Prediction method, device, system and readable storage medium based on visceral fat tissue area

Through segmentation algorithm and machine learning model, the area of ​​visceral adipose tissue was extracted from the image, and the problem of poor diagnosis of early pelvic adipose lipidemia was solved, and the accurate prediction of pelvic fat volume and efficient diagnosis of PL were achieved.

CN116797592BActive Publication Date: 2024-11-08PEKING UNION MEDICAL COLLEGE HOSPITAL
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310816522.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-07-05
Publication Date
2024-11-08
Estimated Expiration
2043-07-05

AI Technical Summary

Technical Problem

The prior art has poor accuracy in the early diagnosis of pelvic fatemia (PL), and the measurement of pelvic fat volume (PFV) is complex, making it difficult to achieve accurate prediction.

Method used

The visceral adipose tissue area is segmented from the image images using a segmentation algorithm, and based on this, machine learning or deep learning model is constructed to predict the presence of pelvic fat volume and PL, simplifying the diagnostic process.

Benefits of technology

It improves the accurate prediction of pelvic fat volume and the accuracy of early diagnosis of PL, reduces the difficulty of diagnosis, and is suitable for large-scale clinical applications.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116797592B_ABST
    Figure CN116797592B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of intelligent medical technology, and more specifically, to a prediction method, device, system and readable storage medium based on visceral fat tissue area. The method comprises: obtaining an image of a sample, wherein the image includes a pelvic image; segmenting the image using a segmentation algorithm to obtain the visceral fat tissue area of ​​the sample; and predicting the pelvic fat volume of the sample or the classification result of whether the sample is pelvic lipomatosis based on the visceral fat tissue area. The present application has important clinical application value.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of intelligent medical technology, and more specifically, to a prediction method, device, system and readable storage medium based on visceral fat tissue area. Background Art

[0002] Pelvic lipomatosis (PL) is a rare benign disease with an unknown etiology, characterized by excessive growth of adipose tissue around the bladder and rectum. Currently, the clinical diagnosis of PL mainly relies on clinical symptoms and imaging manifestations, but the diagnostic accuracy is poor and the detection is difficult for PL patients who have no typical morphological changes in early pelvic organs. Pelvic fat volume (PFV) is an objective and intuitive indicator reflecting the content of pelvic adipose tissue with high sensitivity, especially for early patients, but the actual measurement is more complicated. Summary of the invention

[0003] To solve the above problems, we segment the patient's image based on a segmentation algorithm to obtain the patient's visceral fat tissue area, predict PFV and PL based on the patient's visceral fat tissue area, and propose a method, device, system and storage medium for predicting PFV and PL based on visceral fat tissue area.

[0004] The present application discloses a prediction method based on visceral fat tissue area, the method steps comprising:

[0005] S101: Acquire an image of a sample, wherein the image includes a pelvic image;

[0006] S102: Segmenting the image using a segmentation algorithm to obtain the visceral fat tissue area of ​​the sample;

[0007] S103: predicting the pelvic fat volume of the sample based on the visceral fat tissue area;

[0008] Optionally, the pelvic image includes any one or more of the following levels: S1, S2, S3, S4, FH;

[0009] Optionally, a segmentation algorithm is used to segment the planes S1, S2, S3, S4 or FH in the pelvic image to obtain the S1, S2, S3, S4 or FH visceral fat tissue area of ​​the sample, and the pelvic fat volume of the sample is predicted based on the S1, S2, S3, S4 or FH visceral tissue area.

[0010] Furthermore, the image also includes an abdominal image; the image is segmented using a segmentation algorithm to obtain the visceral fat tissue area of ​​the abdominal cavity of the sample; the pelvic fat volume of the sample is predicted based on the visceral fat tissue area of ​​the abdominal cavity of the sample;

[0011] Optionally, the abdominal cavity image includes any one or more of the following levels: L -94 -L -100 ;

[0012] Optionally, a segmentation algorithm is used to segment the abdominal image in the middle plane L -94 , L -95 , L -96 , L -97 , L -98 , L -99 or L -100 , and obtain the sample L -94 , L -95 , L -96 , L -97 , L -98 , L -99 or L -100 Visceral adipose tissue area, based on the L -94 , L -95 , L -96 , L -97 , L -98 , L -99 or L -100 Visceral adipose tissue area predicted pelvic fat volume in the sample;

[0013] Optionally, a segmentation algorithm is used to segment the pelvic image and the abdominal image to obtain the pelvic visceral fat tissue area and the abdominal visceral fat tissue area of ​​the sample; and the pelvic fat volume of the sample is predicted based on the pelvic visceral fat tissue area and the abdominal visceral fat tissue area of ​​the sample.

[0014] Furthermore, step S103 of the method is to input the visceral fat tissue area into the constructed model and output the predicted pelvic fat volume of the sample; preferably, a regression model between the visceral fat area and the pelvic fat volume is constructed based on the training set to obtain the constructed model; more preferably, the model is a machine learning algorithm or a deep learning algorithm model.

[0015] Further, step S103 of the method may be replaced by: predicting whether the sample is a classification result of pelvic lipomatosis based on the visceral fat tissue area;

[0016] Optionally, the pelvic image includes any one or more of the following levels: S1, S2, S3, S4, FH;

[0017] Optionally, a segmentation algorithm is used to segment the plane S1, S2, S3 or S4 in the pelvic image to obtain the S1, S2, S3 or S4 visceral fat tissue area of ​​the sample, and a classification result of whether the sample is pelvic lipomatosis is predicted based on the S1, S2, S3 or S4 visceral fat tissue area;

[0018] Optionally, a segmentation algorithm is used to segment the FH in the middle plane of the pelvic image to obtain the FH visceral fat tissue area of ​​the sample, and a classification result of whether the sample is pelvic lipomatosis is predicted based on the FH visceral fat tissue area;

[0019] Optionally, a segmentation algorithm is used to segment the planes S4 and FH in the pelvic image to obtain the S4 and FH visceral fat tissue areas of the sample, and a classification result of whether the sample is pelvic lipomatosis is predicted based on the S4 and FH visceral fat tissue areas.

[0020] Further, obtaining the pelvic fat volume of the sample, and predicting whether the sample is classified as pelvic lipomatosis based on the pelvic fat volume and visceral fat tissue area;

[0021] Optionally, the classification result of whether the sample is pelvic lipomatosis is predicted based on the pelvic fat volume and the S4 and FH visceral fat tissue areas; preferably, the pelvic fat volume is a directly measured pelvic fat volume or a pelvic fat volume of the sample predicted based on the visceral fat tissue area;

[0022] Optionally, the pelvic fat volume is the pelvic fat volume directly measured using MRIcron or Syngo Volume;

[0023] Optionally, the visceral fat tissue area predicts the pelvic fat volume of the sample by using the above-mentioned method to predict the pelvic fat volume of the sample.

[0024] Furthermore, step S103 of the method is to input the visceral fat tissue area into the constructed model, and output a classification result of predicting whether the sample is pelvic lipomatosis; preferably, a regression model between the visceral fat area and pelvic lipomatosis is constructed based on the training set to obtain the constructed model; more preferably, the model is a machine learning algorithm or a deep learning algorithm model.

[0025] Furthermore, the segmentation network includes any one of the following: U-net (a neural network that can perform image segmentation on two-dimensional images), PSP-net (Pyramid Scene Parsing Network), FCN (Fully Convolutional Network), Res Net (Residual Neural Network), Inception (stacking convolutional features and pooling features of various sizes such as 1x1 and 3x3 together) or a segmentation network optimized based on the above architecture. Preferably, the segmentation network is a nnU-Net segmentation network.

[0026] The purpose of the present application is to provide a prediction device based on visceral fat tissue area, comprising: a memory and a processor;

[0027] The memory is used to store program instructions;

[0028] The processor is used to call program instructions, and when the program instructions are executed, the above-mentioned prediction method steps based on visceral fat tissue area are implemented.

[0029] The purpose of the present application is to provide a computer program product having a computer program stored thereon, which implements the above-mentioned steps of the prediction method based on visceral fat tissue area when the computer program is executed by a processor.

[0030] The purpose of the present application is to provide a computer-readable storage medium having a computer program stored thereon, which implements the above-mentioned steps of the prediction method based on visceral fat tissue area when executed by a processor.

[0031] Advantages of this application:

[0032] 1. In view of the current problem that PFV measurement is difficult, time-consuming and labor-intensive, the present application provides a new method for predicting PFV, that is, segmenting pelvic images based on a segmentation algorithm to obtain the patient's visceral fat tissue area, and constructing a regression model between the visceral fat area and PFV based on a training set to predict the patient's PFV;

[0033] 2. In order to further reduce the difficulty and facilitate clinical application, this application selected 5 levels (S1, S2, S3, S4, FH) in the pelvic image, and 201 levels (L -100 —L+ 100 ) and selected L -94 -L -100 For subsequent predictions;

[0034] 3. In view of the problem of poor accuracy in detecting early PL patients without typical morphological changes, the present application provides a new method for predicting PL, namely, segmenting pelvic images based on a segmentation algorithm to obtain the patient's visceral fat tissue area, and constructing the visceral fat tissue area and PL based on a training set.

[0035] The regression model between predicted the risk of PL in patients;

[0036] 4. The research results based on this application show that if a single-cut image is used, the visceral fat tissue area can be directly obtained based on the segmentation in the S4 and / or FH levels. The model constructed based on the aforementioned training set can be directly used for the early prediction of PFV and PL. It is simple and convenient, suitable for large-scale clinical promotion and application, and has very good clinical application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0038] Figure 1 is a schematic flow chart of a prediction method based on visceral fat tissue area provided by an embodiment of the present invention;

[0039] Figure 2 Schematic diagram of different anatomical position levels of pelvic and abdominal images provided by an embodiment of the present invention. (A) Taking the L3 vertebra as the zero point (L0), a total of 201 levels are selected with an interval of 1 mm (L -100 -L +100 ); (B) Five levels of the pelvis were selected, including four levels from the anterior superior edge of the S1 vertebra to the anterior superior edge of the S4 vertebra and the level passing through the upper edge of the bilateral femoral heads (FH); (C) The VAT and SAT areas at the umbilicus level of the patient obtained by automatic segmentation, the red area represents SAT, and the green area represents VAT; (D) Three-dimensional software was used to calculate the patient's PFV, the green area represents pelvic fat, and the deformed bladder can be observed; (E) The body width was measured at the level of the right renal hilum, which is equal to the left-right diameter (AB) plus the anteroposterior diameter (CD) of the body. The yellow arrow shows bilateral hydronephrosis;

[0040] Figure 3 is a schematic diagram of a prediction device based on visceral fat tissue area provided by an embodiment of the present invention;

[0041] Figure 4 is a linear regression model of VAT area at different anatomical locations and PFVs provided by an embodiment of the present invention; Figure 5 : is a correspondence diagram between an actual PFV and a predicted PFV provided by an embodiment of the present invention; (A) a Bland-Altman diagram is drawn to show the difference between the actual PFV and the predicted PFV at the FH level, the dotted line represents the upper and lower limits of the 95% CI, and the black solid line represents the deviation; (B) the standard deviation S of the percentage difference between the predicted PFV and the actual PFV of the visceral fat area at a single level at different anatomical locations d %;

[0042] Figure 6: is a ROC curve diagram of a logistic regression model provided by an embodiment of the present invention at various anatomical locations; the combined model including S4 and FH showed a higher diagnostic efficiency (AUC=0.927), when Y was greater than 0.462, it was diagnosed as PL, and the overall accuracy of the logistic regression model in predicting PL was 88%;

[0043] Figure 7 It is a schematic diagram of a prediction system based on visceral fat tissue area provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0044] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0045] In some of the processes described in the specification and claims of the present invention and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0046] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0047] Figure 1 : is a schematic flow chart of a prediction method based on visceral fat tissue area provided by an embodiment of the present invention, the method steps comprising:

[0048] S101: Acquire an image of a sample, wherein the image includes a pelvic image;

[0049] S102: Segmenting the image using a segmentation algorithm to obtain the visceral fat tissue area of ​​the sample;

[0050] S103: Predicting the pelvic fat volume of the sample based on the visceral fat tissue area.

[0051] In one embodiment, the pelvic image includes any one or more of the following layers: S1, S2, S3, S4, FH, and the layer S1, S2, S3, S4 or FH in the pelvic image is segmented by a segmentation algorithm to obtain the S1, S2, S3, S4 or FH visceral fat tissue area of ​​the sample, and the pelvic fat volume of the sample is predicted based on the S1, S2, S3, S4 or FH visceral fat tissue area. The visceral fat tissue area (VAT) refers to the area of ​​abdominal visceral fat in the CT image.

[0052] The pelvic fat volume (PFV) is an objective indicator reflecting the content of pelvic fat tissue, and is usually a quantitative indicator calculated by 3D modeling software and has good repeatability.

[0053] In one embodiment, the image is an abdominal image; the abdominal image includes any one or more of the following levels: L -94 -L -100 , using a segmentation algorithm to segment the image, and obtain the visceral fat tissue area of ​​the abdominal cavity of the sample; based on the visceral fat tissue area of ​​the abdominal cavity of the sample, the pelvic fat volume of the sample is predicted. Specifically, using a segmentation algorithm to segment the abdominal cavity image in the middle plane L -94 , L -95 , L -96 , L -97 , L -98 , L -99 or L -100 , and obtain the sample L -94 , L -95 , L -96 , L -97 , L -98 , L -99 or L -100 Visceral adipose tissue area, based on the L -94 , L -95 , L -96 , L -97 , L -98 , L -99 or L -100 Visceral adipose tissue area predicted pelvic fat volume in the samples.

[0054] In a specific embodiment, the pelvic image and / or abdominal image is a CT image.

[0055] In one embodiment, a segmentation algorithm is used to segment the pelvic image and the abdominal image to obtain the pelvic visceral fat tissue area and the abdominal visceral fat tissue area of ​​the sample; the pelvic fat volume of the sample is predicted based on the pelvic visceral fat tissue area and the abdominal visceral fat tissue area of ​​the sample. Specifically, the segmentation algorithm is used to segment the plane S1, S2, S3, S4 or FH in the pelvic image to obtain the S1, S2, S3, S4 or FH visceral fat tissue area of ​​the sample, and the segmentation algorithm is used to segment the plane L in the abdominal image -94 , L -95 , L -96 , L -97 , L -98 , L -99 or L -100 , and obtain the sample L -94 , L -95 , L -96 , L -97 , L -98 , L -99 or L -100 Visceral adipose tissue area, based on S1, S2, S3, S4, FH, L -94 , L -95 , L -96 , L -97 , L -98 , L -99 and / or L -100 The visceral fat tissue area predicts the pelvic fat volume of the sample. More specifically, the visceral fat tissue area is input into a constructed model, and the pelvic fat volume of the predicted sample is output; preferably, a regression model between the visceral fat area and the pelvic fat volume is constructed based on a training set to obtain a constructed model; more preferably, the model is a machine learning algorithm or a deep learning algorithm model.

[0056] In one embodiment, the model construction method includes: obtaining training set data and an initial regression model; training the initial regression model using an optimization algorithm based on the training set data; when the loss function value of the regression model reaches a preset threshold, the regression model training is completed. In a specific embodiment, S1, S2, S3, S4, FH, L in the training set data are obtained. -94 , L -95 , L -96 , L -97 , L -98 , L -99 and / or L -100 The visceral tissue area and the initial regression model are trained, the pelvic fat volume of the output predicted sample is compared with the actual pelvic fat volume, and the loss function is optimized. When the loss function value of the regression model reaches the preset threshold, the model training is completed.

[0057] An embodiment of the present invention provides a schematic flow chart of a method for predicting visceral fat tissue area. Specifically, the method comprises the following steps:

[0058] S101: Acquire an image of a sample, wherein the image includes a pelvic image;

[0059] S102: Segmenting the image using a segmentation algorithm to obtain the visceral fat tissue area of ​​the sample;

[0060] S103: Predicting whether the sample is classified as pelvic lipomatosis based on the visceral tissue area.

[0061] In one embodiment, the pelvic image includes any one or more of the following layers: S1, S2, S3, S4, FH, and a segmentation algorithm is used to segment the layers S1, S2, S3 or S4 in the pelvic image to obtain the S1, S2, S3 or S4 visceral fat tissue area of ​​the sample, and the classification result of whether the sample is pelvic lipomatosis is predicted based on the S1, S2, S3 or S4 visceral fat tissue area.

[0062] In one embodiment, a segmentation algorithm is used to segment the FH layer in the pelvic image to obtain the FH visceral fat tissue area of ​​the sample, and the classification result of whether the sample is pelvic lipomatosis is predicted based on the FH visceral fat tissue area. In a specific embodiment, a segmentation algorithm is used to segment the layers S4 and FH in the pelvic image to obtain the S4 and FH visceral fat tissue areas of the sample, and the classification result of whether the sample is pelvic lipomatosis is predicted based on the S4 and FH visceral fat tissue areas.

[0063] In one embodiment, the pelvic fat volume of the sample is obtained, and the classification result of whether the sample is pelvic lipomatosis is predicted based on the pelvic fat volume and the visceral tissue area. Specifically, the pelvic fat volume is the pelvic fat volume directly measured by MRIcron or Syngo Volume.

[0064] In one embodiment, the classification result of whether the sample is pelvic lipomatosis is predicted based on the pelvic fat volume and the S4 and FH visceral fat tissue areas; preferably, the pelvic fat volume is a directly measured pelvic fat volume or the pelvic fat volume of the sample is predicted based on the visceral fat tissue area.

[0065] In a specific embodiment, a pelvic image of the sample is obtained, and the pelvic image includes any one or more of the following levels: S1, S2, S3, S4, and FH. The pelvic image is segmented using a segmentation algorithm to obtain the pelvic visceral fat tissue area of ​​the sample. The pelvic fat volume of the sample is predicted based on the pelvic visceral fat tissue area. Based on the pelvic fat volume and the S4 and FH visceral fat tissue areas, it is predicted whether the sample has pelvic lipomatosis.

[0066] In one embodiment, the visceral fat tissue area is input into the constructed model, and the classification result of whether the sample is pelvic lipomatosis is output; preferably, a regression model between the visceral fat area and pelvic lipomatosis is constructed based on the training set to obtain the constructed model; more preferably, the model is a machine learning algorithm or a deep learning algorithm model. Exemplarily, S1, S2, S3, S4, FH, L in the training set data are obtained. -94 , L -95 , L -96 , L -97 , L -98 , L -99 and / or L -100 The visceral tissue area and the initial regression model are trained, the predicted classification results of the output prediction samples are compared with the actual classification results, and the loss function is optimized. When the loss function value of the regression model reaches the preset threshold, the model training is completed.

[0067] In one embodiment, the segmentation network includes any one of the following: U-net (a neural network that can perform image segmentation on two-dimensional images), PSP-net (Pyramid Scene Parsing Network), FCN (Fully Convolutional Network), Res Net (Residual Neural Network), Inception (stacking convolutional features and pooling features of various sizes such as 1x1 and 3x3 together) or a segmentation network optimized based on the above architecture. Preferably, the segmentation network is a nnU-Net segmentation network.

[0068] In one embodiment, the segmentation network is a full convolutional network; the full convolutional network includes a pooling layer and a convolutional layer; the training sample image is input into the segmentation network to obtain a segmentation map corresponding to the training sample image, including: inputting the training sample image into the full convolutional network, performing average pooling processing on the training sample image through the pooling layer of the full convolutional network; performing convolution processing on the image obtained by the pooling processing through the convolutional layer of the full convolutional network to obtain a segmentation map corresponding to the training sample image. By obtaining the visceral fat tissue area label of the sample corresponding to the object entity in the training sample image and the training sample image, the training sample image is input into the segmentation network to obtain a segmentation map corresponding to the training sample image, and the image of the visceral fat tissue area of ​​the sample corresponding to the object entity is intercepted from the training sample image according to the segmentation map, and the image is input into the trained segmentation network to obtain the corresponding predicted segmentation result, which is compared with the actual segmentation result, and the loss function is calculated. According to the loss function, the network parameters of the segmentation network are adjusted until the training stop condition is met to obtain a trained segmentation network. Among them, the type of loss function can adopt common loss functions, such as cross-entropy loss function, absolute value loss function, log loss function, exponential loss function, etc.

[0069] In a specific embodiment, a pelvic image of the sample is obtained, and the pelvic image includes any one or more of the following levels: S4, FH. The image is segmented using the nnU-Net segmentation network to obtain the S4 and / or FH visceral fat tissue area of ​​the sample, and the pelvic fat volume of the sample is predicted based on the S4 and / or FH visceral fat tissue area, or whether the sample is classified as pelvic lipomatosis based on the S4 and / or FH visceral fat tissue area.

[0070] In a specific embodiment, patients who underwent CTU examination in our hospital from September 2014 to December 2019 were retrospectively included. The inclusion criteria for patients in the PL group were as follows: (1) CT imaging diagnosis of "pelvic lipomatosis"; (2) clinical diagnosis including "pelvic lipomatosis". Fifty consecutive patients were obtained. Among the 50 PL patients, 5 cases were diagnosed with pelvic lipomatosis by laparoscopic pathology, and 23 cases were diagnosed with glandular cystitis after transurethral resection of bladder tumor (TURBT). Fifty controls matched with the PL group in both age and sex were selected according to the following criteria: (1) CTU examination was performed, and (2) normal or diseased but without effect on adipose tissue (see Table 1).

[0071] Table 1 Demographic characteristics of the study population

[0072]

[0073]

[0074] CT acquisition: All patients underwent CTU examination using Somatom Definition Flash (Siemens Healthcare, Forchheim, Germany). The patient was placed in the supine position, and the scanning range was from above the diaphragm to the pubic symphysis. The abdomen was scanned in plain scan, renal cortical phase, renal parenchymal phase, and excretory phase. The CT parameters were as follows: slice thickness 1 mm, pitch 0.9, collimation 128 × 0.6 mm, rotation time 0.28 s, kernel filter B30f (medium smoothing), tube voltage 120 kV, and real-time dynamic exposure dose adjustment Care4D. For contrast-enhanced scanning, 90 ml of non-ionic contrast agent (Ultravist370, Bayer Schering Pharma) was injected at a rate of 4 ml / s using a double-barreled high-pressure syringe, followed by 100 ml of normal saline. The ROI was located at the junction of the descending aorta and the abdominal aorta, and the automatic trigger threshold was set to 100 HU. The renal parenchymal phase scan was started 7 seconds after reaching the threshold, and the excretory phase scan was performed 55 seconds later.

[0075] Adipose tissue area measurement: Before analysis, all patient information was deleted from the CT images, and thin-section renal parenchymal images were collected for all patients and controls. A total of 201 slices were selected with the L3 vertebra as the zero point (L0) at an interval of 1 mm (L -100 -L +100 ), covering the upper edge of the L1 vertebra and the L5 / S1 intervertebral space. Five levels were selected for the pelvis, including four levels from the anterior upper edge of the S1 vertebra to the anterior upper edge of the S4 vertebra and a level passing through the upper edge of the bilateral femoral head (FH) ( Figure 2 A-2B). We used the nnU-Net architecture to automatically segment SAT and VAT at each level (9). The areas of SAT and VAT were measured at each level in patients and controls ( Figure 2 C).

[0076] Pelvic fat volume measurement: Pelvic fat volume was measured in 50 patients and 50 controls using MRIcron (macOS version 1.0.20190902) and SyngoVolume (Siemens), respectively. The measurement range was from the anterior superior edge of the first sacral vertebra to the pelvic floor (CT value between -150HU and -30HU) ( Figure 2 D).

[0077] Statistical analysis: All data were analyzed by IBM SPSS Statistics Version 26.0 (IBM, SPSS; Chicago, IL, USA). Two independent sample t-tests or rank-sum tests were used to compare the statistical differences in area and volume between patients and controls, and P values ​​< 0.05 were considered statistically significant. The intraclass correlation coefficient of PFV was used to show the consistency between groups. Pearson correlation was used to calculate the correlation between adipose tissue area and PFV at each level. A linear regression model for estimating PFV using adipose tissue area was established, and a binary logistic regression model was developed to show the accuracy of adipose tissue area at different anatomical locations in predicting PL. The receiver operating characteristic (ROC) curve of the logistic regression model was plotted to reflect the sensitivity (SEN) and specificity (SPE). The area under the ROC curve (AUC) value represents the diagnostic efficiency of the model.

[0078] Abdominal SAT and VAT areas: Obtained the SAT and VAT areas of each abdominal plane for each patient and control. The VAT areas of 29 of the 201 abdominal planes were statistically different between patients and controls. -94 -L -100 (9.4cm to 10cm below the L3 vertebra) and L +79 -L +100 (7.9cm to 10cm above the L3 vertebra) (P<0.05) (Table 2). There was no statistically significant difference in the SAT area at each level of the abdomen. The five levels measured in the pelvis showed statistical differences (P<0.05). Table 2 shows the mean and standard deviation of the VAT area at each level of the abdomen and pelvis in the two groups.

[0079] Pelvic fat volume: The average PFV of patients calculated by the two softwares was 924.60 cm 3 and 924.72cm 3 The average PFV of the control group calculated by the two softwares was 621.15 cm 3 and 621.33cm 3 The intraclass correlation coefficient of PFV was 1.000 (P < 0.001), indicating that the intergroup consistency of PFV was excellent. On average, the mean PFV of the patients was 924.66 cm 3 , the standard deviation is 235.95cm 3 , while the mean and standard deviation of PFV in the control group were 621.24 cm 3 and 248.21cm 3 Two independent sample t tests showed that the difference in PFV between the patient group and the control group was statistically significant (P < 0.001) (Table 2).

[0080] Table 2

[0081]

[0082]

[0083] The patients were rematched with the control group by BMI, and the differences in SAT area, VAT area, and PFV between the two groups were compared by paired t-test or Wilcoxon rank sum test. There were statistical differences in PFV and VAT area at 5 pelvic levels (P<0.05), and there was no statistical difference in SAT area between the patients and the control group, which was consistent with the results before BMI matching. There were statistical differences in VAT area between the patients and the control group in 28 of the 201 abdominal levels—L-81-L -100 (8.1cm to 10cm below the L3 vertebra) and L +93 -L +100 (9.3cm to 10cm above the L3 vertebra).

[0084] Correlation between VAT and PFV and linear regression model: The pelvic VAT area was strongly correlated with PFV (correlation coefficient range: 0.624-0.745), while the abdominal VAT area was poorly correlated with PFV (maximum correlation coefficient: 0.596). The VAT area at the FH level was best correlated with PFV (r = 0.745). The linear regression model used the VAT areas at seven abdominal levels and five pelvic levels as independent variables and PFV as the dependent variable (Table 3, Figure 4 A-4F).

[0085] Table 3

[0086]

[0087] Linear regression model established at the FH level:

[0088] PFV(cm 3 )=288.806+5.925×FH(cm 2 )(R 2 =0.555,P<0.001)

[0089] In addition to the binary regression model, we also developed a multivariate regression model using a linear regression input algorithm with VAT areas at seven abdominal and five pelvic levels as independent variables, and the resulting model only included FH as an independent variable. We plotted the Bland-Altman plot of the equation to compare the difference between the estimated PFV calculated using the VAT area at the FH level and the actual PFV ( Figure 5 A). Figure 5B summarizes the percentage differences between the estimated and actual values ​​of PFV calculated using VAT area at five pelvic levels, with standard deviation S d % = (PFV ~ -PFV) / PFV * 100%. The minimum S d The % value is located at the FH level S3 level Sd% value is 7.34%.

[0090] Prediction of PL using VAT area: Logistic regression equations were established with VAT areas at five pelvic levels as independent variables. Logistic regression equations including S4 or FH were highly efficient in predicting PL (Table 4).

[0091] Y=-3.357+0.038×S4(cm 2 )(SEN:88%,SPE:66%,AUC:0.800)

[0092] Y=-5.253+0.066×FH(cm 2 )(SEN:74%,SPE:94%,AUC:0.893) In addition, we

[0093] A binary logistic regression model was developed that combined the VAT area of ​​S4 and FH.

[0094] Y=4.209+0.105×S4(cm 2 )-0.170×FH(cm 2 )

[0095] The AUC of the logistic regression model was 0.927 (95% CI: 0.874-0.980, SEN = 90%, SPE = 88%, threshold P0> 0.462). The overall accuracy of the logistic regression model in predicting PL was 88% ( Figure 6 A-6B). We also performed univariate logistic regression analysis using the area ratios of different pelvic locations. The area under the ROC curve for the FH / S1 ratio was the largest (AUC: 0.823, 95% CI: 0.742-0.904, P < 0.001), followed by the S4 / S1 ratio (AUC: 0.696, 95% CI: 0.592-0.799, P < 0.001). The ROC curve showed that the optimal threshold of PFV for predicting PL was 733.42 cm 3 (AUC=0.834, 95% CI: 0.751-0.916, SEN=80%, SPE=82%).

[0096] Table 4

[0097]

[0098] Figure 7A prediction system based on visceral fat tissue area provided by an embodiment of the present invention comprises:

[0099] An acquisition unit, used for acquiring an image of the sample, wherein the image includes a pelvic image;

[0100] A segmentation unit, used for segmenting the image using a segmentation algorithm to obtain the visceral fat tissue area of ​​the sample;

[0101] The PFV prediction unit is used to predict the pelvic fat volume of the sample based on the visceral fat tissue area.

[0102] In one embodiment, a prediction system based on visceral fat tissue area comprises:

[0103] An acquisition unit, used for acquiring an image of the sample, wherein the image includes a pelvic image;

[0104] A segmentation unit, used for segmenting the image using a segmentation algorithm to obtain the visceral fat tissue area of ​​the sample;

[0105] The PL prediction unit is used to predict whether the sample is classified as pelvic lipomatosis based on the visceral fat tissue area.

[0106] In a specific embodiment, a prediction system based on visceral fat tissue area includes:

[0107] An acquisition unit, used for acquiring an image of the sample, wherein the image includes a pelvic image;

[0108] A segmentation unit, used for segmenting the image using a segmentation algorithm to obtain the visceral fat tissue area of ​​the sample;

[0109] a PFV prediction unit, configured to predict the pelvic fat volume of the sample based on the visceral fat tissue area;

[0110] The PL prediction unit is used to predict whether the sample is classified as pelvic lipomatosis based on the pelvic fat volume and visceral fat tissue area.

[0111] Figure 3 A prediction device based on visceral fat tissue area provided by an embodiment of the present invention includes: a memory and a processor;

[0112] The memory is used to store program instructions;

[0113] The processor is used to call program instructions, and when the program instructions are executed, the above-mentioned prediction method steps based on visceral fat tissue area are implemented.

[0114] One object of the present invention is to provide a computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the above-mentioned prediction method steps based on visceral fat tissue area are implemented.

[0115] The validation results of this validation example show that assigning intrinsic weights to indications can moderately improve the performance of the method relative to the default setting.

[0116] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0117] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0118] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0119] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0120] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, and the storage medium may include: a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, etc.

[0121] A person skilled in the art can understand that all or part of the steps in the above-mentioned embodiment method can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. The above-mentioned storage medium can be a read-only memory, a disk or an optical disk, etc.

[0122] The above is a detailed introduction to a computer device provided by the present invention. For a person skilled in the art, according to the concept of the embodiments of the present invention, there may be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A prediction method based on visceral fat tissue area, characterized in that: The method steps include: S101: Acquire an image of a sample, wherein the image includes a pelvic image; S102: Segmenting the image using a segmentation algorithm to obtain the visceral fat tissue area of ​​the sample; S103: Predicting whether the sample is classified as pelvic lipomatosis based on the visceral fat tissue area; the pelvic image includes any one or more of the following levels: S1, S2, S3, S4, FH; using a segmentation algorithm to segment the pelvic image level S1, S2, S3 or S4 to obtain the S1, S2, S3 or S4 visceral fat tissue area of ​​the sample, and predicting whether the sample is classified as pelvic lipomatosis based on the S1, S2, S3 or S4 visceral fat tissue area; using a segmentation algorithm to segment the pelvic image level FH to obtain the FH visceral fat tissue area of ​​the sample, and predicting whether the sample is classified as pelvic lipomatosis based on the FH visceral fat tissue area; using a segmentation algorithm to segment the pelvic image level S4 and FH to obtain the S4 and FH visceral fat tissue areas of the sample, and predicting whether the sample is classified as pelvic lipomatosis based on the S4 and FH visceral fat tissue areas.

2. The prediction method based on visceral fat tissue area according to claim 1, characterized in that: The method further includes: predicting the pelvic fat volume of the sample based on the visceral fat tissue area in S102.

3. The prediction method based on visceral fat tissue area according to claim 1, characterized in that: The pelvic image also includes the FH level.

4. The prediction method based on visceral fat tissue area according to claim 3, characterized in that: A segmentation algorithm is used to segment the planes S1, S2, S3, S4 or FH in the pelvic image to obtain the S1, S2, S3, S4 or FH visceral fat tissue area of ​​the sample, and the pelvic fat volume of the sample is predicted based on the S1, S2, S3, S4 or FH visceral fat tissue area.

5. The prediction method based on visceral fat tissue area according to claim 2, characterized in that: The image also includes an abdominal image; a segmentation algorithm is used to segment the image to obtain the visceral fat tissue area of ​​the abdominal cavity of the sample; and the pelvic fat volume of the sample is predicted based on the visceral fat tissue area of ​​the abdominal cavity of the sample.

6. The method for predicting visceral fat tissue area according to claim 5, characterized in that: The abdominal cavity images include any one or more of the following levels: -94 -L -100 .

7. The prediction method based on visceral fat tissue area according to claim 6, characterized in that: Use segmentation algorithm to segment the middle layer L of the abdominal image -94 , L -95 , L -96 , L -97 , L -98 , L -99 or L -100 , and obtain the sample L -94 , L -95 , L -96 , L -97 , L -98 , L -99 or L -100 Visceral adipose tissue area, based on the L -94 , L -95 , L -96 , L -97 , L -98 , L -99 or L -100 Visceral adipose tissue area predicted pelvic fat volume in the samples.

8. The method for predicting visceral fat tissue area according to claim 5, characterized in that: A segmentation algorithm is used to segment the pelvic image and the abdominal image to obtain the pelvic visceral fat tissue area and the abdominal visceral fat tissue area of ​​the sample; and the pelvic fat volume of the sample is predicted based on the pelvic visceral fat tissue area and the abdominal visceral fat tissue area of ​​the sample.

9. The prediction method based on visceral fat tissue area according to claim 2, characterized in that: The method of predicting the pelvic fat volume of the sample based on the visceral fat tissue area in S102 is to input the visceral fat tissue area into the constructed model and output the predicted pelvic fat volume of the sample.

10. The prediction method based on visceral fat tissue area according to claim 9, characterized in that: A regression model between the visceral fat area and the pelvic fat area is constructed based on the training set to obtain a constructed model.

11. The method for predicting visceral fat tissue area according to claim 9, characterized in that: The model is a machine learning algorithm or a deep learning algorithm model.

12. The method for predicting visceral fat tissue area according to claim 2, characterized in that: The pelvic fat volume of the sample is obtained, and based on the pelvic fat volume and the area of ​​visceral fat tissue, a classification result of whether the sample has pelvic lipomatosis is predicted.

13. The method for predicting visceral fat tissue area according to claim 12, characterized in that: The classification result of whether the sample has pelvic lipomatosis is predicted based on the pelvic fat volume and the S4 and FH visceral fat tissue areas.

14. The method for predicting visceral fat tissue area according to claim 12 or 13, characterized in that: The pelvic fat volume is a directly measured pelvic fat volume or a pelvic fat volume of a sample predicted based on the visceral fat tissue area.

15. The method for predicting visceral fat tissue area according to claim 12 or 13, characterized in that: The pelvic fat volume is the pelvic fat volume directly measured by MRIcron or Syngo Volume.

16. The prediction method based on visceral fat tissue area according to claim 1, characterized in that: Step S103 of the method is to input the visceral fat tissue area into the constructed model, and output a classification result of whether the predicted sample is pelvic lipomatosis.

17. The method for predicting visceral fat tissue area according to claim 16, characterized in that: A regression model between the visceral fat area and pelvic lipomatosis is constructed based on the training set to obtain a constructed model.

18. The method for predicting visceral fat tissue area according to claim 16 or 17, characterized in that: The model is a machine learning algorithm or a deep learning algorithm model.

19. The method for predicting visceral fat tissue area according to claim 1, characterized in that: The network used by the segmentation algorithm includes any one of the following: U-net, PSP-net, FCN, Res Net, Inception or a segmentation network optimized based on the above architecture.

20. The prediction method based on visceral fat tissue area according to claim 19, characterized in that: The network used in the segmentation algorithm is the nnU-Net segmentation network.

21. A prediction device based on visceral fat tissue area, characterized in that: include: Memory and processor; The memory is used to store program instructions; The processor is used to call program instructions, and when the program instructions are executed, the steps of the prediction method based on visceral fat tissue area described in any one of claims 1-20 are implemented.

22. A computer program product having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the prediction method based on visceral fat tissue area according to any one of claims 1 to 20 are implemented.

23. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the prediction method based on visceral fat tissue area according to any one of claims 1 to 20 are implemented.