A method for evaluating lymph node metastasis of gastric cancer based on pet / ct
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- YANCHENG NO 1 PEOPLES HOSPITAL
- Filing Date
- 2026-06-02
- Publication Date
- 2026-08-07
AI Technical Summary
胃癌淋巴结转移的评估主要依赖术后病理检查与术前传统影像诊断，在手术切除后无法为术前治疗方案的制定提供依据；而传统术前影像诊断多依赖肉眼观察淋巴结的大小、形态及代谢活性，受医生经验影响较大，对微小转移或早期转移的识别能力有限，且难以量化评估肿瘤的整体侵袭潜能，容易出现漏诊或误判
[0024]本发明通过术前影像组学与临床风险因素的融合，实现胃癌淋巴结转移风险的精准、无创评估，通过分割胃癌原发灶的感兴趣体积并提取多维度影像组学特征，能够从PET代谢、PET纹理及CT解剖层面，全面量化肿瘤的生物学特性，突破传统影像诊断仅依赖肉眼观察的局限，深度挖掘影像数据中与肿瘤侵袭性相关的隐藏信息，为淋巴结转移风险评估提供客观、可量化的依据，通过影像组学特征评分与预设临床风险因素的结合，构建逻辑回归预测模型，有效整合影像组学的肿瘤内在特征与临床血清学、传统影像诊断的外在指标，实现多模态信息的互补，显著提升预测模型的准确性与稳定性，能够同时输出有无淋巴结转移的分类结果与N2至3b期转移的概率值，为临床提供从定性到定量的多层次评估结果。在术前精准识别高淋巴结转移风险，从而避免不必要的扩大手术或过度治疗。
Smart Images
Figure CN122531743A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of gastric cancer assessment technology, and more specifically, to a PET / CT-based method for assessing lymph node metastasis in gastric cancer. Background Technology
[0002] Lymph node metastasis is a key factor influencing patient prognosis and treatment planning, and accurate assessment of lymph node metastasis risk is of great guiding significance for clinical diagnosis and treatment. The assessment of lymph node metastasis in gastric cancer mainly relies on postoperative pathological examination and traditional preoperative imaging diagnosis. However, after surgical resection, it cannot provide a basis for preoperative treatment planning. Traditional preoperative imaging diagnosis relies heavily on visual observation of lymph node size, morphology, and metabolic activity, which is greatly influenced by physician experience and has limited ability to identify micrometastases or early metastases. Furthermore, it is difficult to quantify the overall invasive potential of the tumor, easily leading to missed diagnoses or misdiagnoses. In addition, the detection of single serum tumor markers also suffers from insufficient specificity and cannot comprehensively reflect the biological behavior of the tumor. With the development of radiomics technology, tumor heterogeneity can be quantified by extracting multidimensional features hidden in images. However, current radiomics models mostly rely on single imaging features and do not fully integrate clinical risk factors; therefore, the predictive accuracy and clinical applicability still need improvement. Summary of the Invention
[0003] In view of the shortcomings of the existing technology, the purpose of this invention is to provide a method for evaluating lymph node metastasis in gastric cancer based on PET / CT.
[0004] To achieve the above objectives, the present invention provides the following technical solution:
[0005] A PET / CT-based method for assessing lymph node metastasis in gastric cancer, comprising the following steps:
[0006] Acquire preoperative images of the target subject and segment the volume of interest of the primary gastric cancer lesion from the preoperative images;
[0007] Extract the target radiomics features corresponding to the volume of interest, wherein the radiomics features include at least one of PET metabolic features, PET texture features, and CT anatomical features;
[0008] Image omics feature scores are obtained based on the target image omics features;
[0009] By combining radiomics feature scores with pre-defined clinical risk factors, a lymph node metastasis prediction model was constructed and the results of gastric cancer lymph node metastasis risk were output.
[0010] Preferably, the PET metabolic characteristics include at least one of the following: maximum standard uptake value, average standard uptake value, tumor metabolic volume, and total glycolysis in the lesion.
[0011] The PET texture features include at least one of the following: gray-level co-occurrence matrix features, gray-level run-length matrix features, neighborhood gray-level difference matrix features, and gray-level region length matrix features.
[0012] The CT anatomical features include at least one of shape features and density features.
[0013] Preferably, the volume of interest of the primary gastric cancer lesion is segmented from the preoperative images, specifically as follows:
[0014] The volume of interest was delineated from the preoperative images with a threshold of SUVmax≥40% for the primary lesion.
[0015] Preferably, extracting the target image omics features corresponding to the volume of interest specifically includes the following steps:
[0016] Mann-Whitney U was used to screen candidate features with differences in the volume of interest;
[0017] Candidate features are processed by dimensionality reduction using the minimum absolute shrinkage and selection operator algorithm to obtain the target image omics features.
[0018] Preferably, the preset clinical risk factors include at least one of carbohydrate antigen 199, carcinoembryonic antigen, and conventional PET / CT lymph node metastasis diagnosis results.
[0019] Preferably, the lymph node metastasis prediction model is a logistic regression model.
[0020] Preferably, the method also includes validating the lymph node metastasis prediction model by dividing the dataset into a training set, an internal validation set, and an external validation set, and evaluating the discrimination and clinical applicability of the lymph node metastasis prediction model through receiver operating characteristic (ROC) curves and decision curves.
[0021] Preferably, the gastric cancer lymph node metastasis risk result includes the classification result of whether or not there is lymph node metastasis, and the probability value of lymph node metastasis in N2-3b stage.
[0022] Preferably, it also includes integrating radiomics feature scores with clinical risk factors into a nomograph to visualize the results of gastric cancer lymph node metastasis risk.
[0023] Compared with the prior art, the present invention has the following beneficial effects:
[0024] This invention achieves precise and non-invasive assessment of the risk of lymph node metastasis in gastric cancer by integrating preoperative radiomics with clinical risk factors. By segmenting the volume of interest (VLOS) of the primary gastric cancer lesion and extracting multi-dimensional radiomics features, it comprehensively quantifies the biological characteristics of the tumor from the perspectives of PET metabolism, PET texture, and CT anatomy. This overcomes the limitations of traditional imaging diagnosis, which relies solely on visual observation, and deeply mines hidden information related to tumor invasiveness within the imaging data. It provides an objective and quantifiable basis for assessing the risk of lymph node metastasis. By combining radiomics feature scores with pre-defined clinical risk factors, a logistic regression prediction model is constructed. This effectively integrates the intrinsic tumor characteristics from radiomics with extrinsic indicators from clinical serology and traditional imaging diagnosis, achieving multimodal information complementarity. This significantly improves the accuracy and stability of the prediction model, simultaneously outputting classification results for the presence or absence of lymph node metastasis and probability values for N2 to 3b stage metastasis, providing clinicians with multi-level assessment results from qualitative to quantitative perspectives. Precise identification of high lymph node metastasis risk preoperatively avoids unnecessary extended surgery or overtreatment. Attached Figure Description
[0025] Figure 1 This is a schematic diagram illustrating a method for assessing lymph node metastasis in gastric cancer based on PET / CT, as provided in an embodiment of the present invention.
[0026] Figure 2 This is a schematic diagram illustrating the target influence omics characteristics obtained in a PET / CT-based method for assessing lymph node metastasis in gastric cancer, as provided in an embodiment of the present invention. Detailed Implementation
[0027] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
[0028] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.
[0029] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single embodiment or an embodiment selectively excluded from other embodiments.
[0030] Reference Figures 1-2 As shown.
[0031] The embodiments further illustrate the PET / CT-based method for assessing lymph node metastasis in gastric cancer proposed in this invention.
[0032] A PET / CT-based method for assessing lymph node metastasis in gastric cancer, comprising the following steps:
[0033] Obtain preoperative images of the target patient, and segment the volume of interest (Voice of Interest) of the primary gastric cancer lesion from the preoperative images, specifically:
[0034] The volume of interest was delineated from the preoperative images with a threshold of SUVmax ≥ 40% for the primary lesion.
[0035] First, preoperative images of the target subject are acquired. These images are typically fused PET and CT images, which simultaneously present the metabolic activity and anatomical structure information of the primary gastric cancer lesion. The volume of interest (VRO) of the primary gastric cancer lesion is segmented from the preoperative images. Using 40% of the SUVmax value of the primary lesion as a threshold, the region of the primary gastric cancer lesion is delineated on the images to determine the target area. SUVmax is a key indicator for measuring the glucose metabolic activity of tumor tissue, representing the voxel value with the highest radioactive uptake within the lesion. Using 40% of SUVmax as a threshold ensures complete coverage of the main active metabolic areas of the tumor, while reducing bias caused by partial volume effects or interference from low-metabolic areas, ensuring that the VRO accurately reflects the overall metabolic distribution of the tumor. This method can distinguish tumor tissue from surrounding normal tissue, avoiding interference from normal tissue signals in the analysis results. It provides a reliable spatial range for the extraction of PET metabolic features, PET texture features, and CT anatomical features, thus providing accurate imaging basis data for constructing predictive models of gastric cancer lymph node metastasis risk. This segmentation method can preserve the integrity of the active tumor region while excluding interference from non-metabolic necrotic tissue and surrounding normal tissue. It ensures that feature extraction is based on biologically active tumor tissue, laying a solid foundation for the extraction, screening, and model construction of radiomics features, thereby ensuring the accuracy and reliability of the gastric cancer lymph node metastasis risk assessment process.
[0036] Extracting the target image omics features corresponding to the volume of interest includes the following steps:
[0037] Mann-Whitney U was used to screen candidate features with differences in the volume of interest;
[0038] Candidate features are processed by dimensionality reduction using the minimum absolute shrinkage and selection operator algorithm to obtain the target image omics features.
[0039] Radiomics features include at least one of PET metabolic features, PET texture features, and CT anatomical features;
[0040] PET metabolic characteristics include at least one of the following: maximum standard uptake, average standard uptake, tumor metabolic volume, and total glycolysis in the lesion.
[0041] PET texture features include at least one of the following: gray-level co-occurrence matrix features, gray-level run-length matrix features, neighborhood gray-level difference matrix features, and gray-level region length matrix features;
[0042] CT anatomical features include at least one of shape features and density features.
[0043] The Mann-Whitney U method was used to initially screen all radiomics features extracted from the volume of interest. This test can effectively distinguish the differences in feature values under different lymph node metastasis states, screen candidate features with statistical differences, and exclude noisy features unrelated to lymph node metastasis. The candidate features were then subjected to dimensionality reduction using a minimum absolute shrinkage and selection operator algorithm. This algorithm can automatically compress feature dimensions, eliminate redundant and highly collinear features, and finally obtain the target radiomics features highly correlated with lymph node metastasis in gastric cancer.
[0044] The target radiomics features include PET metabolic features, PET texture features, and CT anatomical features. Among them, PET metabolic features mainly reflect the glucose metabolic activity of the tumor, including maximum standardized uptake (MSU), mean standardized uptake (MSU), tumor metabolic volume, and total glycolysis in the lesion. The maximum standardized uptake is the voxel value with the highest radioactive uptake within the lesion, which can reflect the highest metabolic activity in the local tumor. The mean standardized uptake reflects the overall metabolic level of the tumor. The tumor metabolic volume represents the volume of tumor tissue with high metabolic activity. The total glycolysis in the lesion combines the metabolic volume and the mean standardized uptake to comprehensively reflect the overall glycolytic level of the tumor. PET texture features reveal the heterogeneity within a tumor by determining the spatial distribution of pixel grayscale within the lesion. These features include grayscale co-occurrence matrix (GCMM), grayscale run-length matrix (GLLM), neighborhood grayscale difference matrix (NMM), and grayscale region length matrix. The GCMM describes the grayscale distribution relationship between pixel pairs, reflecting the texture roughness and regularity within the tumor. The GLLM reflects the length distribution of consecutive pixels with the same grayscale value, reflecting the spatial arrangement pattern of the tumor tissue. The NMM determines the grayscale differences between adjacent pixels, reflecting local heterogeneity within the tumor. The GLLM statistically analyzes the size distribution of different grayscale regions, further revealing the spatial structural features of the tumor. CT anatomical features reflect the anatomical characteristics of the tumor from both morphological and density perspectives. These include shape and density features. Shape features describe the overall morphological parameters of the tumor, such as volume, sphericity, and surface area, reflecting the tumor's growth pattern and invasiveness. Density features, by analyzing the distribution of CT values, reflect the differences in tissue composition within the tumor, such as necrosis, cystic degeneration, or fibrosis. It can quantify the biological characteristics of primary gastric cancer lesions from multiple dimensions, providing a stable feature basis for constructing a lymph node metastasis prediction model.
[0045] Image omics feature scores are obtained based on the target image omics features;
[0046] Radiomics feature scores are calculated based on target radiomics features. The radiomics feature score is a quantitative index obtained by linearly combining selected key radiomics features. Each target radiomics feature is assigned a corresponding weight value based on its regression coefficient obtained in the minimum absolute contraction and selection operator algorithm. The feature value is then multiplied by its corresponding weight and summed to obtain the radiomics feature score. This score integrates information from all key features in PET metabolic features, PET texture features, and CT anatomical features, comprehensively reflecting the biological behavior and invasive potential of the primary gastric cancer lesion from multiple dimensions, including internal heterogeneity of tumor metabolic activity and anatomical morphology. As a continuous quantitative index, the radiomics feature score can transform complex, multi-dimensional imaging feature information into a single risk quantification value, providing a unified input variable for constructing a lymph node metastasis prediction model combined with clinical risk factors, thereby achieving a quantitative assessment of the risk of lymph node metastasis in gastric cancer.
[0047] By combining radiomics feature scores with pre-defined clinical risk factors, a lymph node metastasis prediction model was constructed and the results of gastric cancer lymph node metastasis risk were output.
[0048] The risk outcome of lymph node metastasis in gastric cancer includes the classification of whether or not lymph node metastasis is present, as well as the probability value of lymph node metastasis in N2-3b stage.
[0049] The lymph node metastasis prediction model is a logistic regression model.
[0050] A lymph node metastasis prediction model in the form of a logistic regression model was constructed by combining radiomics feature scores with pre-defined clinical risk factors, ultimately outputting the risk of lymph node metastasis in gastric cancer. The logistic regression model uses radiomics feature scores and pre-defined clinical risk factors as input variables. These pre-defined clinical risk factors include at least one of carbohydrate antigen 199, carcinoembryonic antigen, and traditional PET-CT lymph node metastasis diagnostic results. These factors are all clinically closely related to lymph node metastasis in gastric cancer, supplementing clinical information not covered by radiomics scores from the perspectives of laboratory examinations and traditional imaging diagnosis. The model constructs a multivariate prediction equation by assigning corresponding regression coefficients to radiomics feature scores and each clinical risk factor, thereby comprehensively quantifying the influence of each factor on lymph node metastasis in gastric cancer. The model output includes two types of results: one is a classification result indicating the presence or absence of lymph node metastasis, which can directly determine whether the target subject has lymph node metastasis; the other is a probability value of N2 to 3b stage lymph node metastasis, which can quantify the risk of the target subject developing intermediate to late-stage lymph node metastasis. It can provide clear binary judgment criteria for clinical practice, and can also reflect the severity of metastasis risk through probability values, helping clinicians to more comprehensively assess the invasiveness of gastric cancer.
[0051] It also includes validating the lymph node metastasis prediction model by dividing the dataset into a training set, an internal validation set, and an external validation set, and evaluating the discrimination and clinical applicability of the lymph node metastasis prediction model through receiver operating characteristic (ROC) curves and decision curves.
[0052] Pre-defined clinical risk factors can supplement radiomics scores from different dimensions, improving the model's predictive efficacy. Carbohydrate antigen 199 (CA199) is a serum marker closely related to gastrointestinal tumors; elevated CA199 levels typically indicate increased tumor invasiveness and are significantly associated with the risk of lymph node metastasis in gastric cancer. Carcinoembryonic antigen (CEA), as a broad-spectrum tumor marker, reflects the proliferation and metastatic potential of tumors; changes in its expression level can provide important reference for assessing the risk of lymph node metastasis in gastric cancer. Traditional PET-CT lymph node metastasis diagnostic results are commonly used imaging assessment indicators in clinical practice. By directly observing the size, morphology, and metabolic activity of lymph nodes, a preliminary judgment of lymph node metastasis is made, providing the model with intuitive clinical imaging evidence. These clinical risk factors complement the radiomics feature scores, enriching the model's input information from both serological and traditional imaging perspectives, thus more comprehensively reflecting the lymph node metastasis risk in gastric cancer patients.
[0053] To ensure the reliability and generalization ability of the model, rigorous validation of the lymph node metastasis prediction model is necessary. First, the dataset is divided into a training set, an internal validation set, and an external validation set. The training set is used for model construction and parameter optimization, the internal validation set is used to evaluate the model's initial performance under the same data distribution, and the external validation set uses data from different sources to test the model's generalization ability in different scenarios. The model's discriminative power is evaluated using receiver operating characteristic (ROC) curves, which intuitively reflect the model's ability to distinguish between metastatic and non-metastatic cases; a larger area under the curve indicates better discriminative performance. Simultaneously, the clinical applicability of the model is evaluated using decision curves. These curves assess the model's net benefit at different thresholds, determining its practical application value in clinical decision-making and effectively avoiding over-reliance on statistical indicators while neglecting clinical benefits. This comprehensive evaluation of the model's performance ensures its reliable discriminative power and practical value in real-world clinical applications, providing a guarantee for the accurate assessment of the risk of lymph node metastasis in gastric cancer.
[0054] It also includes integrating radiomics feature scores with clinical risk factors into a nomograph, enabling the visualization of gastric cancer lymph node metastasis risk results.
[0055] To enable clinicians to use the model results intuitively and conveniently, this solution integrates radiomics feature scores with clinical risk factors into a nomograph, enabling the visualization of gastric cancer lymph node metastasis risk results. The nomograph is a visualization tool based on a multivariate regression model, converting the regression coefficients of each variable in the logistic regression model into intuitive scales and line segments. Each predictor variable corresponds to a line segment with a score scale, the value of which is determined by the variable's influence on lymph node metastasis. In use, clinicians simply locate the corresponding position on the line segment based on the patient's radiomics feature score and the values of each clinical risk factor, read the corresponding score, sum it, and then directly read the probability of lymph node metastasis and the risk value of N2 to 3b stage lymph node metastasis from the total score scale. This approach transforms the complex model calculation process into a simple scoring and reading operation, eliminating the need for doctors to master professional statistical analysis methods, and quickly obtaining accurate risk assessment results. This greatly improves the clinical practicality and ease of use of the model, while also facilitating doctors' explanation of the risk to patients, providing an intuitive basis for doctor-patient communication.
[0056] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0057] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.
[0058] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for assessing lymph node metastasis in gastric cancer based on PET / CT, characterized in that, Includes the following steps: Acquire preoperative images of the target subject and segment the volume of interest of the primary gastric cancer lesion from the preoperative images; Extract the target radiomics features corresponding to the volume of interest, wherein the radiomics features include at least one of PET metabolic features, PET texture features, and CT anatomical features; Image omics feature scores are obtained based on the target image omics features; By combining radiomics feature scores with pre-defined clinical risk factors, a lymph node metastasis prediction model was constructed and the results of gastric cancer lymph node metastasis risk were output.
2. The method for assessing lymph node metastasis in gastric cancer based on PET / CT according to claim 1, characterized in that, The PET metabolic characteristics include at least one of the following: maximum standard uptake value, average standard uptake value, tumor metabolic volume, and total glycolysis in the lesion. The PET texture features include at least one of the following: gray-level co-occurrence matrix features, gray-level run-length matrix features, neighborhood gray-level difference matrix features, and gray-level region length matrix features. The CT anatomical features include at least one of shape features and density features.
3. The method for assessing lymph node metastasis in gastric cancer based on PET / CT according to claim 1, characterized in that, The volume of interest of the primary gastric cancer lesion is segmented from preoperative images, specifically as follows: The volume of interest was delineated from the preoperative images with a threshold of SUVmax≥40% for the primary lesion.
4. The method for assessing lymph node metastasis in gastric cancer based on PET / CT according to claim 1, characterized in that, Extracting the target image omics features corresponding to the volume of interest includes the following steps: Mann-Whitney U was used to screen candidate features with differences in the volume of interest; Candidate features are processed by dimensionality reduction using the minimum absolute shrinkage and selection operator algorithm to obtain the target image omics features.
5. The method for assessing lymph node metastasis in gastric cancer based on PET / CT according to claim 1, characterized in that, Pre-defined clinical risk factors include at least one of carbohydrate antigen 199, carcinoembryonic antigen, and conventional PET / CT lymph node metastasis diagnosis results.
6. The method for assessing lymph node metastasis in gastric cancer based on PET / CT according to claim 1, characterized in that, The lymph node metastasis prediction model is a logistic regression model.
7. The method for assessing lymph node metastasis in gastric cancer based on PET / CT according to claim 1, characterized in that, It also includes validating the lymph node metastasis prediction model by dividing the dataset into a training set, an internal validation set, and an external validation set, and evaluating the discrimination and clinical applicability of the lymph node metastasis prediction model through receiver operating characteristic (ROC) curves and decision curves.
8. The method for assessing lymph node metastasis in gastric cancer based on PET / CT according to claim 1, characterized in that, The risk of lymph node metastasis in gastric cancer includes a classification result for the presence or absence of lymph node metastasis, as well as the probability value of lymph node metastasis in N2-3b stage.
9. The method for assessing gastric cancer lymph node metastasis based on PET / CT according to claim 8, characterized in that, It also includes integrating radiomics feature scores with clinical risk factors into a nomograph, enabling the visualization of gastric cancer lymph node metastasis risk results.