame v i >T thresh If the corresponding frame is affected by metal artifacts, its index i is marked as an anomaly. Interpolation repair: For frames v marked as anomalous i Using adjacent normal frames v prev and v next Perform linear interpolation repair: ; and They are and The corresponding frame index; Final output: After eliminating metal artifacts, accurate peak parotid gland metabolic indicators were obtained; among them This is the repaired sequence.

5. The method for extracting maximum standard value uptake analysis of the parotid gland and prostate gland based on CT and MRI images according to claim 4, characterized in that: Obtaining the maximum standard uptake value of the prostate involves the following steps: The prostate mask obtained from MRI image segmentation is mapped to CT-PET space using the corresponding bone alignment information to construct a CT-PET mask and obtain the prostate region. Each frame of prostate image is obtained based on the prostate region, and the maximum standard uptake value within the prostate mask is calculated frame by frame. The value obtained in each frame is stored in an array, namely the maximum standard uptake value array of the prostate. Based on the obtained array of maximum standard uptake values ​​of the prostate, and using a truncation algorithm to eliminate the interference of the maximum standard uptake value of the bladder on the maximum standard uptake value of the prostate, the local maximum value in the array of maximum standard uptake values ​​of the prostate without interference sources is the target tissue; the maximum value of the target tissue is taken as the maximum standard uptake value of the prostate.

6. The method for extracting maximum standard value uptake analysis of the parotid gland and prostate gland based on CT and MRI images according to claim 6, characterized in that: The truncation algorithm is a high-order bladder signal truncation algorithm based on gradient mutation detection, which includes: adopting an automatic truncation algorithm based on one-dimensional sequential gradient mutation; when the mask contains the bladder, the maximum standard ingress value of the first few frames will be approximately equal and abnormally large compared to the subsequent values, so the maximum standard ingress value of the mask containing the bladder is set to zero.

7. The method for extracting maximum standard uptake analysis of the parotid gland and prostate gland based on CT and MRI images according to claim 7, characterized in that: To address the interference from high metabolic signals in the bladder that often appear in the head-side of prostate PET image sequences, an automatic truncation algorithm based on one-dimensional sequence gradient mutation is adopted. According to physiological principles, bladder signals exhibit a "continuous high-value plateau" at the beginning of the sequence, while prostate signals are located behind it and have lower intensity.

8. The method for extracting maximum standard value uptake analysis of the parotid gland and prostate gland based on CT and MRI images according to claim 8, characterized in that: The truncation algorithm includes the following steps; Within the registered prostate mask region, the maximum SUV value of each frame is extracted along the Z-axis to construct a one-dimensional metabolic signal sequence. Where N is the total number of slices; to eliminate the small fluctuations caused by imaging noise, the sequence is first smoothed by moving average: ； Where w is the radius of the sliding window, which enables sequence construction and preprocessing.

9. The first-order forward difference of the smoothed sequence is calculated using forward difference gradient to quantify metabolic variability between adjacent slices: ;when Furthermore, a larger value indicates a drop from a "high metabolism zone" to a "low metabolism zone"; The strategy involves identifying the "cliff-like drop point" at the beginning of the sequence to lock in the boundary of the high-level consolidation platform; among other things... A "cliff-like drop" must simultaneously satisfy two constraints: "gradient maximization" and "high-level platform verification". The constraints for maximizing the gradient include: finding the peak gradient within the first M frames of the sequence. ; The constraints for high-level platform verification include: verification k peak Previous sequence mean Is it significantly higher than the subsequent sequence mean? To conform to the characteristic of "larger first few frames": ； in, The average signal value before the peak. The average signal value after the peak value. is the index of the peak point in the sequence, and N is the total number of frames in the entire sequence; This represents the preprocessed signal value of the j-th frame. When satisfied Where α is the preset drop factor, then determine k peak k represents the true physical boundary between the bladder and prostate. * ; Based on the locked boundary k * Construct a binary weighted mask vector W: ； Multiply the original sequence by the mask to obtain the cleaned sequence Sclean, and extract the final result from it: ,in, S represents the maximum standardized uptake value of the prostate region, S is the original signal sequence, and W is the weight matrix; this achieves interference zeroing and target extraction.

10. A system for extracting maximum standard values ​​of the parotid gland and prostate based on CT and MRI images, implemented using the maximum standard value extraction and analysis method for the parotid gland and prostate based on CT and MRI images as described in any one of claims 1-9, characterized in that... include: The input module is used to acquire detection data, which includes paired axial pelvic region computed tomography (CT) image sequences and axial pelvic region magnetic resonance imaging (MRI) image sequences of the subject, as well as axial head region computed tomography (CT) image sequences. The diagnostic analysis module obtains the maximum standard uptake value of the parotid gland and the maximum standard uptake value of the prostate gland based on the test data, compares the maximum standard uptake value of the prostate gland with the maximum standard uptake value of the parotid gland, determines whether the metabolic level of the prostate gland is abnormal, and obtains corresponding risk warnings or status warnings. Output module; Used for analyzing results.

Images