Marker combination related to benign and malignant pulmonary nodules, method for distinguishing benign and malignant pulmonary nodules and application

Abstract

The invention belongs to the technical field of biomedicine, and particularly relates to a marker combination related to benign and malignant pulmonary nodules, a pulmonary nodule benign and malignant distinguishing method and application. The marker composition related to the benign and malignant pulmonary nodules comprises four genes of SHOX2, CDO1, H4C6 and PTGER4, the combination of the four genes has high sensitivity to the malignant pulmonary nodules and can be stably detected in ctDNA, and a new effective detection way is provided for judging the benign and malignant pulmonary nodules. On the basis of molecular detection data brought by the marker composition, in combination with CT image data and clinical parameters, a complete multi-modal data fusion method is provided, the benign and malignant pulmonary nodules of a to-be-detected object can be preliminarily graded, and accurate, dynamic and low-cost pulmonary nodule risk assessment is achieved.

Description

Technical Field

[0001] The present invention belongs to the field of biomedical technology, and specifically relates to a combination of gene markers related to the benign and malignant nature of pulmonary nodules, a method for distinguishing benign and malignant pulmonary nodules based on multimodal data fusion, and its application. Background Art

[0002] In clinical medicine, assessing the benign or malignant nature of lung nodules has always been a challenging problem. Traditional detection methods rely heavily on the physician's experience and subjective judgment, resulting in significant discrepancies between different physicians, leading to frequent misdiagnoses and missed diagnoses. Furthermore, the complex and diverse characteristics of lung nodules, such as morphology, size, and density, make it difficult to accurately identify some tiny nodules using a single diagnostic method, making early-stage lung cancer screening extremely challenging.

[0003] Existing technologies also have significant limitations in the diagnosis of lung nodules. Currently, commonly used detection technologies such as CT imaging and molecular testing are mostly based on single-modality data analysis. Single-modality data can only reflect partial characteristics of lung nodules and cannot fully and accurately describe the biological characteristics of lung nodules. For example, while CT images can clearly display the morphology and location of lung nodules, their accuracy in detecting some subtle lesions and early-stage lung cancer is limited. Methylation testing can also be subject to errors due to factors such as sample quality and detection methods.

[0004] Among existing technologies, the Lung Imaging Reporting and Data System (Lung-RADS) proposed by the American College of Radiology is the current international mainstream standard, but it is mainly based on data from Western populations and lacks validation and optimization for the Chinese population. Clinical data show that the sensitivity of Lung-RADS for high-risk lung nodules in China is only 63.3%, which poses a significant risk of missed diagnosis. In addition, the system relies solely on CT image features (such as nodule size and density) and does not integrate clinical parameters or molecular markers, resulting in insufficient ability to distinguish benign from malignant nodules. Although AI lung nodule analysis methods (such as Yizhun AI and United Imaging Intelligence) can automatically detect nodules and conduct quantitative analysis (size, density, volume), their core is still based on a single imaging modality and has the following problems: 1. High false positive rate: Ground glass nodules overlap with imaging features of benign lesions such as inflammation and tuberculosis, and the AI ​​misjudgment rate can reach 30%. 2. Lack of dynamic assessment: Most systems lack dynamic tracking of follow-up data (such as nodule volume doubling time), making it impossible to optimize long-term risk assessment.

[0005] While non-invasive diagnostic technology based on cfDNA methylation in blood can improve the specificity of distinguishing benign from malignant tumors (up to 93.3%), it still faces the following challenges: 1. Insufficient sensitivity. For small nodules with a diameter of <6 mm, the ctDNA concentration is below the detection threshold, and the sensitivity drops sharply to below 50%. 2. High cost: Whole-genome methylation sequencing (cfMeDIP-seq) is expensive, making it difficult to popularize in primary care. 3. Biological heterogeneity: Some benign lesions (such as tuberculosis and old inflammation) may release similar methylation signals, resulting in false positives.

[0006] The emergence of multimodal fusion technology has become the key to solving these problems. By integrating and analyzing data from different modalities (such as CT images, molecular data, and clinical information), it is possible to evaluate lung nodules from multiple dimensions, improve diagnostic accuracy and reliability, reduce false positive rates and missed diagnosis rates, and provide patients with more precise diagnosis and treatment options.

[0007] Domestic and international studies have attempted to integrate imaging, clinical, and molecular data to construct predictive models (such as the Mayo model and the Brock model). However, these models face key challenges: 1. Single data source: Western models (such as the Mayo model) fail to incorporate risk factors specific to the Chinese population (such as family history and occupational exposures). 2. Static modeling: Most models are based on cross-sectional data from a single time point and do not dynamically integrate follow-up information.

[0008] It can be seen that multimodal data fusion technology provides new ideas and methods to solve these problems, but the existing multimodal fusion methods still face many technical challenges.

[0009] Therefore, providing a new method for judging the benign and malignant nature of lung nodules based on multimodal fusion technology to break through the bottleneck of existing technology and provide a better reference for the selection of clinical treatment plans is a technical problem that needs to be solved urgently. Summary of the Invention

[0010] In order to solve the above technical problems, one of the objectives of the present invention is to provide a marker combination related to benign and malignant lung nodules, wherein the marker combination includes four genes: SHOX2, CDO1, H4C6, and PTGER4.

[0011] A second object of the present invention is to provide the use of the marker combination described above in the preparation of a product for detecting benign and malignant pulmonary nodules.

[0012] Preferably, the product comprises a kit, a preparation or a chip.

[0013] Preferably, the product is a kit comprising primer pairs and probes for detecting the methylation levels of four genes: SHOX2, CDO1, H4C6, and PTGER4.

[0014] The second purpose of the present invention is to provide a method for classifying the malignant risk of lung nodules based on multimodal data fusion. This method uses the combined detection results of the markers as described above, and combines imaging data and clinical information to make a preliminary risk classification of the lung nodules of the test subject. The results obtained cannot be used as a direct diagnosis of the benign or malignant nature of the lung nodules, but can only be used as intermediate information or research.

[0015] The following steps are involved:

[0016] S1. Initial Assessment:

[0017] Obtain CT imaging data of the subjects to be tested and extract nodule size data. Based on the nodule size data, the subjects are classified as low-potential risk and those requiring attention. Low-potential risk subjects undergo annual follow-up, while those requiring attention undergo further evaluation.

[0018] S2. Comprehensive Assessment:

[0019] S21. Continue to extract radiomics features based on the CT image data obtained in step S1; simultaneously, obtain clinical parameters of the subject of interest, obtain peripheral blood free DNA methylation detection data, and extract the methylation level of the marker combination according to claim 1;

[0020] S22. Calculate the score S according to the following dynamic weighted logistic regression model:

[0021] S=0.6×P img +0.3×P meth +0.1×P clinical

[0022] In the model, P img represents the prediction score based on radiomics features, P meth represents the prediction score of methylation level based on the marker combination; P clinical represents the prediction score based on clinical parameters;

[0023] Among them, the P img The calculation formula is as follows:

[0024]

[0025] Where σ is the sigmoid function, x′ i is the eigenvalue of the input i-th radiomics feature, w i is the weight coefficient corresponding to the input i-th radiomics feature, and b is the bias term;

[0026] The P meth The calculation formula is as follows:

[0027]

[0028] Where σ is the sigmoid function, w i is the weight coefficient corresponding to the input i-th methylation gene, is the Ct value of the methylation level of the i-th gene input, Preset the maximum Ct threshold for the input methylation level of the i-th gene;

[0029] The P clinical The calculation formula is as follows:

[0030]

[0031] {Clinical score} = 0.15 × {age} + 0.15 × {smoking history} + 0.2 × {occupational exposure}

[0032] +0.1×{family history}

[0033] Wherein, {age} is 0 if under 50 years old and 0.1 / 5 years if over 50 years old; {smoking history} is 1 if smoking more than 20 packs / year, otherwise 0; {occupational exposure} is 1 if there is occupational exposure, otherwise 0; {family history} is 1 if there is family history, otherwise 0;

[0034] S23. Determine the malignancy risk level of the pulmonary nodule based on the calculated score S output in step S22 as follows:

[0035] 0≤S<0.5, classified as low-risk subjects;

[0036] 0.5≤S<0.65, classified as medium-risk subjects;

[0037] 0.65≤S<0.89, classified as high-risk subjects;

[0038] S ≥ 0.89, classified as extremely high-risk subjects;

[0039] Regular follow-up is performed for low-risk and moderate-risk subjects, and clinical intervention is recommended for high-risk and very high-risk subjects.

[0040] Preferably, in step S1, the classification is based on: if the nodule size is ≤4 mm, it is a low potential risk object; otherwise, it is an object requiring attention.

[0041] Preferably, in step S21, the imaging genomics features include: maximum diameter of the nodule, depth of lobulation, length of burr sign, grade of vascular clustering sign, presence of cavitation sign, average CT value, proportion of solid component, standard deviation of CT value heterogeneity, position of lung lobe, distance from pleura, grayscale co-occurrence matrix contrast, and wavelet transform HLH energy.

[0042] Preferably, the x′i and w i Take values ​​according to the following table:

[0043]

[0044]

[0045]

[0046] Preferably, in step S21, w i and Take values ​​according to the following table:

[0047]

[0048] When the The value exceeds is considered as not detected, and The value of is recorded as -0.3.

[0049] Preferably, the method further includes dynamically adjusting the calculated score S, as follows:

[0050] When at least two of the genes tested among SHOX2, CDO1, H4C6, and PTGER were positive, P meth The weight is adjusted to 0.35, P omg The weight is adjusted to 0.6, P clinical The weight of is adjusted to 0.35;

[0051] When the test results of SHOX2, CDO1, H4C6, and PTGER were all negative and at least one of the following conditions existed in the imaging features: lobulation depth ≥ 2 mm, spiculation length ≥ 2 mm, vascular clustering grade ≥ 2, and nodule diameter > 15 mm, P meth The weight is adjusted to 0.15, P img The weight is adjusted to 0.75, P clinical The weight is adjusted to 0.1.

[0052] Preferably, it also includes post-follow-up data correction:

[0053] Obtain follow-up data after any follow-up period, including volume doubling time (VDT), newly added lobulation sign or spiculation sign a, and methylation level change rate b of the marker combination, and substitute them into the following formula for calculation:

[0054] S new =S+0.08×{VDT}+0.15×a+0.1×b

[0055] Where S is the initial calculation score, {VDT} is recorded as 1 when it is less than 400 days, otherwise it is 0, when the nodule becomes smaller, {VDT} is recorded as -1; when there is a new lobulation sign or spiculation sign, a is recorded as 1, otherwise it is 0; when the number of positive genes in the marker combination increases, b = P meth,new / P meth , otherwise 0.

[0056] Finally, the present invention provides a chip storing a program for running the above-mentioned method for classifying the malignant risk of pulmonary nodules.

[0057] The beneficial effects of the present invention are:

[0058] By providing a combination of markers related to the benign and malignant nature of lung nodules, the combination has a sensitivity of 78.7% for malignant lung nodules. Compared with traditional markers (such as CEA, with a sensitivity of approximately 10%-40%), the combination is stably detected in ctDNA, with a detection rate increased by 30%, providing a new and effective detection approach for judging the benign and malignant nature of lung nodules.

[0059] Based on the molecular detection data brought by the marker composition, combined with CT imaging data and clinical parameters (age, smoking history, family cancer history, occupational exposure, etc.), a complete multimodal data fusion method is provided, which can preliminarily classify the benign and malignant nature of the lung nodules of the subjects to be tested, and realize the accurate, dynamic and low-cost risk assessment of lung nodules.

[0060] This risk classification method establishes a cross-modal dynamic weight allocation strategy and a clinical-imaging-molecular ternary fusion architecture. Compared with traditional single-modality models, it has the following advantages: ① It incorporates risk characteristics of the Chinese population (such as occupational exposure and family history) and optimizes the classification threshold; ② A cross-modal attention mechanism integrates imaging, molecular, and clinical data to reduce misclassification; ③ A low-cost qPCR methylation panel and dynamic weight allocation improve sensitivity (the AUC of the simple CT model is 0.83 or the AUC of the methylation model is 0.79, while the AUC of this method is increased to 0.93); ④ Follow-up data (such as VDT) can be introduced to update the risk score in real time, enabling dynamic tracking.

[0061] It should be emphasized that the results obtained by the risk classification method provided in this application cannot be used as a direct diagnosis of the benign or malignant nature of lung nodules. They can only be used as intermediate information for clinical diagnosis reference or research, such as assisting doctors in determining clinical intervention priorities or assisting in confirming treatment plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 These are the methylation detection results of the four genes SHOX2, CDO1, H4C6, and PTGER4 in different lung nodule samples in Example 1.

[0063] Figure 2 Comparison of ROC curves for clinical, imaging, molecular and multimodal fusion methods of this application. DETAILED DESCRIPTION

[0064] To facilitate understanding, the technical solution of the present invention is described in more detail below with reference to embodiments.

[0065] Unless otherwise specified, all raw materials, reagents, instruments, and equipment used in this article can be purchased from the market or prepared by existing methods.

[0066] Example 1

[0067] Screening of marker combinations

[0068] In 2023, the inventors' team conducted a study in collaboration with three tertiary hospitals, enrolling 367 patients with pulmonary nodules. Inclusion criteria included patients aged 18-80 years with pulmonary nodules 3-30 mm in diameter detected by CT screening, and primary nodules rather than metastases from other lesions. Exclusion criteria included patients with other serious cardiopulmonary diseases, mental illness, and inability to cooperate with the trial.

[0069] The blind verification process adopts a double-blind design, and neither the doctors nor the patients involved in the diagnosis are aware of the patient's true diagnosis results. In terms of the multi-center collaboration mechanism, multiple tertiary hospitals participate, and each center collects and analyzes data according to a unified plan. The follow-up collection cycle is once every 3 months, and blood samples and imaging data are collected simultaneously. At the same time as the blood samples are collected, a CT scan is performed to ensure the temporal consistency of the methylation data and the imaging data. The collected blood samples are immediately sent to the designated laboratory for processing, and the imaging data are uploaded to a unified data platform for storage and analysis.

[0070] like Figure 1 As shown in the results, the methylation results of the four genes SHOX2, CDO1, H4C6, and PTGER4 were significantly different in benign and malignant nodules, which is meaningful for the differentiation of benign and malignant lung nodules.

[0071] SHOX2, CDO1, H4C6, and PTGER4 gene methylation detection is performed using a PCR fluorescent probe method, which includes two steps:

[0072] Step 1: Use a plasma free DNA extraction kit to extract free DNA from plasma, then use sulfite to convert unmethylated cytosine to produce uracil sulfonate through a deamination reaction. Methylated cytosine will not be converted by sulfite.

[0073] In step 2, the methylated DNA is amplified by triple PCR. The blockers and probes in the PCR reaction distinguish between methylated and unmethylated sequences, allowing amplification of the methylated sequences. Fluorescent probes that specifically bind to methylated SHOX2 / CDO1 and H4C6 / PTGER4 gene sequences allow for the exclusive detection of methylated sequences in the PCR reaction. An internal control, the ACTB (β-actin) gene, is used to assess the adequacy of the DNA quantity used in the assay. Positive and negative controls are provided in the kit and should be included in every assay.

[0074] As shown in Table 1, the combined detection of the four genes has better identification accuracy than the identification results of a single gene, and the sensitivity for malignant lung nodules is 78.7%, indicating that it can be used as a marker combination for judging the benign and malignant nature of lung nodules.

[0075] Table 1 Comparison of gene methylation detection results

[0076]

[0077] Example 2

[0078] A method for classifying the malignancy risk of pulmonary nodules based on multimodal data fusion includes the following steps:

[0079] S1. Initial Assessment:

[0080] CT imaging data of the subjects to be tested are obtained, and nodule size data are extracted. Based on the size data of nodules (including solid / ground glass / mixed types), those ≤4mm are classified as low-potential risk subjects and undergo annual follow-up; those larger than 4mm are classified as subjects requiring attention and undergo the next comprehensive assessment.

[0081] S2. Comprehensive Assessment:

[0082] S21. Data acquisition:

[0083] 1) Based on the CT image data obtained in step S1, continue to extract radiomic features. The radiomic features include: morphological features (maximum diameter of nodules, lobulation depth, burr sign length, vascular clustering grade, and presence of cavitation signs), density features (mean CT value, proportion of solid components, and standard deviation of CT value heterogeneity), position features (lobe partition location and distance from the pleura), and texture features (gray-level co-occurrence matrix contrast and wavelet transform HLH energy).

[0084] 2) Obtain clinical parameters of the subjects of concern, including age, smoking history, family history of cancer, and occupational exposure.

[0085] 3) Obtain peripheral blood free DNA methylation detection data and extract the methylation level results of SHOX2, CDO1, H4C6, and PTGER4 genes.

[0086] S22. Calculate the score S according to the following dynamic weighted logistic regression model:

[0087] S=0.6×P img +0.3×P meth +0.1×P clinical

[0088] In the model, P img represents the prediction score based on radiomics features, P meth represents the prediction score of methylation level based on the marker combination; P clinical represents the prediction score based on clinical parameters;

[0089] Among them, the P omg The calculation formula is as follows:

[0090]

[0091] Where σ is the sigmoid function (output 0-1, indicating the probability of malignancy), x′ i is the characteristic value of the input i-th radiomics feature (standardized / normalized), w i is the weight coefficient corresponding to the i-th radiomics feature of the input, and b is the bias term.

[0092] Image feature extraction training is performed through a multi-head cross-attention mechanism for the detection and classification of lung nodules. The dataset uses LIDC-IDRI (The Lung Image Database Consortium and Image Database Resource Initiative). The extracted image features and their weights are shown in Table 2.

[0093] Table 2 Image features and weight distribution

[0094]

[0095]

[0096] The P meth The calculation formula is as follows:

[0097]

[0098] Where σ is the sigmoid function, which maps the linear output to a 0-1 probability; w i is the weight coefficient corresponding to the input i-th methylated gene, which is determined by regression analysis of the training set; is the Ct value of the methylation level of the i-th gene input; The maximum Ct threshold is preset for the methylation level of the input gene i (a value exceeding this value is considered as not detected); if Exceed but

[0099] The methylation marker thresholds and weights are given in Table 2:

[0100] Table 2 Thresholds and weight values

[0101]

[0102] The P clinical The calculation formula is as follows:

[0103]

[0104] [Clinical score] = 0.15 × {age} + 0.15 × {smoking history} + 0.2 × {occupational exposure}

[0105] +0.1×{family history}

[0106] Wherein, {age} is 0 if under 50 years old and 0.1 / 5 years if over 50 years old; {smoking history} is 1 if smoking more than 20 packs / year and 0 otherwise; {occupational exposure} is 1 if there is occupational exposure and 0 otherwise; {family history} is 1 if there is a family history and 0 otherwise.

[0107] According to the above calculation, the output calculation score S is dynamically adjusted:

[0108] 1) When at least two of the genes tested among SHOX2, CDO1, H4C6, and PTGER were positive, in the dynamic weighted logistic regression model, P meth The weight is adjusted to 0.35, P img The weight is adjusted to 0.6, P clonical The weight of is adjusted to 0.35;

[0109] 2) When the test results of SHOX2, CDO1, H4C6, and PTGER were all negative and the imaging features contained at least one of the following: lobulation depth ≥ 2 mm, spiculation length ≥ 2 mm, vascular clustering grade ≥ 2, and nodule diameter > 15 mm, P meth The weight is adjusted to 0.15, P img The weight is adjusted to 0.75, P clinical The weight is adjusted to 0.1.

[0110] S23. Determine the malignancy risk level of the pulmonary nodule based on the calculated score S outputted in step S22 as follows:

[0111] 0≤S<0.5, classified as low-risk subjects;

[0112] 0.5≤S<0.65, classified as medium-risk subjects;

[0113] 0.65≤S<0.89, classified as high-risk subjects;

[0114] S ≥ 0.89, classified as extremely high-risk subjects;

[0115] Regular follow-up is conducted for low-risk and medium-risk subjects, for example, annual regular follow-up is conducted for low-risk subjects, and regular follow-up is conducted every 3 or 6 months for medium-risk subjects; clinical intervention is recommended for high-risk and extremely high-risk subjects, for example, high-risk subjects can be given priority intervention, and extremely high-risk subjects undergo clinical intervention.

[0116] Furthermore, based on the follow-up, data correction is also included, the method is as follows:

[0117] Obtain follow-up data after any follow-up period, including volume doubling time (VDT), newly added lobulation sign or spiculation sign a, and methylation level change rate b of the marker combination, and substitute them into the following formula for calculation:

[0118] S new =S+0.08×{VDT}+0.15×a+0.1×b, where S is the initial calculation score, {VDT} is 1 when it is less than 400 days, otherwise it is 0, and when the nodule becomes smaller, {VDT} is recorded as -1; when there is a new lobulation sign or spiculation sign, a is recorded as 1, otherwise it is 0; when the number of positive genes in the marker combination increases, b=P meth,new / P meth , otherwise 0.

[0119] Example 3

[0120] Based on the samples of Example 1, ROC curves were calculated and drawn according to clinical, methylation, and imaging data. The AUC value of the clinical-imaging-molecular multimodal fusion ROC curve of this method (0.93) was significantly better than that of the individual clinical (AUC 0.56), imaging (AUC 0.83), and molecular (AUC 0.79) models. Figure 2 Clinical comparative data showed that this method missed <2% of malignant nodules with medium-to-high risk or above (score ≥ 0.5), misdiagnosed <2% of benign nodules with high risk or above (score ≥ 0.65), and exempted some benign nodules from invasive examinations for medium-to-low risk (score < 0.65), as shown in Table 2.

[0121] Table 2 Comparison of clinical pathological results (cases) and risk levels of this method

[0122]

[0123] Example 4

[0124] The information of a patient with pulmonary nodules is as follows:

[0125] Female, 49 years old, no smoking history, no family history of cancer, and no occupational exposure.

[0126] CT features: 1) Pure ground-glass nodule (pGGN) in the right upper lobe, 10 mm in diameter; 2) Lobulation sign (-), spiculation sign (-), cavitation sign (+); 3) Average CT value: -580 HU, solid component proportion 0%.

[0127] Methylation detection: SHOX2 (not detected), CDO1 (not detected), H4C6 (not detected), PTGER4 (not detected).

[0128] The analysis was performed using the pulmonary nodule malignancy risk classification method based on multimodal data fusion provided in this application. The results are as follows:

[0129] S1. Preliminary assessment: Ground-glass nodules with a diameter of 10 mm and greater than 4 mm are classified as requiring attention and undergo a second-step comprehensive assessment.

[0130] S2. Comprehensive evaluation: further extract data to calculate P img =0.65, P met =0.43,

[0131] P clinical =0.5;

[0132] Finally, the score S was calculated by the dynamic weighted logistic regression model to be 0.5, 0.5≤0.54<0.65, and the subjects were classified as medium-risk subjects, and regular follow-up was recommended.

[0133] After 3 months of follow-up, the nodule had shrunk and the score dropped to 0.49, turning the patient into a low-risk subject.

[0134] The patient's surgical pathology results showed lung inflammation.

[0135] The above results show that by stratifying patients' risks according to the method of this application and conducting regular follow-up, unnecessary puncture biopsies can be avoided and overdiagnosis and treatment can be reduced.

[0136] Example 5

[0137] The information of a patient with pulmonary nodules is as follows:

[0138] Male, 58 years old, no smoking history, no family history of cancer, and no occupational exposure.

[0139] CT features: 1) solid nodule in the left lower lobe, 5 mm in diameter; 2) lobulation sign (+, depth 2 mm), spiculation sign (+, length 1.5 mm); 3) average CT value: 180 HU, with solid components accounting for 100%.

[0140] Methylation detection: SHOX2 (Ct=34.8), CDO1 (not detected), H4C6 (Ct=36.1), PTGER4 (37.3).

[0141] The analysis was performed using the pulmonary nodule malignancy risk classification method based on multimodal data fusion provided in this application. The results are as follows:

[0142] S1. Preliminary assessment: Solid nodules with a diameter of 5 mm or greater than 4 mm are classified as requiring attention and undergo a second-step comprehensive assessment.

[0143] S2. Comprehensive evaluation: Because two markers were tested positive, in the dynamic weighted logistic regression model, P meth The weight is adjusted to 0.35, P img The weight is adjusted to 0.6, P clinical The weight is adjusted to 0.05.

[0144] Further extract data to calculate P img =0.57, P meth =0.86, P clinical =0.51;

[0145] The final calculated score was S = 0.67, 0.65≤0.67<0.89, and the subject was classified as a high-risk subject, and intervention was recommended.

[0146] Surgical resection confirmed minimally invasive adenocarcinoma (MIA).

[0147] This patient was classified as low risk according to traditional Lung-RADS. The above results show that the risk classification of patients according to this application method has broken through the sensitivity limitation of traditional imaging for small lesions.

[0148] Example 6

[0149] The information of a patient with pulmonary nodules is as follows:

[0150] Patient information: Female, 47 years old, no smoking history, family history of lung cancer (mother).

[0151] CT features: 1) mixed ground-glass nodule in the right middle lobe, 8 mm in diameter (30% solid); 2) lobulation sign (-), spiculation sign (+, 1.8 mm in length).

[0152] Methylation detection: only SHOX2 was weakly positive, SHOX2 (Ct=37.9).

[0153] The analysis was performed using the pulmonary nodule malignancy risk classification method based on multimodal data fusion provided in this application. The results are as follows:

[0154] S1. Preliminary assessment: Solid nodules with a diameter of 8 mm or greater than 4 mm are classified as requiring attention and undergo a second-step comprehensive assessment.

[0155] S2. Comprehensive evaluation: further extract data to calculate P img =0.63, P meth =0.49,

[0156] P clinical =0.52;

[0157] The final score calculated by the dynamic weighted logistic regression model was S=0.58, which classified the patient as a medium-risk subject and recommended regular follow-up.

[0158] After 3 months of follow-up, the nodule increased to 9.5 mm, and the solid proportion increased to 45%; methylation retest: SHOX2 (Ct=35.1) and CDO1 (Ct=36.7), both positive.

[0159] VDT = 280 days (<400-day threshold, corresponding to a = 0.18).

[0160] The final score S = 0.58 (initial score) + 0.08 (VDT) + 0.16 (methylation change rate) = 0.82, which is classified as a high-risk subject.

[0161] After surgical intervention, pathological results showed invasive adenocarcinoma (IA).

[0162] The above results show that the dynamic tracking provided by the method of this application can improve the ability to identify rapidly progressing nodules.

[0163] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A combination of markers related to benign and malignant pulmonary nodules, characterized in that: The marker combination includes four genes: SHOX2, CDO1, H4C6, and PTGER4.

2. Use of the marker combination as claimed in claim 1 in the preparation of a product for detecting benign and malignant pulmonary nodules.

3. The use according to claim 2, characterized in that The product includes a kit, a preparation or a chip.

4. The use according to claim 2, characterized in that The product is a kit, which includes primer pairs and probes for detecting the methylation levels of four genes: SHOX2, CDO1, H4C6, and PTGER4.

5. A method for classifying the malignant risk of pulmonary nodules based on multimodal data fusion, characterized in that: The following steps are involved: S1. Initial Assessment: Obtain CT imaging data of the subjects to be tested and extract nodule size data. Based on the nodule size data, the subjects are classified as low-potential risk and those requiring attention. Low-potential risk subjects undergo annual follow-up, while those requiring attention undergo further evaluation. S2. Comprehensive Assessment: S21. Continue to extract radiomics features based on the CT image data obtained in step S1; simultaneously, obtain clinical parameters of the subject of interest, obtain peripheral blood free DNA methylation detection data, and extract the methylation level of the marker combination according to claim 1; S22. Calculate the score S according to the following dynamic weighted logistic regression model: S=0.6×P img +0.3×P meth +0.1×P clinical In the model, P img represents the prediction score based on radiomics features, P meth represents the prediction score of methylation level based on the marker combination; P clinical represents the prediction score based on clinical parameters; Among them, the P img The calculation formula is as follows: Where σ is the sigmoid function, x′ i is the eigenvalue of the input i-th radiomics feature, w i is the weight coefficient corresponding to the input i-th radiomics feature, and b is the bias term; The P meth The calculation formula is as follows: Where σ is the sigmoid function, w i is the weight coefficient corresponding to the input i-th methylation gene, is the Ct value of the methylation level of the i-th gene input, Preset the maximum Ct threshold for the input methylation level of the i-th gene; The P clinival The calculation formula is as follows: {Clinical score} = 0.15 × {age} + 0.15 × {smoking history} + 0.2 × {occupational exposure} + 0.1 × {family history} Wherein, {age} is 0 if under 50 years old and 0.1 / 5 years if over 50 years old; {smoking history} is 1 if smoking more than 20 packs / year, otherwise 0; {occupational exposure} is 1 if there is occupational exposure, otherwise 0; {family history} is 1 if there is family history, otherwise 0; S23. Determine the malignancy risk level of the pulmonary nodule based on the calculated score S output in step S22 as follows: 0≤S<0.5, classified as low-risk subjects; 0.5≤S<0.65, classified as medium-risk subjects; 0.65≤S<0.89, classified as high-risk subjects; S ≥ 0.89, classified as extremely high-risk subjects; Regular follow-up is performed for low-risk and moderate-risk subjects, and clinical intervention is recommended for high-risk and very high-risk subjects.

6. The method for classifying malignant risk of pulmonary nodules based on multimodal data fusion according to claim 5, characterized in that: In step S1, the classification is based on the following criteria: if the nodule size is ≤4 mm, it is a low potential risk object; otherwise, it is an object requiring attention.

7. The method for classifying malignant risk of pulmonary nodules based on multimodal data fusion according to claim 5, wherein: In step S21, the imaging features include: maximum diameter of nodules, lobulation depth, burr length, vascular clustering grade, presence of cavitation, average CT value, proportion of solid components, standard deviation of CT value heterogeneity, lobe division position, distance from pleura, gray-level co-occurrence matrix contrast, and wavelet transform HLH energy.

8. The method for classifying malignant risk of pulmonary nodules based on multimodal data fusion according to claim 6, wherein: The x′ i and w i Take values ​​according to the following table:

9. The method for classifying malignant risk of pulmonary nodules based on multimodal data fusion according to claim 6, wherein: In the step S21, Take values ​​according to the following table: When the The value exceeds is considered as not detected, and The value of is recorded as -0.

3.

10. The method for classifying malignant risk of pulmonary nodules based on multimodal data fusion according to claim 6, wherein: It also includes dynamic adjustment of the calculated score S, which is as follows: When at least two of the genes tested among SHOX2, CDO1, H4C6, and PTGER were positive, P meth The weight is adjusted to 0.35, P img The weight is adjusted to 0.6, P clinical The weight of is adjusted to 0.35; When the test results of SHOX2, CDO1, H4C6, and PTGER were all negative and at least one of the following conditions existed in the imaging features: lobulation depth ≥ 2 mm, spiculation length ≥ 2 mm, vascular clustering grade ≥ 2, and nodule diameter > 15 mm, P meth The weight is adjusted to 0.15, P img The weight is adjusted to 0.75, P clinical The weight is adjusted to 0.

1.

11. The method for classifying malignant risk of pulmonary nodules based on multimodal data fusion according to claim 6, wherein: Also includes post-follow-up data revisions: Obtain follow-up data after any follow-up period, including volume doubling time (VDT), newly added lobulation sign or spiculation sign a, and methylation level change rate b of the marker combination, and substitute them into the following formula for calculation: S new =S+0.08×{VDT}+0.15×a+0.1×b Where S is the initial calculation score, {VDT} is recorded as 1 when it is less than 400 days, otherwise it is 0, when the nodule becomes smaller, {VDT} is recorded as -1; when there is a new lobulation sign or spiculation sign, a is recorded as 1, otherwise it is 0; when the number of positive genes in the marker combination increases, b = P meth,new / P meth , otherwise 0.

12. A chip, characterized in that: A program for running the method for classifying the malignant risk of pulmonary nodules as described in any one of claims 5 to 12 is stored.

Images