Method and system for deducing rib fracture time based on 99mTc-MDP SPECT semi-quantitative analysis

Abstract

The invention provides a method and system for deducing rib fracture time through < 99m > Tc-MDP SPECT semi-quantitative analysis, and belongs to the technical field of biology. The method comprises the following steps: S1, after < 99m > Tc-MDP is injected into a subject, performing whole body bone imaging by using SPECT (Single Photon Emission Computed Tomography); s2, on the 99mTc-MDP SPECT image, sketching a fracture rib T and a normal rib NT at the same position of the uninjured side, and calculating the ratio of T / NT; s3, constructing a regression three-classification model by adopting a Neural Network method, and inputting training set and test set data into the constructed visual analysis network to obtain a training model; and S4, inputting the data of the sample to be tested into the test set to obtain an inference result of the fracture time. The method is high in accuracy, simple to operate and suitable for wide application.

Description

Technical Field

[0001] This invention belongs to the field of biotechnology, specifically relating to... 99m A systematic method for inferring rib fracture time using Tc-MDP SPECT semi-quantitative analysis. Background Technology

[0002] Rib fractures are among the most common chest injuries, often caused by indirect or direct trauma. Accurately determining the exact number of rib fractures is crucial for forensic clinical assessments of disability and injury severity, providing key information for grading. Because early-stage rib fractures are not easily detected on multi-slice spiral CT or X-rays, a follow-up examination of the fracture area is usually required after three weeks to determine the exact number of rib fractures based on the fracture line or callus formation. However, this prolonged interval between examinations can easily lead to forensic clinical assessment conclusions being influenced by fabricated injuries, secondary injuries, or pre-existing injuries.

[0003] Chinese patent CN116843629A discloses a method for detecting rib fractures, comprising: S1: acquiring a preset number of CT images of marked rib fractures and preprocessing them to obtain a sample set; S2: constructing a rib fracture detection model, the model comprising: a backbone network constructed using ResNet50, an FPN network connected to the backbone network, a compression excitation module connected to both the backbone network and the FPN network, an Anchor module for removing negative samples from image X, and a prediction module for obtaining the predicted category and predicted offset vector using the output of the Anchor module; wherein image X is the FPN network. The output image; the negative sample is an anchor for the non-bone tissue foreground; the predicted category is fracture or non-fracture; S3: Initialize the model parameters in the rib fracture detection model, train the rib fracture detection model through the sample set, and calculate the total loss function during training. Based on the calculated value of the total loss function, the model parameters are updated backward using the gradient descent algorithm until the calculated value of the total loss function is less than the preset value, thus obtaining the rib fracture detection target model; S4: Obtain the predicted category and predicted offset vector corresponding to the image to be detected through the rib fracture detection target model, and obtain the location box corresponding to the fracture region through the predicted offset vector. It solves the problem of severe imbalance between positive and negative samples and lack of targeted learning during training by removing negative samples in image X through the Anchor module in the model, thus improving the prediction accuracy of the rib fracture detection model, but its model construction is complex.

[0004] Whole-body bone scintigraphy is a nuclear medicine examination technique administered via intravenous injection. 99m Tc-labeled methylene diphosphonate ( 99mTc-MDP imaging agents are used with SPECT (single-photon emission computed tomography) to assess bone metabolic activity. Bone damage often begins with metabolic abnormalities (such as changes in bone blood flow and increased osteoblast activity), which then gradually develop into abnormal bone morphology and structure. 99m Tc-MDP can be specifically taken up by active osteoblasts. By detecting differences in radioactivity count distribution through single-photon emission computed tomography (SPECT), bone damage can be detected before obvious changes in bone morphology and structure are observed (such as early fractures). In particular, occult rib fractures may be missed by CT due to volume effects or structural overlap, while SPECT can more clearly show these occult lesions through differences in metabolic activity.

[0005] The prior art does not disclose a basis based on 99m Tc-MDP SPECT semi-quantitative analysis method for estimating rib fracture time. Summary of the Invention

[0006] To address the above problems, the present invention provides a method based on... 99m A systematic method for inferring rib fracture time using Tc-MDP SPECT semi-quantitative analysis.

[0007] On the one hand, the present invention provides a basis 99m The method for inferring rib fracture time using Tc-MDP SPECT semi-quantitative analysis includes the following steps: S1, Subject injection 99m After Tc-MDP, whole-body bone scintigraphy was performed using SPECT. S2, in 99m On Tc-MDP SPECT images, the T of the fractured rib and the NT of the normal rib at the same position on the healthy side are delineated, and the T / NT ratio is calculated. S3. Construct a regression tri-classification model using the Neural Network method. Input the training set and test set data into the constructed visualization analysis network to obtain the training model. S4. Input the data of the sample to be tested into the test set to obtain the inference result of the fracture time.

[0008] Specifically, in step S1 99m The injection volume of Tc-MDP is 10-30 mCi.

[0009] According to some embodiments of the present invention, in step S1 99mThe injection volume of Tc-MDP can be 10mCi, 11mCi, 12mCi, 13mCi, 14mCi, 15mCi, 16mCi, 17mCi, 18mCi, 19mCi, 20mCi, 21mCi, 22mCi, 23mCi, 24mCi, 25mCi, 26mCi, 27mCi, 28mCi, 29mCi, 30mCi, and any intermediate value can be selected.

[0010] Furthermore, in step S1 99m The injection volume of Tc-MDP is 10-20 mCi.

[0011] According to some embodiments of the present invention, in step S1 99m The injection volume of Tc-MDP can be 10mCi, 11mCi, 12mCi, 13mCi, 14mCi, 15mCi, 16mCi, 17mCi, 18mCi, 19mCi, 20mCi, and any intermediate value can be selected.

[0012] Preferably, in step S1 99m The injection volume of Tc-MDP is 20 mCi.

[0013] Specifically, in step S1, the subject is injected with... 99m Whole-body bone scintigraphy is performed 2-4 hours after Tc-MDP.

[0014] According to some embodiments of the present invention, in step S1, the subject is injected with... 99m Whole-body bone scintigraphy was performed 2, 2.5, 3, 3.5, and 4 hours after Tc-MDP.

[0015] Preferably, in step S1, the subject is injected with... 99m Whole-body bone scintigraphy was performed 3 hours after Tc-MDP.

[0016] Specifically, the subject can be injected via any route. 99m Tc-MDP; According to some embodiments of the present invention, the injection routes include, but are not limited to, injection via the dorsal vein of the hand, injection via the cubital vein, injection via the scalp vein, injection via the dorsal vein of the foot, and injection via the femoral triangle vein.

[0017] Preferably, the subject receives the injection via the antecubital vein. 99m Tc-MDP.

[0018] Specifically, in step S2, the parameters of the SPECT collimator are: inherent spatial resolution: effective field of view half-width ≤ 3.9 mm; effective field of view half-width ≤ 7.7 mm; planar sensitivity of 202 cpm / μCi; acquisition speed of 13 cm / min; and acquisition using dual probes in front and behind positions.

[0019] Specifically, in step S2, the ROI technique is used to delineate the fractured rib T and the normal rib NT at the same position on the healthy side.

[0020] Specifically, in step S3, a regression three-classification model is constructed based on the Neural Network method in Orange software.

[0021] Specifically, in step S3, the Neurons in hidden layers parameter is set to 100, Activation is set to ReLu, and Solver is set to Adam.

[0022] Specifically, the data used to train the model in step S3 is: When the T / NT ratio is 2.475±0.495, the rib fracture time is within 30 days. When the T / NT ratio is 1.914±0.337, the rib fracture time is 30-60 days. When the T / NT ratio is 1.525±0.297, the rib fracture time is more than 60 days.

[0023] On the other hand, the present invention provides a basis 99m A system for Tc-MDP SPECT semi-quantitative analysis to infer rib fracture time includes: an electronic device, including a processing device, for loading a program and performing the above-described method for analyzing and inferring rib fracture time.

[0024] The beneficial effects of this invention are as follows: This invention clarifies a method based on 99m A semi-quantitative Tc-MDP SPECT method for estimating rib fracture time was used. A Neural Network classification model was constructed based on Orange 3.38.1 software. This model achieved an accuracy of 82.8% in predicting fractures within 30 days, 80.4% in predicting fractures within 30-60 days, and 87.1% in predicting fractures after 60 days. This method uses conventional... 99m The Tc-MDP SPECT image acquisition method is used, and data analysis is performed based on Orange interactive open-source software. It is simple to operate and suitable for a wide range of applications. Attached Figure Description

[0025] Figure 1 To train the T / NT ratios for different fracture time groups, where *** represents P ≤ 0.0005, **** indicates P ≤ 0.0001. Detailed Implementation

[0026] The present invention will be further described in detail below with reference to specific embodiments. The following embodiments are not intended to limit the present invention, but only to illustrate the present invention. Unless otherwise specified, the experimental methods used in the following embodiments are generally performed under conventional conditions. Unless otherwise specified, the materials and reagents used in the following embodiments are commercially available.

[0027] Example 1 1. Sample Collection Collected 111 rib fractures from 32 patients 99m Tc-MDP SPECT and CT image data. Bone scintigraphy was performed using a Siemens SPECT scanner. 99 mg technetium-methylene diphosphate was injected via the antecubital vein. 99m Whole-body bone scintigraphy was performed 3 hours after administering 20 mCi of Tc-MDP imaging agent, with collimator parameters set to low energy and high resolution. Effective field of view (FWHM) ≤ 3.9 mm; effective field of view (FWHM) ≤ 7.7 mm; planar sensitivity of 202 cpm / μCi; acquisition speed of 13 cm / min; and anteroposterior acquisition using dual probes. 99m On Tc-MDP SPECT images, the fractured rib (T) and the normal rib at the same location on the healthy side (NT) were repeatedly delineated using ROI technology. The ratio of the radioactivity count of the fractured rib to that of the normal rib at the same location on the healthy side, i.e., the T / NT ratio, was calculated. This was done three times, and the average of the three ratios was taken. The morphology and structure of the rib fracture on CT were also recorded (Tables 1-6). All fracture data were divided into three groups according to the time of injury: fractures occurring more than 60 days prior were included in the "more than 60 days" group (40 cases); fractures occurring between -60 days prior were included in the "30-60 days" group (28 cases); and fractures occurring within 30 days prior were included in the "within 30 days" group (43 cases).

[0028] Table 1 Data Collection Table

[0029] Table 2 Data Collection Table

[0030] Table 3 Data Collection Table

[0031] Table 4 Data Collection Table

[0032] Table 5 Data Collection Table

[0033] Table 6 Data Collection Table

[0034] 2. Model Building Using R software, 111 fracture cases were randomly divided into a 70% training set and a 30% test set. A regression tri-classification model was constructed based on the Neural Network method in Orange 3.38.1 software, with the Neurons in hidden layers parameter set to 100, Activation set to ReLU, and Solver set to Adam. In the constructed visualization analysis network, the training and test set data were entered into the "File" field. The "Test & Score" field displays the AUC values ​​of the test and training sets, respectively, and the "Predictions" field displays the prediction results for the test set. In future use, the training set (Tables 7-9) from this dataset can be directly applied, and the test data can be used as the test set. The fracture time inference results for the test data can be viewed in "Predictions". The statistical results of the T / NT ratio for different fracture time groups in the training set are shown below. Figure 1 As shown.

[0035] Table 7 Training Set Data

[0036] Table 8 Training Set Data

[0037] Table 9 Training Set Data

[0038] Example 2 Model Validation The area under the ROC curve (AUC) was used as an indicator to evaluate model performance. Using 34 fracture cases as a validation set, the model was constructed with the following accuracy: 82.8% for fractures within 30 days, 80.4% for fractures within 30-60 days, and 87.1% for fractures after 60 days.

[0039] Table 10. Fracture time estimation results for the independent test set.

[0040] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. based on 99m The method for inferring the time of rib fracture by Tc-MDP SPECT semi-quantitative analysis is characterized in that, The method comprises the following steps: S1, injection of subject 99m Whole body bone imaging with SPECT after Tc-MDP; S2, in 99m On Tc-MDP SPECT images, the fractured ribs T and the normal ribs NT in the same position of the healthy side were outlined, and the ratio of T / NT was calculated. S3, a regression three-classification model is constructed by using a Neural Network method, training set and test set data are input into the constructed visual analysis network, and a training model is obtained; S4, sample data to be tested is input into the test set, and an inference result of the fracture time is obtained.

2. The method of claim 1, wherein, In step S1 99m The injected amount of Tc-MDP is 10-30 mCi.

3. The method of claim 2, wherein, In step S1 99m The injected amount of Tc-MDP was 20 mCi.

4. The method of claim 1, wherein, Injection of the subject in step S1 99m Whole body bone imaging was performed 2-4 hours after Tc-MDP.

5. The method of claim 4, wherein, In step S1 the subject is injected 99m Whole body bone imaging was performed 3 hours after Tc-MDP.

6. The method according to any one of claims 1 to 5, characterized in that, The subjects were injected intravenously in the elbow vein 99m Tc-MDP.

7. The method of claim 1, wherein, In step S2, the parameters of the SPECT collimator are inherent spatial resolution: effective field of view half-height width ≤ 3.9 mm; effective field of view half-full width ≤ 7.7 mm, plane sensitivity is 202 cpm / μCi, and the acquisition speed is 13 cm / min, which is collected in a double-probe front and rear position.

8. The method of claim 1, wherein, In step S2, the ROI technology is used to delineate the fractured rib T and the normal rib NT at the same position on the healthy side.

9. The method of claim 1, wherein, In step S3, the Neurons in hidden layers parameter is set to 100, the Activation is set to ReLu, and the Solver is set to Adam.

10. based on 99m A system for inferring the time of a rib fracture based on Tc-MDP SPECT semi-quantitative analysis, characterized by, The method comprises: An electronic device comprising a processing device to load a program and execute the method of analyzing and inferring the rib fracture time according to any one of claims 1-9.

Images