Metabolism analysis method and metabolism analysis system for living body

Abstract

A metabolism analysis method for a living body is used for analyzing the metabolic state of an anatomical structure of a target living body. The metabolism analysis method comprises the following steps: step S10, aiming at each pair of anatomical structures to be analyzed, according to a standard linear regression model of relative SUV values or metabolic parameter values of a pair of anatomical structures, calculating student residual errors of the relative SUV values or metabolic parameter values of the pair of anatomical structures of a target organism in the standard linear regression model, the pair of anatomical structures belong to the brain of an organism or belong to the body of the organism; and step S20, aiming at each pair of anatomical structures to be analyzed, judging whether the pair of anatomical structures of the target organism have metabolic abnormality or not according to the student residual error. The metabolism analysis method can sensitively identify a potential lesion-causing area under the condition that the sample size is small. The invention further provides a metabolism analysis system capable of executing the metabolism analysis method.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image analysis, and in particular to a metabolic analysis method and a metabolic analysis system for an organism. Background Art

[0002] Traditional methods of identifying potential pathogenic areas mainly rely on threshold judgments of metabolic indicators of a single anatomical structure. Taking the liver as an example, when the maximum standardized uptake value (maxSUV) of the liver is greater than 2.5, it is usually judged to be functionally abnormal. However, this detection method has obvious limitations. On the one hand, it only focuses on the local characteristics of a single anatomical structure and completely ignores the intrinsic relationship between different anatomical structures at the metabolic level; on the other hand, the threshold-based judgment method is highly dependent on past experience and it is difficult to achieve accurate consideration of individual differences. For example, for obese people, the metabolic characteristics of their physiological state are different from those of normal people, resulting in the threshold standards originally applicable to normal people being inaccurate for them and unable to effectively reflect the actual functional status of the anatomical structure.

[0003] While anomaly detection models built using deep learning methods currently demonstrate advantages in certain areas, they also present significant challenges. These models are essentially black-box models with no interpretability, making it difficult to fully understand their logic. Furthermore, with small sample sizes, model training accuracy struggles to achieve optimal levels, potentially leading to inaccurate metabolic analysis results. Summary of the Invention

[0004] The purpose of the present invention is to provide a metabolic analysis method for an organism, which can sensitively identify potential pathogenic focus areas with a small sample size.

[0005] Another object of the present invention is to provide a metabolic analysis system capable of performing the above-mentioned metabolic analysis method for an organism.

[0006] The present invention provides a metabolic analysis method for an organism, which is used to analyze the metabolic state of the anatomical structures of a target organism. The metabolic analysis method includes: step S10, for each pair of anatomical structures to be analyzed, based on a standard linear regression model of the relative SUV values or metabolic parameter values of the pair of anatomical structures, calculating the studentized residual of the relative SUV values or metabolic parameter values of the pair of anatomical structures of the target organism in the standard linear regression model, where the pair of anatomical structures belongs to the same brain or body of the organism; and step S20, for each pair of anatomical structures to be analyzed, determining whether the pair of anatomical structures of the target organism has a metabolic abnormality based on the studentized residual.

[0007] Based on a standard linear regression model of relative SUV values or metabolic parameter values for a pair of anatomical structures, the studentized residuals of the relative SUV values or metabolic parameter values for the target organism are calculated to determine whether metabolic abnormalities exist in the target organism's pair of anatomical structures. This metabolic analysis method requires a small sample size and can sensitively identify potential pathogenic areas.

[0008] In another exemplary embodiment of the metabolic analysis method of an organism, the standard linear regression model is represented by the formula shown in formula (2):

[0009] y=β0+β1x Formula (2)

[0010] Where x and y are the relative SUV values or metabolic parameter values of the pair of anatomical structures, respectively;

[0011] β1 is the regression coefficient, which is calculated by the formula shown in formula (3):

[0012]

[0013] Where n is the number of samples, 1≤i≤n, and the samples are healthy organisms.

[0014] x i and y i are the relative SUV values or metabolic parameter values of the anatomical structures in the i-th sample,

[0015] and For all x i and all y i The average value of

[0016] w i is the weight of age and gender of the i-th sample;

[0017] β0 is the intercept term, which is calculated by the formula shown in formula (4):

[0018]

[0019] This helps to statistically obtain a standard linear regression model for the relative SUV values or metabolic parameter values of a pair of anatomical structures.

[0020] In another exemplary embodiment of the metabolic analysis method of an organism, the weight w i Calculated by the formula shown in formula (5):

[0021]

[0022] Where n(age group, gender) is the number of samples in that age group and gender among all samples. This helps eliminate demographic bias.

[0023] In another exemplary embodiment of the method for analyzing the metabolism of an organism, the error of the standard linear regression model is calculated by the formula shown in formula (6):

[0024]

[0025] Where E is the error;

[0026] n is the number of samples, 1≤i≤n;

[0027] e i is the residual of the i-th sample, which is calculated by the formula shown in formula (7):

[0028] e i =y i -(β0+β1x i )Formula (7).

[0029] In another exemplary embodiment of the metabolic analysis method of an organism, in step S10, the studentized residual is calculated using the formula shown in formula (8):

[0030]

[0031] Among them, S r is the studentized residual;

[0032] x′ and y′ are the relative SUV values or metabolic parameter values of the pair of anatomical structures of the target organism, respectively;

[0033] β0 is the intercept term;

[0034] β1 is the regression coefficient;

[0035] E is the error.

[0036] This is useful for calculating the studentized residuals.

[0037] In another exemplary embodiment of the method for analyzing the metabolism of an organism, the P value of the standard linear regression model is less than 0.01, and R 2 The value is greater than 0.2, which helps to obtain a standard linear regression model that is statistically significant and interpretable.

[0038] In another exemplary embodiment of the metabolic analysis method for an organism, step S20 specifically includes: for each pair of anatomical structures to be analyzed, if the studentized residual is greater than 3, then determining that the pair of anatomical structures in the target organism has metabolic abnormalities. This facilitates sensitive identification of potential pathogenic areas.

[0039] In another exemplary embodiment of the metabolic analysis method for an organism, step S30 is further included: generating an analysis report. The analysis report includes a metabolic association map. The metabolic association map identifies all anatomical structures to be analyzed and lines connecting anatomical structures with metabolic abnormalities. This facilitates intuitive identification of potential disease-causing areas.

[0040] In another exemplary embodiment of a method for analyzing metabolic activity in a living organism, the relative SUV value of an anatomical structure is the ratio of the absolute SUV value of the anatomical structure to the absolute SUV value of a reference structure. In the brain, the reference structure is the brainstem, and in the body, the reference structure is the liver. This facilitates eliminating differences in equipment and individual metabolic profiles.

[0041] In another exemplary embodiment of the metabolic analysis method for an organism, the pair of anatomical structures of the target organism are divided from a PET-CT image of the target organism, and the pair of anatomical structures of the sample are divided from a PET-CT image of the sample.

[0042] The present invention also provides a metabolic analysis system comprising a memory and a processor. The memory stores a standard linear regression model. When the processor loads the standard linear regression model, the aforementioned metabolic analysis method can be implemented. This metabolic analysis method is capable of performing the aforementioned metabolic analysis method for an organism, sensitively identifying potential pathogenic foci even with a small sample size. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The following drawings are only used to schematically illustrate and explain the present invention and are not intended to limit the scope of the present invention.

[0044] Figure 1 The present invention is a flow chart illustrating a schematic embodiment of a method for analyzing the metabolism of an organism.

[0045] Figure 2 FIG. 1 is a flow chart illustrating an exemplary embodiment of constructing a standard linear regression model.

[0046] Figure 3 A schematic diagram illustrating an analysis report. DETAILED DESCRIPTION

[0047] In order to have a clearer understanding of the technical features, purposes and effects of the invention, specific embodiments of the present invention are now described with reference to the accompanying drawings, in which the same reference numerals represent the same parts.

[0048] In this document, “illustrative” means “serving as an example, instance or illustration”, and any diagram or implementation described in this document as “illustrative” should not be interpreted as a more preferred or more advantageous technical solution.

[0049] To simplify the drawings, each figure schematically shows only the parts related to the present invention, which do not represent the actual structure of the product.

[0050] Figure 1 FIG. 1 is a flow chart illustrating a schematic embodiment of a method for analyzing the metabolism of an organism. Figure 1 In an exemplary embodiment, a metabolic analysis method for an organism is used to analyze the metabolic state of an anatomical structure of a target organism. The metabolic analysis method includes steps S10 and S20.

[0051] Step S10: For each pair of anatomical structures to be analyzed, the studentized residuals of the relative SUV values or metabolic parameter values of the pair of anatomical structures in the target organism are calculated based on a standard linear regression model of the relative SUV values or metabolic parameter values of the pair of anatomical structures. The pair of anatomical structures belongs to the same brain or body of the organism. Due to the presence of structures such as the blood-brain barrier, the metabolic correlations between brain anatomical structures and body anatomical structures are weak and can be ignored. Therefore, when analyzing the metabolic correlations of a pair of anatomical structures, the pair of anatomical structures belongs to the same brain or body of the organism.

[0052] In an exemplary embodiment, a standard linear regression model is constructed, for example, as Figure 2 As shown, it includes steps S1 to S5.

[0053] Step S1: Segmentation of anatomical structures. The anatomical structures are obtained, for example, by segmenting a sample's PET-CT (positron emission tomography-computer tomography) image. The PET-CT image can be a whole-body image or a localized image.

[0054] In a specific embodiment, 206 anatomical structures are segmented in a whole-body PET-CT image of a sample, of which 83 are located in the brain and 123 are located in the body. The anatomical structures can be segmented using a deep learning model well known to those skilled in the art, or manually delineated.

[0055] Step S2: Calculate the relative SUV value or metabolic parameter value of each anatomical structure. The relative SUV value of an anatomical structure is, for example, the ratio of the absolute SUV value of the anatomical structure to the absolute SUV value of a reference structure. In the brain, the reference structure is the brainstem, and in the body, the reference structure is the liver. The brainstem and liver are selected as reference structures because the absolute SUV values of these two anatomical structures are relatively stable in the organism and are not prone to drastic fluctuations with changes in the physiological state of the organism. Compared with the absolute SUV value, the relative SUV value is beneficial in eliminating errors caused by measuring equipment and errors caused by different individual metabolism.

[0056] The absolute SUV value of the anatomical structure is calculated using the formula shown in formula (1):

[0057]

[0058] The tissue radioactivity concentration is the radioactivity activity in the area where the lesion is located, which is calculated using the grayscale value of the PET-CT image and is expressed in kBq / ml. The injection dose is the total activity of the imaging agent injected into the subject, which is recorded before the PET-CT image is taken and is expressed in MBq. The subject's weight is the subject's actual weight and is expressed in kg.

[0059] The metabolic parameter values of the anatomical structure can be the fitting parameters (K1, k2, ..., k n ,f v ,K i ), calculated by the gray value of PET-CT images, the unit is ml / min / ml or min -1 .

[0060] Step S3: Construct a standard linear regression model based on the relative SUV values or metabolic parameter values of any pair of anatomical structures. For example, in the aforementioned specific embodiment, 83 anatomical structures are located in the brain and 123 anatomical structures are located in the body. Thus, it can be obtained A standard linear regression model is constructed for each pair of anatomical structures and for the relative SUV values or metabolic parameter values of each pair of anatomical structures. In an exemplary embodiment, the standard linear regression model is represented by the formula shown in formula (2):

[0061] y=β0+β1x Formula (2)

[0062] Wherein, x and y are the relative SUV values or metabolic parameter values of the pair of anatomical structures, respectively, and the pair of anatomical structures belong to the brain or the body of the organism;

[0063] β1 is the regression coefficient, which is calculated by the formula shown in formula (3):

[0064]

[0065] Among them, n is the number of samples, 1≤i≤n, and the samples are healthy organisms. The standard linear regression model constructed by this is of reference value.

[0066] x i and y i are the relative SUV values or metabolic parameter values of the anatomical structures in the i-th sample,

[0067] and For all x i and all y i The average value of

[0068] w i is the weight of age and gender of the i-th sample;

[0069] β0 is the intercept term, which is calculated, for example, by the formula shown in formula (4):

[0070]

[0071] For weight w i , which performs logarithmic smoothing based on the age and gender distribution of the subjects to eliminate demographic bias, and the weight w i For example, the formula (5) is used to calculate:

[0072]

[0073] Among them, n(age group, gender) is the number of samples in this age group and gender among all samples.

[0074] In other exemplary embodiments, the weight w i It can also be processed by reciprocal smoothing according to the age and gender distribution of the subjects, which is done by the formula Calculation is performed, where n(age group, gender) represents the number of samples in the sample set that are in that age group and of that gender.

[0075] In an exemplary embodiment, the error of the standard linear regression model is calculated, for example, by the formula shown in equation (6):

[0076]

[0077] Where E is the error;

[0078] n is the number of samples, 1≤i≤n;

[0079] e iis the residual of the i-th sample, where the residual is calculated by the formula shown in formula (7):

[0080] e i =y i -(β0+β1x i )Formula (7).

[0081] Step S4: Screening a standard linear regression model that meets both the significance and interpretability requirements. For example, the standard linear regression model has an F test, and its P value is less than 0.01, and R 2 For example, the value is greater than 0.2. And those that do not meet the P value requirements and R 2 If the value does not meet the standard linear regression model required, it means that the metabolic relationship between the corresponding pair of anatomical structures does not conform to the linear relationship and is removed in the screening process. This helps to obtain a standard linear regression model with statistical significance and interpretability.

[0082] Step S5: Storing the parameters of each standard linear regression model that meets the requirements. The parameters of each standard linear regression model include, for example, the regression coefficient β1, the intercept term β0, and the error E. As previously mentioned, the standard linear regression model is independently constructed for different pairs of anatomical structures. Therefore, the parameters of each standard linear regression model are different.

[0083] In step S10, the target organism is an organism to be analyzed, such as a human body. The method for dividing its anatomical structure can refer to step S1, and the method for calculating the relative SUV value or metabolic parameter value of its anatomical structure can refer to step S2. After each standard linear regression model retrieves the relative SUV value or metabolic parameter value of its corresponding anatomical structure to be analyzed, the studentized residual of the relative SUV value or metabolic parameter value of the anatomical structure in the standard linear regression model is calculated. The studentized residual is calculated using the formula shown in formula (8):

[0084]

[0085] Among them, S r is the studentized residual;

[0086] x′ and y′ are the relative SUV values or metabolic parameter values of the pair of anatomical structures of the target organism, respectively;

[0087] β0 is the intercept term;

[0088] β1 is the regression coefficient;

[0089] E is the error.

[0090] Step S20: For each pair of anatomical structures to be analyzed, determine whether metabolic abnormalities exist in the target organism's anatomical structures based on the studentized residual. In an exemplary embodiment, step S20 specifically includes: for each pair of anatomical structures to be analyzed, if the studentized residual is greater than 3, then determine that metabolic abnormalities exist in the target organism's anatomical structures. This facilitates sensitive identification of potential pathogenic areas.

[0091] Based on a standard linear regression model of relative SUV values or metabolic parameter values for a pair of anatomical structures, the studentized residuals of the relative SUV values or metabolic parameter values of a pair of anatomical structures of the target organism are calculated to determine whether metabolic abnormalities exist in these anatomical structures. This metabolic analysis method requires a small sample size and can sensitively identify potential pathogenic areas.

[0092] In an exemplary embodiment, see Figure 1 The method for metabolic analysis of an organism further comprises step S30: generating an analysis report, the analysis report including a metabolic association map, and an application diagram of the metabolic association map is shown in FIG. Figure 3 shown. Figure 3 Figures a and b show the metabolic association maps of Alzheimer's (AD) in the brain and lung cancer in the body, respectively. Figure 3 In a, the left side is the metabolic association map of the patient in the mild cognitive impairment (MCI) stage, which is the pre-stage of Alzheimer's disease, and the right side is the metabolic association map of the patient who has already suffered from Alzheimer's disease. Figure 3 Figure b shows the metabolic association maps of the third and fourth stages of lung cancer. Figure 3 In b, the left side is the metabolic association map of the patient in the third stage of lung cancer, and the right side is the metabolic association map of the patient in the fourth stage of lung cancer. Figure 3 The a and b of the diagram clearly mark all the anatomical structures to be analyzed and the lines connecting the anatomical structures with metabolic abnormalities. Figure 3 Green, red, and blue dots are shown in a and b. Green dots represent the central nodes of anatomical structures with normal metabolism, while red and blue dots represent the central nodes of anatomical structures with abnormal metabolism. Red dots represent central nodes with a high number of abnormal metabolic connections (e.g., more than five abnormal metabolic connections), while blue dots represent central nodes with a low number of abnormal metabolic connections (e.g., fewer than five abnormal metabolic connections). Each pair of anatomical structures is connected by a red line.

[0093] By comparison Figure 3 The left and right metabolic association maps of a show that when patients progress from mild cognitive impairment to Alzheimer's disease, the number of anatomical structures with metabolic abnormalities in their brains increases, and the connections between these anatomical structures become denser. Figure 3The left and right metabolic association maps of b show that as patients progress from stage 3 to stage 4 lung cancer, the number of metabolically abnormal anatomical structures increases, the connections between these structures become denser, and the abnormalities tend to spread distally. This facilitates intuitive identification of potential lesions.

[0094] The present invention also provides a metabolic analysis system comprising a memory and a processor. The memory stores a standard linear regression model. When the processor loads the standard linear regression model, the aforementioned metabolic analysis method can be implemented. This metabolic analysis method can perform the aforementioned metabolic analysis method for an organism, sensitively identifying potential pathogenic foci even with a small sample size.

[0095] It should be understood that although this specification is described according to various embodiments, not every embodiment contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

[0096] The series of detailed descriptions listed above are only specific descriptions of feasible embodiments of the present invention. They are not intended to limit the scope of protection of the present invention. Any equivalent implementation scheme or changes that do not deviate from the technical spirit of the present invention, such as the combination, division or repetition of features, should be included in the scope of protection of the present invention.

Claims

1. A method for analyzing the metabolic state of an anatomical structure of a target organism, characterized in that: The metabolic analysis method comprises: Step S10: for each pair of anatomical structures to be analyzed, calculating, based on a standard linear regression model of the relative SUV values or metabolic parameter values of a pair of anatomical structures, the studentized residuals of the relative SUV values or metabolic parameter values of the pair of anatomical structures of the target organism in the standard linear regression model, wherein the pair of anatomical structures belong to the same brain or body of the organism; and Step S20: for each pair of anatomical structures to be analyzed, determining whether metabolic abnormalities exist in the pair of anatomical structures of the target organism according to the studentized residuals.

2. The method for metabolic analysis of an organism according to claim 1, wherein The standard linear regression model is represented by the formula shown in formula (2): y=β0+β1x Formula (2) Where x and y are the relative SUV values or metabolic parameter values of the pair of anatomical structures, respectively; β1 is the regression coefficient, which is calculated by the formula shown in formula (3): Wherein, n is the number of samples, 1≤i≤n, and the samples are healthy organisms. x i and y i are the relative SUV values or metabolic parameter values of the anatomical structures in the i-th sample, and For all x i and all y i The average value of w i is the weight of age and gender of the i-th sample; β0 is the intercept term, which is calculated by the formula shown in formula (4):

3. The method for metabolic analysis of an organism according to claim 2, wherein: The weight w i Calculated by the formula shown in formula (5): Among them, n(age group, gender) is the number of samples in this age group and gender among all samples.

4. The method for metabolic analysis of an organism according to claim 2, wherein: The error of the standard linear regression model is calculated by the formula shown in formula (6): Wherein, E is the error; n is the number of samples, 1≤i≤n; e i is the residual of the i-th sample, which is calculated by the formula shown in formula (7): e i = y i -(β0 + β1x i ) Equation (7).

5. The method for metabolic analysis of an organism according to claim 4, wherein: In step S10, the studentized residual is calculated using the formula shown in equation (8): Among them, S r is the studentized residual; x′ and y′ are the relative SUV values or metabolic parameter values of the pair of anatomical structures of the target organism, respectively; β0 is the intercept term; β1 is the regression coefficient; E is the error.

6. The method for metabolic analysis of an organism according to claim 1, wherein The P value of the standard linear regression model was less than 0.01, R 2 The value is greater than 0.

2.

7. The method for metabolic analysis of an organism according to claim 1, wherein Step S20 specifically includes: for each pair of anatomical structures to be analyzed, if the studentized residual is greater than 3, determining that metabolic abnormalities exist in the pair of anatomical structures of the target organism.

8. The method for metabolic analysis of an organism according to claim 1, wherein The method further includes step S30: generating an analysis report, wherein the analysis report includes a metabolic association map, and the metabolic association map marks the connection lines between all anatomical structures to be analyzed and anatomical structures with metabolic abnormalities.

9. The method for metabolic analysis of an organism according to claim 1, wherein The relative SUV value of the anatomical structure is the ratio of the absolute SUV value of the anatomical structure to the absolute SUV value of a reference structure. In the brain, the reference structure is the brainstem, and in the body, the reference structure is the liver.

10. The method for metabolic analysis of an organism according to claim 2, wherein: The pair of anatomical structures of the target organism are divided from a PET-CT image of the target organism, and the pair of anatomical structures of the sample are divided from a PET-CT image of the sample.

11. A metabolic analysis system, characterized in that The method comprises a memory and a processor, wherein the memory stores a standard linear regression model, and when the processor loads the standard linear regression model, the metabolic analysis method according to any one of claims 1 to 10 can be implemented.

Images