Adrenal gland volume abnormality determination method based on CT image and related device

Abstract

The application discloses an adrenal gland volume anomaly judgment method and related device based on a CT image, and the method comprises the following steps: acquiring a CT scanning image of a subject, and performing pre-processing reconstruction to obtain CT image information; utilizing a pre-trained deep learning segmentation model to perform three-dimensional automatic segmentation and volume quantification processing on the CT image information, so as to obtain the adrenal gland volume of the subject; acquiring clinical characteristic parameters of the subject, and inputting the clinical characteristic parameters into a pre-set quantile regression model to obtain a volume reference interval of the subject; and outputting a judgment result according to the adrenal gland volume, the reference interval and the clinical characteristic parameters. Through fusion analysis on the clinical phenotype of the subject and the low-dose CT image, the reference range boundary, the volume measurement absolute value and the risk score of the patient are intuitively provided for medical personnel, and the accuracy of adrenal gland volume anomaly judgment is improved.

Description

Technical Field

[0001] This application relates to the field of medical image processing technology, and in particular to a method and related apparatus for judging adrenal gland volume abnormalities based on CT images. Background Technology

[0002] Abnormal adrenal gland volume is closely associated with a variety of diseases, including primary aldosteronism, Cushing's syndrome, adrenal tumors, diabetes, cardiovascular disease, and hypothalamic-pituitary-adrenal (HPA) axis disorders. If not diagnosed and intervened in a timely manner, it may increase the risk of cardiovascular disease and mortality by 1.5 to 10 times. Enlarged adrenal gland volume can serve as a biomarker for metabolic syndromes such as hypertension and diabetes. Therefore, accurate measurement of adrenal gland volume and automated detection and screening of abnormal volumes are of great significance in clinical practice.

[0003] In current technologies, clinical assessment of adrenal abnormalities primarily relies on conventional two-dimensional linear measurements. Due to the high variability in adrenal morphology, this traditional method has limited sensitivity and specificity in assessing abnormal size, and its consistency and diagnostic performance are relatively low. Furthermore, existing models linking adrenal volume to metabolic diseases rely on data from enhanced CT scans, resulting in limited sample sizes and neglecting individual variability. While some three-dimensional volume measurements can more comprehensively reflect adrenal size, they currently rely heavily on manual or semi-automatic contour drawing by medical professionals, a time-consuming and cumbersome process that hinders large-scale application in routine clinical diagnosis. Moreover, the relatively low image quality of low-dose chest CT scans used in routine health checkups further complicates automatic adrenal segmentation and volume measurement. Summary of the Invention

[0004] This application provides a method and related device for judging adrenal volume abnormalities based on CT images, which can improve the accuracy of judging adrenal volume abnormalities by fusing analysis of the subject's clinical phenotype and low-dose CT images.

[0005] In a first aspect, embodiments of this application provide a method for determining adrenal gland volume abnormalities based on CT images, the method comprising the following steps: CT scan images of the subjects were acquired and preprocessed to reconstruct CT image information. The CT image information is automatically segmented in three dimensions and its volume is quantized using a pre-trained deep learning segmentation model to obtain the adrenal gland volume of the subject. The clinical characteristic parameters of the subject are obtained and input into a preset quantile regression model to obtain the volume reference interval of the subject; The judgment result is output based on the adrenal gland volume, the reference interval, and the clinical characteristic parameters.

[0006] In some embodiments, acquiring the subject's CT scan images and performing preprocessing and reconstruction includes: The CT scan images of the subject are read by a scanner; Thin-layer reconstruction is performed on the CT scan image based on a preset thickness value to obtain the CT image information.

[0007] In some embodiments, the step of using a pre-trained deep learning segmentation model to perform three-dimensional automatic segmentation and volume quantization processing on the CT image information includes: The CT image information is input into a deep learning segmentation model based on the nnU-Net architecture to generate an initial 3D label map; A morphological dilation operation is performed on the initial three-dimensional label map to expand the adrenal gland segmentation boundary and restore missing edges, resulting in an expanded three-dimensional label map; Based on the pixel spacing, slice thickness, and the dilated three-dimensional label map of the CT image information, the adrenal gland volume of the subject is calculated by volumetric summation. The adrenal gland volume includes the left adrenal gland volume, the right adrenal gland volume, and the total adrenal gland volume.

[0008] In some embodiments, obtaining the clinical characteristic parameters of the subject and inputting the clinical characteristic parameters into a preset quantile regression model includes: The clinical characteristic parameters of the subject are obtained, including age, gender, body surface area and preset blood metabolic indicators, including at least one of blood pressure, blood glucose and blood lipids; The clinical characteristic parameters are input into the quantile regression model to obtain the lower and upper volume reference values ​​of the subject. The quantile regression model is constructed based on a preset interval of large-scale healthy population data.

[0009] In some embodiments, the step of outputting a judgment result based on the adrenal gland volume, the reference interval, and the clinical characteristic parameters includes: Based on the volume of the left adrenal gland and the volume of the right adrenal gland, a first determination result based on quantitative screening is determined; For samples with abnormalities in the first determination result, the clinical characteristic parameters and adrenal gland volume of the subject are input into an anomaly detector based on a Gaussian mixture model to determine a second determination result based on anomaly detection; The first determination result and the second determination result are combined to generate and output the determination result.

[0010] In some embodiments, determining a first determination result based on quantitative screening according to the volume of the left adrenal gland and the volume of the right adrenal gland includes: Calculate the difference between the volume of the left adrenal gland and the volume of the right adrenal gland; When the difference value is greater than a preset anatomical threshold, the first determination result is an anatomical abnormality.

[0011] In some embodiments, inputting the clinical characteristic parameters and adrenal gland volume of the subject into an anomaly detector based on a Gaussian mixture model includes: The clinical characteristic parameters of the subject are combined with the adrenal gland volume to construct a high-dimensional feature vector, which is used to characterize the individual physiological and metabolic state of the subject. The high-dimensional feature vector is input into the Gaussian mixture model, wherein the Gaussian mixture model is obtained by pre-training based on large-scale healthy population data; Calculate the negative log-likelihood of the high-dimensional feature vector in the Gaussian mixture model to obtain the anomaly score; When the abnormal score is greater than a preset probability threshold, the second determination result is a feature abnormality.

[0012] Secondly, embodiments of this application provide a device for determining adrenal gland volume abnormalities based on CT images, which is used to implement the method for determining adrenal gland volume abnormalities based on CT images in the first aspect, including: The acquisition module is used to acquire the subject's CT scan images, perform preprocessing and reconstruction, and obtain CT image information; The calculation module is used to perform three-dimensional automatic segmentation and volume quantization processing on the CT image information using a pre-trained deep learning segmentation model to obtain the adrenal gland volume of the subject. The reference module is used to acquire the clinical characteristic parameters of the subject and input the clinical characteristic parameters into a preset quantile regression model to obtain the volume reference interval of the subject. The judgment module is used to output a judgment result based on the adrenal gland volume, the reference interval, and the clinical characteristic parameters.

[0013] Thirdly, embodiments of this application provide an electronic device, characterized in that it includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it implements the adrenal gland volume abnormality judgment method based on CT images of the first aspect.

[0014] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the adrenal gland volume abnormality judgment method based on CT images of the first aspect.

[0015] The method, apparatus, and storage medium for judging adrenal volume abnormalities based on CT images in this application have at least the following beneficial effects: By acquiring CT scan images of the subject and performing preprocessing and reconstruction, CT image information is obtained; a pre-trained deep learning segmentation model is used to perform three-dimensional automatic segmentation and volume quantization processing on the CT image information to obtain the subject's adrenal volume; clinical characteristic parameters of the subject are acquired and input into a preset quantile regression model to obtain the subject's volume reference interval; and the judgment result is output based on the adrenal volume, reference interval, and clinical characteristic parameters. Through the fusion analysis of the subject's clinical phenotype and low-dose CT images, the system intuitively provides medical personnel with patient-specific reference range boundaries, absolute values ​​of volume measurements, and risk scores, improving the accuracy of adrenal volume abnormality judgment, reducing the reliance on experienced specialists for adrenal lesion identification, and enabling routine physical examination centers or primary healthcare personnel to obtain accurate and comprehensive early screening and diagnostic results using the information provided by this system.

[0016] Other features and advantages of this application will be set forth in the following description and will be apparent in part from the description or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the description and the accompanying drawings. Attached Figure Description

[0017] Figure 1 This is a flowchart of a method for judging adrenal gland volume abnormalities based on CT images according to an embodiment of the present invention; Figure 2 for Figure 1 Flowchart of step S1000; Figure 3 for Figure 1 Flowchart of step S2000; Figure 4 for Figure 1 Flowchart of step S3000; Figure 5 for Figure 1 Flowchart of step S4000; Figure 6 for Figure 5 Flowchart of step S4100; Figure 7 for Figure 5 Flowchart of step S4200; Figure 8 This is a structural diagram of an adrenal gland volume abnormality judgment device based on CT images according to an embodiment of the present invention; Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. Furthermore, the features, operations, or characteristics described in the specification can be combined in any suitable manner to form various implementations. Simultaneously, the steps or actions described in the method description can be rearranged or adjusted in a manner readily apparent to those skilled in the art. Therefore, the various orders in the specification and drawings are merely for the clear description of a particular embodiment and do not imply a mandatory order, unless otherwise stated that a particular order must be followed.

[0019] In the description of this application, "several" means one or more, "more than" means two or more, "greater than," "less than," and "exceeding" are understood to exclude the stated number, while "above," "below," and "within" are understood to include the stated number. The use of "first" and "second" in the description is merely for distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features.

[0020] The serial numbers assigned to components in this document, such as "first" and "second," are used only to distinguish the described objects and have no sequential or technical meaning. The terms "connection" and "linkage" used in this application, unless otherwise specified, include both direct and indirect connections (linkages).

[0021] This invention relates to a method and related apparatus for assessing adrenal gland volume abnormalities based on CT images. Computed tomography (CT) utilizes X-rays to penetrate the human body and combines rotational scanning with computer reconstruction technology to generate tomographic images of the body's internal structures. The core principle of CT is to acquire X-ray attenuation data from different angles and then reconstruct cross-sectional images using mathematical algorithms, aiding in the observation of tissue density differences. CT is a diagnostic method with high image quality, high diagnostic value, and is non-invasive, painless, and risk-free, enabling diagnosticians to reconstruct various layers of the body's structure at any depth or angle. Furthermore, CT can display the location of post-traumatic brain injury sequelae, damage, brain tumors, and other lesions, and is widely used in scientific research and clinical fields.

[0022] Abnormal adrenal gland volume is closely associated with a variety of diseases, including primary aldosteronism, Cushing's syndrome, adrenal tumors, diabetes, cardiovascular disease, and hypothalamic-pituitary-adrenal (HPA) axis disorders. If not diagnosed and intervened in a timely manner, it may increase the risk of cardiovascular disease and mortality by 1.5 to 10 times. Enlarged adrenal gland volume can serve as a biomarker for metabolic syndromes such as hypertension and diabetes. Therefore, accurate measurement of adrenal gland volume and automated detection and screening of abnormal volumes are of great significance in clinical practice.

[0023] In current technologies, the clinical assessment of adrenal abnormalities mainly relies on conventional two-dimensional linear measurements. Due to the high variability in adrenal morphology, this traditional measurement method has limited sensitivity and specificity in assessing abnormal size, and its consistency and diagnostic performance are relatively low. Furthermore, existing models linking adrenal volume to metabolic diseases rely on data from enhanced CT scans, resulting in limited sample sizes and neglecting individual differences. While some three-dimensional volume measurements can more comprehensively reflect adrenal size, they currently rely heavily on manual or semi-automatic contour drawing by medical professionals, a time-consuming and cumbersome process that is difficult to apply on a large scale in routine clinical diagnosis. In addition, routine health checkups often use low-dose chest CT scans, resulting in relatively low image quality, further increasing the difficulty of automatic adrenal segmentation and volume measurement. Low-dose CT refers to computed tomography (CT) scanning technology that significantly reduces radiation dose by optimizing scanning parameters, such as reducing tube voltage and current, based on conventional CT scans. It is widely used in routine health checkups for lung cancer and other conditions.

[0024] Currently, some systems and studies have attempted to assess the normal range of adrenal gland volume and perform disease association analysis. However, these existing techniques often rely on small sample data or non-healthy populations with potential complications, resulting in results that cannot represent the general healthy population. Specifically, most studies use fixed volume cutoff values, i.e., fixed thresholds, as the criteria for judging abnormalities, failing to fully consider the significant impact of individual physiological and metabolic factors such as age, sex, body size (e.g., body surface area, body mass index), blood pressure, and blood sugar on adrenal gland volume. Due to the lack of personalized normal reference ranges based on a large-scale healthy population, this single standardized threshold method has extremely limited applicability in practical clinical applications and is highly likely to lead to misdiagnosis or missed diagnosis of adrenal gland lesions.

[0025] Based on the above, this invention provides a method and related device for judging adrenal volume abnormalities based on CT images. The method involves acquiring CT scan images of the subject, performing preprocessing and reconstruction to obtain CT image information, using a pre-trained deep learning segmentation model to perform three-dimensional automatic segmentation and volume quantization of the CT image information to obtain the subject's adrenal volume, acquiring the subject's clinical characteristic parameters, and inputting these parameters into a preset quantile regression model to obtain the subject's volume reference interval, and outputting the judgment result based on the adrenal volume, reference interval, and clinical characteristic parameters. Through the fusion analysis of the subject's clinical phenotype and low-dose CT images, the system intuitively provides medical personnel with patient-specific reference range boundaries, absolute volume measurements, and risk scores, improving the accuracy of adrenal volume abnormality judgment, reducing reliance on experienced specialists for adrenal lesion identification, and enabling routine medical examination centers or primary healthcare personnel to obtain accurate and comprehensive early screening and diagnostic results using the information provided by this system.

[0026] Please see Figure 1 , Figure 1 The flowchart illustrates a method for determining adrenal gland volume abnormalities based on CT images, provided by an embodiment of the present invention. For example... Figure 1 As shown, the method for determining adrenal gland volume abnormalities based on CT images according to an embodiment of the present invention includes the following steps: Step S1000: Acquire the CT scan image of the subject and perform preprocessing and reconstruction to obtain CT image information.

[0027] Understandably, to accurately identify CT image information that can be directly read and processed by the model, it is necessary to acquire the subject's CT scan images and preprocess and reconstruct them. To accurately acquire CT scan images corresponding to the subject's adrenal glands, a CT scan of the subject's chest is required. Specifically, the adrenal glands are important endocrine glands located above both kidneys. Due to their small size, weighing approximately 5-7 grams, they secrete various key hormones that participate in regulating metabolism, blood pressure levels, stress responses, and water-electrolyte balance, serving as indispensable chemical signaling stations for maintaining life. There is one adrenal gland on each side, located above and inward of each kidney, protected by the renal fascia and adipose tissue. They have a rich blood supply, making them one of the organs with the highest blood flow per unit weight in the body, ensuring rapid release of hormones into the bloodstream. Therefore, a precise, high-resolution CT scan of the subject's adrenal glands is necessary to obtain CT image information.

[0028] Please see Figure 2 , Figure 2 A schematic diagram illustrating the specific implementation process of step S1000 above is shown. For example... Figure 2 As shown, step S1000 includes at least the following steps: Step S1100: Read the subject's CT scan image using a scanner.

[0029] Understandably, the subject lies on the examination bed, and the X-ray tube and detectors of the CT scanner work together to scan the subject's chest, acquiring a CT scan image of the chest. During operation, the X-ray tube emits a fan-shaped X-ray beam from one side of the body; the rays attenuate as they pass through different tissues, and the detector array on the other side of the X-ray tube receives the remaining X-rays and converts them into electrical signals. Then, the X-ray tube and detectors rotate 360° around the body, acquiring projection data from multiple angles. This projection data is then transmitted to a computer, which uses filtered backprojection or iterative reconstruction algorithms to calculate the attenuation coefficient at each location. The attenuation coefficient is then converted into a grayscale image, resulting in a single-layer tomographic CT scan image. As the examination bed moves forward, multiple layers are scanned consecutively, forming a tomographic sequence covering the adrenal glands, thus creating the subject's CT scan image. Specifically, reading the subject's CT scan image using a scanner is existing technology and will not be elaborated upon here.

[0030] Step S1200: Perform thin-layer reconstruction on the CT scan image based on the preset thickness value to obtain CT image information.

[0031] Understandably, after acquiring the subject's CT scan images, the original CT scan images are thick slices. To facilitate data processing and analysis in subsequent steps, it is necessary to recalculate thinner tomographic images from the original projected CT scan images using computer algorithms—that is, to perform thin-slice reconstruction. Specifically, since specific parameters are used when acquiring low-dose chest CT scan images, such as a tube voltage of 100kVp and a CT scan pitch of 1.0, thin-slice reconstruction with a thickness of 1mm is performed on the acquired scan images, with a reconstruction interval of 5mm, using a 512×512 matrix. In practical applications, thin-slice reconstruction with a thickness of 1mm can better display the morphological details of the adrenal glands, providing basic data for high-precision segmentation. Specifically, thin-slice reconstruction of CT scan images based on preset thickness values ​​is existing technology and will not be elaborated here.

[0032] Step S2000: Use a pre-trained deep learning segmentation model to perform three-dimensional automatic segmentation and volume quantization of CT image information to obtain the adrenal gland volume of the subject.

[0033] Understandably, low-dose CT images often exhibit blurred edges. Applying morphological dilation post-processing can help effectively expand the segmentation boundaries of CT image information and recover potentially missed peripheral edge regions, thus avoiding incorrect segmentation by deep learning models under low contrast conditions.

[0034] Please see Figure 3 , Figure 3 A schematic diagram illustrating the specific implementation process of step S2000 above is shown. For example... Figure 3 As shown, step S2000 includes at least the following steps: Step S2100: Input the CT image information into the deep learning segmentation model based on the nnU-Net architecture to generate the initial 3D label map.

[0035] Understandably, a large amount of labeled data is used as input samples for training the deep learning segmentation model. nnU-Net, or Adaptive U-Net framework, is a deep learning network architecture that can adapt to various medical image segmentation tasks without manual hyperparameter tuning, automatically configuring preprocessing, network topology, and post-processing strategies. During training, a stochastic gradient descent (SGD) optimizer with momentum set to 0.99 is used, with an initial learning rate of 0.01 and a multinomial decay strategy. To overcome noise interference from low-dose CT and augment the dataset, data augmentation strategies are randomly applied during training, including random rotation, elastic deformation, and adding Gaussian noise, with each augmentation strategy having an independent application probability of 0.2. Preferably, a loss function combining Dice loss and cross-entropy loss is used to address the problem of extreme foreground-background imbalance in medical images.

[0036] It should be noted that a pre-trained deep learning segmentation model is used to automatically segment the input low-dose chest CT image information to generate a 3D label map of the adrenal glands. This involves classifying each pixel in the CT image information and assigning a category label to each pixel, thereby accurately segmenting the target region from the background and generating an initial 3D label map for the adrenal glands.

[0037] Step S2200: Perform morphological dilation on the initial 3D label map to expand the adrenal segmentation boundary and restore missing edges, resulting in a dilated 3D label map.

[0038] Understandably, because low-dose CT images are prone to blurred edges, a morphological dilation operation using a 3x3 structuring element is needed as a post-processing step on the initial 3D label image to expand the adrenal gland segmentation boundary and recover any potentially missed surrounding areas. Specifically, a structuring kernel is slid across the initial 3D label image; if any pixel within the kernel is foreground, the current pixel becomes foreground. The morphological dilation operation on the initial 3D label image expands the adrenal gland segmentation boundary and recovers any missed edges, resulting in an expanded 3D label image.

[0039] Step S2300: Based on the pixel spacing, slice thickness, and dilation three-dimensional label map of the CT image information, calculate the adrenal volume of the subject by volume accumulation. The adrenal volume includes the volume of the left adrenal gland, the volume of the right adrenal gland, and the total adrenal volume.

[0040] Understandably, pixel spacing in CT image information represents the actual physical distance between two adjacent pixels within the same CT slice. Slice thickness in CT image information is the physical slice thickness of the X-ray beam during scanning, representing the actual tissue thickness contained in that CT slice; common slice thicknesses include 1.0mm, 1.25mm, 5mm, and 7mm. After acquiring the dilated 3D label image and determining the pixel spacing and slice thickness of the CT image information, the voxel size of the CT image information can be determined, converting the number of pixels of the segmented adrenal gland into its actual volume. Single voxel volume. The specific calculation formula is as follows:

[0041] in, The row width is in pixels. Pixel column width, The thickness is the layer thickness.

[0042] This gives the adrenal gland volume. The calculation formula is as follows:

[0043] in, The number of pixels for the left adrenal gland, right adrenal gland, or total adrenal gland.

[0044] Step S3000: Obtain the clinical characteristic parameters of the subject and input the clinical characteristic parameters into the preset quantile regression model to obtain the subject's volume reference interval.

[0045] Understandably, unlike the fixed thresholds used in traditional assessment methods, this application's embodiments dynamically calculate the lower and upper limits of the reference interval based on the subject's multidimensional clinical characteristic parameters. Specifically, the normal range for the subject's adrenal gland volume is between the 2.5th and 97.5th percentiles. When the subject's clinical indicators are fully collected, a preset quantile regression model is invoked to calculate the lower and upper reference values ​​for adrenal gland volume, thus obtaining the volume reference interval for the subject's clinical characteristic parameters, improving the reliability and scientific validity of the assessment results.

[0046] Please see Figure 4 , Figure 4This diagram illustrates the specific implementation process of step S3000 above when the pathology label is an HE staining label. Figure 4 As shown, step S3000 includes at least the following steps: Step S3100: Obtain the subject's clinical characteristic parameters, including age, gender, body surface area, and preset blood metabolism indicators, including at least one of blood pressure, blood glucose, and blood lipids.

[0047] Understandably, compared to the limitations of existing technologies that rely on single, fixed volume thresholds for assessment, incorporating multi-dimensional physiological and metabolic indicators helps establish personalized, dynamic reference standards for subjects, significantly reducing the risk of misdiagnosis or missed diagnosis due to individual physiological differences. Specifically, clinical characteristic parameters include age, sex, body surface area, and preset blood metabolic indicators; blood metabolic indicators include at least one of blood pressure, blood glucose, and blood lipids. In practical applications, clinical characteristic parameters can be obtained by directly reading the subject's physical examination report, blood test report, etc., or by having the subject register or upload them themselves; no restrictions are placed here.

[0048] Step S3200: Input the clinical characteristic parameters into the quantile regression model to obtain the lower and upper volume reference values ​​of the subjects. The quantile regression model is constructed based on the preset interval of large-scale healthy population data.

[0049] Understandably, quantile regression (QR) is a regression analysis method that does not target the mean but estimates the linear or non-linear relationship between specific quantiles (such as the 2.5% and 97.5%) of the response variable and the independent variable, making it suitable for constructing medical reference intervals. Specifically, by modeling different clinical characteristic parameters, such as age or gender, the 2.5%, 5%, 50%, 95%, and 97.5% quantiles of medical indicators directly construct individualized medical reference intervals that vary with the independent variable. Since medical data is generally right-skewed and heavy-tailed, QR does not require normality or homogeneity of variance, making it a natural fit. Secondly, quantile regression models can construct covariate-dependent dynamic reference intervals. For example, for age, the upper and lower reference limits automatically and smoothly change with each year of age increase, which is more in line with physiological patterns. For outliers in medical data, quantile regression models, based on absolute loss, are more robust than algorithms such as least squares and mean-based methods. Quantile regression models can also adjust for confounding factors, such as correcting for gender, body mass index (BMI), and smoking history, to obtain corrected standardized reference intervals.

[0050] In practical applications, the lower and upper reference values ​​for total adrenal gland volume (unit: cm) are as follows: 3 The calculation formula for ) is as follows: Lower limit=-2.739+0.016×Age+0.431×Sex+0.032×WC+0.874× BSA+0.007×DBP+0.081×TG-0.445×HDL+0.306×FPG Upper limit=-3.549+0.031×Age+0.339×Sex+0.030×WC+2.721× BSA+0.020×DBP+0.004×TG-1.244×HDL+0.447×FPG Where Age is age, Sex is gender (male is assigned a value of 1, female is assigned a value of 0), WC is waist circumference, BSA is body surface area, DBP is diastolic blood pressure, TG is triglycerides, HDL is high-density lipoprotein cholesterol, and FPG is fasting blood glucose.

[0051] When some blood biochemical indicators are missing in the subject, a simplified indicator model is used. The formulas for calculating the lower and upper reference values ​​of the total adrenal volume are as follows: Lower limit=-1.575+0.022×Age+0.629×Sex+0.103×BMI+0.957× BSA Upper limit=-3.894+0.035×Age+1.039×Sex+0.225×BMI+ 2.444×BSA BMI stands for Body Mass Index.

[0052] Step S4000: Output the judgment result based on adrenal gland volume, reference range, and clinical characteristic parameters.

[0053] Understandably, comparing and analyzing the adrenal volume, reference range, and clinical characteristic parameters obtained through the above steps can provide accurate judgment results regarding the subject's adrenal volume abnormalities. Specifically, the judgment result output process is divided into two stages, with the union of the identified abnormal results. The first stage screening, by comparing the size of the adrenal glands on both sides, can quickly capture anatomical asymmetry caused by unilateral adrenal space-occupying lesions; the second stage, through multidimensional feature assessment, can track occult adrenal hyperplasia caused by systemic metabolic disorders. This two-stage progressive screening improves the robustness of the model from the perspective of scientific probability distribution.

[0054] Please see Figure 5 , Figure 5 A schematic diagram illustrating the specific implementation process of step S4000 above is shown. For example... Figure 5 As shown, step S4000 includes at least the following steps: Step S4100: Determine the first judgment result based on quantitative screening according to the volume of the left adrenal gland and the volume of the right adrenal gland.

[0055] Understandably, because the adrenal glands are symmetrical structures, with one on each side, the vast majority of adrenal lesions are unilateral and solitary. Bilateral lesions are mostly metastatic or systemic diseases, and their incidence is even lower. The probability of unilateral adrenal lesions is higher than bilateral lesions; this is a common clinical finding. Therefore, in the first stage of screening, comparing the size difference between the left and right adrenal glands can accurately determine whether a unilateral space-occupying lesion has occurred. Specifically, to identify a unilateral space-occupying lesion—an abnormal adrenal gland that occurs only on one side of the body, occupies space, and compresses surrounding tissues—the first diagnostic result based on quantitative screening can be accurately obtained by calculating the difference in volume between the left and right adrenal glands and the magnitude of this difference.

[0056] Please see Figure 6 , Figure 6 A schematic diagram illustrating the specific implementation process of step S4100 above is shown. For example... Figure 6 As shown, step S4100 includes at least the following steps: Step S4110: Calculate the difference between the volume of the left adrenal gland and the volume of the right adrenal gland.

[0057] It is understandable that the difference between the left and right adrenal gland volumes obtained from the above steps can be calculated by subtracting the two volumes. In practical applications, the absolute value of the difference between the left and right adrenal gland volumes can be used as the difference between the two volumes.

[0058] Step S4120: When the difference value is greater than the preset anatomical threshold, the first determination result is anatomical abnormality.

[0059] Understandably, when the difference value exceeds a preset anatomical threshold—for example, 75% of the volume difference value in a normal population—the sample is initially classified as anatomically abnormal, suggesting a unilateral lesion or space-occupying lesion. In this case, the first determination is anatomical abnormality. In other embodiments, when the difference value is less than or equal to the anatomical threshold, the first determination is normal.

[0060] Step S4200: For samples with abnormalities in the first judgment result, input the subject's clinical characteristic parameters and adrenal gland volume into the anomaly detector based on Gaussian mixture model to determine the second judgment result based on anomaly detection.

[0061] Understandably, when there are abnormal samples in the first judgment result, in order to make more accurate anomaly judgment, a deep evaluation is carried out using a pre-constructed high-dimensional joint probability distribution model of multidimensional clinical features and adrenal volume to identify occult abnormalities of the adrenal gland.

[0062] Please see Figure 7 , Figure 7 A schematic diagram illustrating the specific implementation process of step S4200 above is shown. For example... Figure 7 As shown, step S4200 includes at least the following steps: Step S4210: Combine the clinical characteristic parameters of the subject with the adrenal gland volume to construct a high-dimensional feature vector. The high-dimensional feature vector is used to characterize the individual physiological and metabolic state of the subject.

[0063] Understandably, the demographic indicators (age, sex), body mass index (BSA, BMI), blood pressure (diastolic blood pressure DBP), blood metabolic indicators (fasting plasma glucose, triglycerides (TG, high-density lipoprotein cholesterol, HDL, etc.), and the adrenal volume data obtained in the above steps are combined to construct a high-dimensional feature vector reflecting the individual's physiological and metabolic state. .

[0064] Step S4220: Input the high-dimensional feature vector into the Gaussian mixture model, wherein the Gaussian mixture model is obtained by pre-training based on large-scale healthy population data.

[0065] It is understandable that the feature vector The input is a Gaussian Mixture Model (GMM) pre-trained based on large-scale healthy population data. A Gaussian Mixture Model is a probability-based generative model that fits the probability density distribution of data through a weighted combination of multiple Gaussian distributions. It is commonly used for clustering and anomaly detection of complex multidimensional data. In this embodiment, the model fits the feature distribution of the healthy population through a weighted combination of multiple Gaussian distributions. The calculation process of its probability density function is shown in the following formula:

[0066] in, For the first The weights of the Gaussian components, and satisfying ; It is the mean vector. Let be the covariance matrix. Step S4230: Calculate the negative log-likelihood of the high-dimensional feature vector in the Gaussian mixture model to obtain the anomaly score.

[0067] Understandably, the negative log-likelihood is a commonly used loss function in statistics and machine learning. By taking the negative value of the log-likelihood function, the problem of maximizing parameter estimation is transformed into a minimization problem of the optimization algorithm. In the embodiments of this application, the negative log-likelihood (NLL) of the subject's feature vector under this model is calculated as the anomaly score. The calculation process of the anomaly score is shown in the following formula:

[0068] Specifically, the abnormality score reflects the degree to which the characteristics of the test subjects deviate from the mean distribution of the healthy population.

[0069] Step S4240: When the abnormal score is greater than the preset probability threshold, the second determination result is a feature abnormality.

[0070] It is understandable that, consistent with step S4120 above, after obtaining the abnormal score, a preset probability threshold is used as the judgment boundary. Specifically, the probability threshold is 95% of the abnormal scores of healthy individuals. If the subject's abnormal score... If the probability threshold is exceeded, the second determination result is a characteristic abnormality. This stage can effectively track occult adrenal hyperplasia caused by systemic metabolic disorders that is easily overlooked in routine anatomical assessments.

[0071] Step S4300: Take the union of the first judgment result and the second judgment result, generate and output the judgment result.

[0072] It should be noted that the first and second judgment results obtained in the above steps are logically ORed, i.e., the union is taken, to generate the final judgment result. The specific process is as follows: If the subject is judged to have anatomical abnormalities in the first stage, focusing on asymmetry caused by unilateral space-occupying lesions, or is judged to have characteristic abnormalities in the second stage, focusing on occult changes associated with multidimensional metabolic indicators, the final comprehensive screening result for adrenal gland volume abnormalities will be output, triggering a clinical warning. Through this two-stage progressive screening, the embodiments of this application can not only objectively quantify lesions caused by changes in anatomical morphology, but also assess volume shifts caused by metabolic factors from the perspective of statistical probability distribution, significantly improving the sensitivity and clinical applicability of the screening system.

[0073] like Figure 8 As shown, Figure 8 This is a schematic diagram of the structure of the adrenal volume abnormality judgment device 500 based on CT images provided in the embodiments of this application. The entire process of the adrenal volume abnormality judgment method based on CT images provided in the embodiments of this application involves the following modules in the adrenal volume abnormality judgment device 500 based on CT images: acquisition module 510, calculation module 520, reference module 530 and judgment module 540.

[0074] The acquisition module 510 is used to acquire the subject's CT scan images, perform preprocessing and reconstruction, and obtain CT image information. The calculation module 520 is used to perform three-dimensional automatic segmentation and volume quantization of CT image information using a pre-trained deep learning segmentation model to obtain the adrenal gland volume of the subject. Reference module 530 is used to acquire the clinical characteristic parameters of the subject and input the clinical characteristic parameters into a preset quantile regression model to obtain the subject's volume reference interval; The judgment module 540 is used to output the judgment result based on the adrenal gland volume, reference range, and clinical characteristic parameters.

[0075] It should be noted that the information interaction and execution process between the modules of the above-mentioned device are based on the same concept as the method embodiment of this application. For details on their specific functions and technical effects, please refer to the method embodiment section, and they will not be repeated here.

[0076] like Figure 9 As shown, Figure 9 This is a schematic diagram of a controller 600 provided in one embodiment of this application.

[0077] The controller 600 in this embodiment includes one or more processors 610 and a memory 620. Figure 9 The example uses a processor 610 and a memory 620.

[0078] The processor 610 and the memory 620 can be connected via a bus or other means. Figure 9 Taking the example of a connection between China and Israel via a bus.

[0079] Memory 620, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory 620 may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory 620 may optionally include memory 620 remotely located relative to processor 610, and these remote memories can be connected to controller 600 via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0080] Those skilled in the art will understand that Figure 9 The device structure shown does not constitute a limitation on the controller 600 and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0081] This application embodiment also provides a storage medium storing computer-executable instructions for executing the above-described method for judging adrenal gland volume abnormalities based on CT images.

[0082] In one embodiment, the storage medium stores computer-executable instructions that are executed by one or more processors 610, such as one of the processors 610 in the controller 600, which can cause the one or more processors 610 to perform the adrenal gland volume abnormality judgment method based on CT images provided in any embodiment of this application.

[0083] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network nodes. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0084] It will be understood by those skilled in the art that all or some of the steps and systems in the methods disclosed above can be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components can be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, as is known to those skilled in the art, communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0085] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0086] In the several embodiments provided in this application, it should be understood that the disclosed systems, instruments, and methods can be implemented in other ways. For example, the instrument embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the shown or discussed mutual couplings, direct couplings, or communication connections may be through some interfaces; indirect couplings or communication connections between instruments or units may be electrical, mechanical, or other forms. Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, i.e., they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0087] It should also be understood that the various implementation methods provided in this application can be combined arbitrarily to achieve different technical effects.

[0088] The above is a detailed description of the preferred embodiments of this application. However, this application is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. A method for determining adrenal gland volume abnormalities based on CT images, characterized in that, The method includes the following steps: CT scan images of the subjects were acquired and preprocessed to reconstruct CT image information. The CT image information is automatically segmented in three dimensions and its volume is quantized using a pre-trained deep learning segmentation model to obtain the adrenal gland volume of the subject. The clinical characteristic parameters of the subject are obtained and input into a preset quantile regression model to obtain the volume reference interval of the subject; The judgment result is output based on the adrenal gland volume, the reference interval, and the clinical characteristic parameters.

2. The judgment method according to claim 1, characterized in that, The acquisition of the subject's CT scan images and the preprocessing and reconstruction include: The CT scan images of the subject are read by a scanner; Thin-layer reconstruction is performed on the CT scan image based on a preset thickness value to obtain the CT image information.

3. The judgment method according to claim 1, characterized in that, The process of automatically segmenting and quantizing the CT image information using a pre-trained deep learning segmentation model includes: The CT image information is input into a deep learning segmentation model based on the nnU-Net architecture to generate an initial 3D label map; A morphological dilation operation is performed on the initial three-dimensional label map to expand the adrenal gland segmentation boundary and restore missing edges, resulting in an expanded three-dimensional label map; Based on the pixel spacing, slice thickness, and the dilated three-dimensional label map of the CT image information, the adrenal gland volume of the subject is calculated by volumetric summation. The adrenal gland volume includes the left adrenal gland volume, the right adrenal gland volume, and the total adrenal gland volume.

4. The judgment method according to claim 3, characterized in that, The step of obtaining the clinical characteristic parameters of the subject and inputting the clinical characteristic parameters into a preset quantile regression model includes: The clinical characteristic parameters of the subject are obtained, including age, gender, body surface area and preset blood metabolic indicators, including at least one of blood pressure, blood glucose and blood lipids; The clinical characteristic parameters are input into the quantile regression model to obtain the lower and upper volume reference values ​​of the subject. The quantile regression model is constructed based on a preset interval of large-scale healthy population data.

5. The judgment method according to claim 4, characterized in that, The step of outputting a judgment result based on the adrenal gland volume, the reference interval, and the clinical characteristic parameters includes: Based on the volume of the left adrenal gland and the volume of the right adrenal gland, a first determination result based on quantitative screening is determined; For samples with abnormalities in the first determination result, the clinical characteristic parameters and adrenal gland volume of the subject are input into an anomaly detector based on a Gaussian mixture model to determine a second determination result based on anomaly detection; The first determination result and the second determination result are combined to generate and output the determination result.

6. The judgment method according to claim 5, characterized in that, The determination of the first judgment result based on quantitative screening, according to the volume of the left adrenal gland and the volume of the right adrenal gland, includes: Calculate the difference between the volume of the left adrenal gland and the volume of the right adrenal gland; When the difference value is greater than a preset anatomical threshold, the first determination result is an anatomical abnormality.

7. The determination method according to claim 5, characterized in that, The step of inputting the clinical characteristic parameters and adrenal gland volume of the subject into an anomaly detector based on a Gaussian mixture model includes: The clinical characteristic parameters of the subject are combined with the adrenal gland volume to construct a high-dimensional feature vector, which is used to characterize the individual physiological and metabolic state of the subject. The high-dimensional feature vector is input into the Gaussian mixture model, wherein the Gaussian mixture model is obtained by pre-training based on large-scale healthy population data; Calculate the negative log-likelihood of the high-dimensional feature vector in the Gaussian mixture model to obtain the anomaly score; When the abnormal score is greater than a preset probability threshold, the second determination result is a feature abnormality.

8. A device for determining adrenal gland volume abnormalities based on CT images, used to implement the method for determining adrenal gland volume abnormalities based on CT images according to any one of claims 1 to 7, characterized in that, include: The acquisition module is used to acquire the subject's CT scan images, perform preprocessing and reconstruction, and obtain CT image information; The calculation module is used to perform three-dimensional automatic segmentation and volume quantization processing on the CT image information using a pre-trained deep learning segmentation model to obtain the adrenal gland volume of the subject. The reference module is used to acquire the clinical characteristic parameters of the subject and input the clinical characteristic parameters into a preset quantile regression model to obtain the volume reference interval of the subject. The judgment module is used to output a judgment result based on the adrenal gland volume, the reference interval, and the clinical characteristic parameters.

9. An electronic device, characterized in that, include: The method for determining adrenal gland volume abnormalities based on CT images as described in any one of claims 1 to 7 includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for determining adrenal gland volume abnormalities based on CT images as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The device contains a computer program that, when executed by a processor, implements the method for determining adrenal gland volume abnormalities based on CT images as described in any one of claims 1 to 7.

Images