CN112071418B - Gastric cancer peritoneal metastasis prediction system and method based on enhanced CT image histology - Google Patents
Gastric cancer peritoneal metastasis prediction system and method based on enhanced CT image histology Download PDFInfo
- Publication number
- CN112071418B CN112071418B CN202010455571.XA CN202010455571A CN112071418B CN 112071418 B CN112071418 B CN 112071418B CN 202010455571 A CN202010455571 A CN 202010455571A CN 112071418 B CN112071418 B CN 112071418B
- Authority
- CN
- China
- Prior art keywords
- image
- gastric cancer
- value
- patient
- module
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10081—Computed x-ray tomography [CT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30092—Stomach; Gastric
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30096—Tumor; Lesion
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Biomedical Technology (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Quality & Reliability (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Radiology & Medical Imaging (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Pathology (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Epidemiology (AREA)
- Primary Health Care (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The invention discloses a gastric cancer peritoneal metastasis prediction system based on enhanced CT image histology, which comprises an enhanced CT image input module, an enhanced CT image preprocessing module, a feature extraction module, a data processing module and a prediction and result output module. The invention also discloses a prediction method of gastric cancer peritoneal metastasis based on enhanced CT image histology, which comprises the following steps: inputting an enhanced CT image; carrying out imaging treatment; manually selecting an image and manually marking the image characteristics; identifying the marked image features; extracting three characteristic data sets; regression analysis and scoring; and qualitatively analyzing the grading value and outputting a prediction result. The invention improves the preoperative prediction capability and prediction accuracy of gastric cancer peritoneal metastasis, has the advantages of no wound, intuitiveness and easy operation, provides more and more accurate decision information for clinicians before operation, and provides more efficient support for whether to perform operation and making accurate treatment scheme before operation.
Description
Technical Field
The invention relates to the technical field of artificial intelligence and medical image analysis, in particular to a gastric cancer peritoneal metastasis prediction system and method based on enhanced CT image histology.
Background
Gastric Cancer (GC) is one of the most common malignant tumors in the world and is the third leading cause of cancer-related death in the world, and the high incidence and mortality of Gastric cancer place a great economic burden worldwide. In our country, the number of new cases of gastric cancer per year is about more than half of the world, and gastric cancer in the progressive stage is the main, among which gastric cancer peritoneal metastasis (Peritoneal metastasis, PM) is an important cause of poor prognosis. Whether the patient has gastric cancer peritoneal metastasis can be accurately grasped before an operation is always the key of research of researchers and is a troublesome problem which afflicts clinicians for many years. The method has the advantages that the peritoneal metastasis condition of the gastric cancer patient can be effectively predicted before operation, more decision information can be provided for clinicians, whether operation and the operation mode are selected or not is guided, the proper treatment population and treatment mode are selected maximally, the disease is timely intervened and treated, and powerful support is provided for effective diagnosis and treatment. The current clinical diagnosis of gastric cancer peritoneal metastasis is mainly based on imaging, tumor markers and cytological examination of peritoneal lavage fluid. However, a large number of studies have shown that for smaller metastases, it is difficult for imaging physicians to make accurate decisions directly from imaging examinations, and that diagnosis of tumor markers and peritoneal cell lavages is lacking in sufficient specificity and sensitivity. There is currently no clinically effective means for predicting peritoneal metastasis from gastric cancer prior to surgery.
Enhancement CT (computed tomography) has been widely used in the auxiliary diagnosis of gastric cancer as a noninvasive early tumor diagnosis method. At present, the utilization of the enhanced CT image information is mainly judged by subjective judgment of an imaging doctor and corresponding diagnosis is given. However, a considerable amount of feature information still present in medical image pictures is yet to be developed and utilized. Different individuals have different characteristic information expressed on the enhanced CT image due to different pathological characteristics, so that the disease state of the patient can be predicted by the characteristic multidimensional texture features of the CT image before operation, and powerful assistance is provided for accurate treatment of gastric cancer. Therefore, the gastric cancer peritoneal metastasis prediction system and method based on the enhanced CT image histology are established, and the prediction of metastasis conditions before surgery has extremely high practical significance.
Disclosure of Invention
In view of the above, it is necessary to provide a prediction system and method for peritoneal metastasis of gastric cancer based on enhanced CT imaging histology, which can provide powerful assistance for accurate treatment of gastric cancer, provide accurate prediction before gastric cancer surgery, and reduce risks caused by subjective judgment of imaging doctors.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
a prediction system for peritoneal metastasis from gastric cancer based on enhanced CT imaging histology, the prediction system comprising: the device comprises an enhanced CT image input module, an enhanced CT image preprocessing module, a feature extraction module, a data processing module and a prediction and result output module; the enhanced CT image input module is used for inputting enhanced CT images for detecting peritoneal metastasis of gastric cancer; the enhanced CT image preprocessing module is used for carrying out imaging processing on the enhanced CT image and identifying the image characteristics which are manually selected and marked; the feature extraction module is used for extracting three groups of feature data sets from the identified image features, wherein the first group of feature data sets comprise a plurality of intensity feature data, the second group of feature data sets comprise a plurality of morphological feature data, and the third group of feature data sets comprise a plurality of gray texture feature data; the intensity characteristic data are used for reflecting image intensity information of gastric cancer tumor focus on a CT image; the morphological characteristic data are used for reflecting morphological information of gastric cancer tumor focus; the gray texture characteristic data are used for reflecting voxel space distribution intensity level information of gastric cancer focus on a CT image and presenting characteristic information of the surface and the interior of tissue corresponding to the gastric cancer focus; the data processing module is used for carrying out regression analysis and scoring on each characteristic data and transmitting the scoring numerical value to the prediction and result output module; the prediction and result output module performs qualitative analysis and prediction result output on the obtained scoring value;
the data processing module comprises a characteristic variable screening sub-module, a variable coefficient acquisition sub-module and a scoring calculation sub-module; the characteristic variable screening submodule screens out a plurality of characteristic variables related to gastric cancer peritoneal transfer from the characteristic data through a LASSO COX regression calculation model; the variable coefficient acquisition submodule acquires regression coefficients corresponding to the characteristic variables one by one through the LASSO COX regression calculation model; the scoring computation submodule obtains a scoring value through a scoring computation model; the formula of the scoring calculation model is Rad-score=Σ (regression coefficient characteristic variable);
the prediction and result output module comprises a patient score matching sub-module, a cut-off value acquisition sub-module, a probability classification sub-module and a result output sub-module; the patient score matching sub-module matches the statistical data of whether the peritoneal transfer occurs to each patient with the corresponding score value, and transmits the matching result to the truncated value acquisition sub-module; the truncated value acquisition submodule analyzes the matching result and generates a truncated value which is used as a judging threshold value of the probability of peritoneal metastasis of the patient; the probability classification submodule is used for comparing the score value of each patient with the cut-off value and taking the comparison result as a judgment result with high or low illness probability.
Further, the LASSO COX regression calculation model is generated by a glmcet function package in the computer programming tool R language.
Further, each characteristic variable is Eccentricity eccenttricity, range extension, gray level co-occurrence matrix_correlation information measurement GLCM_IMC and gray level co-occurrence matrix_maximum probability GLCM_maximum probability; the Eccentricity eccenttricity is an image morphological feature variable and is used for reflecting original stomach cancer focus morphological information in a CT image; the range extension is an image morphological characteristic variable and is used for reflecting the size of the area range of the original gastric cancer focus in the CT image; the gray level co-occurrence matrix-correlation information measure GLCM-IMC is an image based on gray level texture characteristic variables and is used for reflecting the correlation of pixels of an area image where a gastric cancer focus is located in a CT image in each vector direction; the gray level co-occurrence matrix-maximum probability GLCM-maximum probability is an image based on gray texture characteristic variables and is used for reflecting the spatial change of the pixel intensity level of an image in the area where the gastric cancer focus is located in the CT image and stably reflecting the image information covered by a specific focus tissue.
Further, the patient score matching submodule calls MedCalc medical statistics analysis software to match the statistics data of whether peritoneal transfer occurs to each patient with the corresponding score value, and transmits the matching result to the cut-off value acquisition submodule; and the cut-off value acquisition submodule calls MedCalc medical statistics analysis software to analyze the matching result and generate a cut-off value which is used as a judging threshold value of the probability of peritoneal metastasis of the patient.
Further, when the patient's score value < cutoff value, then the patient's probability of developing peritoneal metastasis is low; when the scoring value of the patient is larger than or equal to the cutoff value, the probability of peritoneal metastasis of the patient is high.
A method for predicting gastric cancer peritoneal metastasis based on enhanced CT image histology, comprising the steps of:
s1, inputting an enhanced CT image for detecting peritoneal metastasis of gastric cancer;
s2, performing imaging processing on the input enhanced CT image;
s3, manually selecting the image obtained in the S2, and manually marking the image characteristics of the selected image;
s4, identifying the marked image features;
s5, extracting three groups of characteristic data sets from the identified image characteristics, wherein the first group of characteristic data sets comprise a plurality of intensity characteristic data, the second group of characteristic data sets comprise a plurality of morphological characteristic data, and the third group of characteristic data sets comprise a plurality of gray texture characteristic data; the intensity characteristic data are used for reflecting image intensity information of gastric cancer tumor focus on a CT image; the morphological characteristic data are used for reflecting morphological information of gastric cancer tumor focus; the gray texture characteristic data are used for reflecting voxel space distribution intensity level information of gastric cancer focus on a CT image and presenting characteristic information of the surface and the interior of tissue corresponding to the gastric cancer focus;
s6, carrying out regression analysis and scoring on the characteristic data obtained in the step S5;
and S7, carrying out qualitative analysis and prediction result output on the score value obtained in the step S6.
Further, S6 includes the following steps:
s61, screening a plurality of characteristic variables related to gastric cancer peritoneal transfer from the characteristic data by using a LASSO COX regression calculation model, and obtaining regression coefficients corresponding to the characteristic variables one by one;
s62, obtaining a scoring value by using a scoring calculation model; the formula of the score calculation model is Rad-score=Σ (regression coefficient characteristic variable).
Further, S7 includes the following steps:
s71, matching the statistical data of whether peritoneal metastasis occurs in each patient with corresponding scoring values;
s72, analyzing the matching result and generating a cut-off value which is used as a judging threshold value of the probability of peritoneal metastasis of the patient;
and S73, comparing the score value of each patient with the cut-off value, and taking the comparison result as a judgment result with high or low illness probability.
Further, the LASSO COX regression calculation model is generated by a glmcet function package in the computer programming tool R language.
Further, in S71, a MedCalc medical statistics analysis software is called to match the statistics of whether peritoneal metastasis occurs to each patient with the corresponding scoring values;
in S72, the medical statistical analysis software of the MedCalc is called to analyze the matching result and generate a cut-off value, and the cut-off value is used as a judging threshold value of the probability of peritoneal metastasis of the patient;
in S73, when the patient' S score value < the cutoff value, it is determined that the patient has a low probability of developing peritoneal metastasis; when the scoring value of the patient is more than or equal to the cutoff value, the probability of the patient to generate peritoneal metastasis is judged to be high.
The beneficial effects of the invention are as follows:
the method has higher use value in preoperative prediction of gastric cancer peritoneal metastasis, has a prediction effect superior to that of the traditional TNM stage alone, makes up the defects of the current technology, provides a new thought for the current clinical preoperative diagnosis of gastric cancer peritoneal metastasis due to the advantage of preoperative noninvasiveness, and reduces the pain and economic burden caused by the invasive diagnosis of disease implementation; as an image histology auxiliary diagnosis technology, the invention has the advantages of no wound, intuitiveness, easy operation and high prediction accuracy, overcomes the defects of the current conventional diagnosis method, and can provide a set of noninvasive prediction scheme for clinical diagnosis of gastric cancer patients before operation development, thereby providing more and more accurate decision information for clinicians before operation and providing more efficient support for whether to develop operation and making accurate treatment scheme before operation.
Drawings
FIG. 1 is a functional block diagram of a prediction system for peritoneal metastasis of gastric cancer based on enhanced CT image histology according to the present invention;
FIG. 2 is a predictive analysis chart of a gastric cancer peritoneal metastasis prediction system and method applied to a training set based on enhanced CT image histology according to the present invention;
FIG. 3 is a predictive analysis graph of a gastric cancer peritoneal metastasis prediction system and method based on enhanced CT imaging histology applied to an internal validation set in accordance with the present invention;
FIG. 4 is a predictive analysis graph of a gastric cancer peritoneal metastasis prediction system and method based on enhanced CT imaging histology applied to an external validation set in accordance with the present invention;
reference numerals illustrate:
1. enhancing a CT image input module; 2. the CT image preprocessing module is enhanced; 3. a feature extraction module; 4. a data processing module; 5. a prediction and result output module; 41. a feature variable screening sub-module; 42. a variable coefficient acquisition sub-module; 43. a score calculation sub-module; 51. a patient score matching sub-module; 52. a truncated value acquisition sub-module; 53. a probability classification sub-module; 54. and outputting a result.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be further clearly and completely described in the following in conjunction with the embodiments of the present invention. It should be noted that the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Example 1
This embodiment is shown in fig. 1:
the present embodiment provides a prediction system for peritoneal metastasis of gastric cancer based on enhanced CT imaging, the prediction system comprising: the device comprises an enhanced CT image input module 1, an enhanced CT image preprocessing module 2, a feature extraction module 3, a data processing module 4 and a prediction and result output module 5; the enhanced CT image input module 1 is used for inputting enhanced CT images for detecting peritoneal metastasis of gastric cancer; the enhanced CT image preprocessing module 2 is used for carrying out imaging processing on the enhanced CT image and identifying the image characteristics which are manually selected and marked; the feature extraction module 3 is configured to extract three sets of feature data sets from the identified image features, where the first set of feature data sets includes a plurality of intensity feature data, the second set of feature data sets includes a plurality of morphological feature data, and the third set of feature data sets includes a plurality of gray texture feature data; the intensity characteristic data are used for reflecting image intensity information of gastric cancer tumor focus on a CT image; the morphological characteristic data are used for reflecting morphological information of gastric cancer tumor focus; the gray texture characteristic data are used for reflecting voxel space distribution intensity level information of gastric cancer focus on a CT image and presenting characteristic information of the surface and the interior of tissue corresponding to the gastric cancer focus; the data processing module 4 is used for carrying out regression analysis and scoring on each characteristic data and transmitting the scoring numerical value to the prediction and result output module 5; the prediction and result output module 5 performs qualitative analysis and prediction result output on the obtained scoring value; specifically, in all the gray texture characteristic data, the gray texture characteristic data can be divided into a plurality of gray level co-occurrence matrix characteristic data, a plurality of gray level run-length matrix characteristic data, a plurality of gray level size area matrix characteristic data and a plurality of neighborhood gray level difference matrix characteristic data, and the gray texture characteristic data plays a role in stably expressing information required by prediction through grasping different characteristic information;
the data processing module 4 comprises a characteristic variable screening sub-module 41, a variable coefficient obtaining sub-module 42 and a scoring calculating sub-module 43; the feature variable screening submodule 41 screens out a plurality of feature variables related to gastric cancer peritoneal transfer from the feature data through a LASSO COX regression calculation model; the variable coefficient obtaining submodule 42 obtains regression coefficients corresponding to the characteristic variables one by one through the LASSO COX regression calculation model; the scoring computation submodule 43 obtains a scoring value through a scoring computation model; the formula of the scoring calculation model is Rad-score=Σ (regression coefficient characteristic variable);
the prediction and result output module 5 comprises a patient score matching sub-module 51, a cut-off value acquisition sub-module 52, a probability classification sub-module 53 and a result output sub-module 54; the patient score matching sub-module 51 matches the statistics of whether peritoneal transfer occurs to each patient with the corresponding score value, and transmits the matching result to the cut-off value obtaining sub-module 52; the truncated value obtaining sub-module 52 analyzes the matching result and generates a truncated value as a judgment threshold for the probability of peritoneal metastasis of the patient; the probability classification sub-module 53 is configured to compare the score value of each patient with the cut-off value, and use the comparison result as a judgment result with high or low probability of being ill.
Further, the LASSO COX regression calculation model is generated by a glmcet function package in the computer programming tool R language.
Further, each characteristic variable is Eccentricity eccenttricity, range extension, gray level co-occurrence matrix_correlation information measurement GLCM_IMC and gray level co-occurrence matrix_maximum probability GLCM_maximum probability; the Eccentricity eccenttricity is an image morphological feature variable and is used for reflecting original stomach cancer focus morphological information in a CT image; the range extension is an image morphological characteristic variable and is used for reflecting the size of the area range of the original gastric cancer focus in the CT image; the gray level co-occurrence matrix-correlation information measure GLCM-IMC is an image based on gray level texture characteristic variables and is used for reflecting the correlation of pixels of an area image where a gastric cancer focus is located in a CT image in each vector direction; the gray level co-occurrence matrix-maximum probability GLCM-maximum probability is a characteristic variable based on gray level texture of an image which presents local image characteristics on visual perception level on a CT image of a stomach cancer focus, is used for reflecting the spatial change of the pixel intensity level of an image in the area where the stomach cancer focus is located in the CT image, has stronger noise reduction capability, and can stably reflect the image information covered by specific focus tissues.
Further, the patient score matching sub-module 51 invokes the MedCalc medical statistics analysis software to match the statistics of whether peritoneal metastasis occurs in each patient with the corresponding score value, and transmits the matching result to the cut-off value obtaining sub-module 52; the truncated value obtaining sub-module 52 calls the MedCalc medical statistics analysis software to analyze the matching result and generate a truncated value, wherein the truncated value is used as a judging threshold value of the probability of peritoneal metastasis of the patient;
further, when the patient's score value < cutoff value, then the patient's probability of developing peritoneal metastasis is low; when the scoring value of the patient is larger than or equal to the cutoff value, the probability of peritoneal metastasis of the patient is high.
Example 2
The embodiment provides a prediction method of gastric cancer peritoneal metastasis based on enhanced CT image histology, which comprises the following steps:
s1, inputting an enhanced CT image for detecting peritoneal metastasis of gastric cancer; specifically, 562 gastric cancer patients subjected to gastric cancer surgery excision in 2016-2018-12 are selected as a training group, 106 gastric cancer patients subjected to gastric cancer surgery in 2019-1-2019-12 are selected as an internal verification group, 287 gastric cancer patients subjected to surgery in 2013-1-2019-12 in other tumor centers are selected as an external verification group, preoperative enhanced CT images of the patients are selected, and the external verification group is used for verifying whether a model is reliable or not;
s2, performing imaging processing on the input enhanced CT image; specifically, the input enhanced CT image is subjected to imaging processing to obtain a portal vein image (the format is dicom), and the obtained portal vein image of the patient is imported into computer software ITK-SNAP (v 3.4) for image preprocessing;
s3, manually selecting the image obtained in the S2, and manually marking the image characteristics of the selected image; specifically, a doctor of an experienced imaging department selects a portal vein image with the largest gastric cancer focus cross-sectional area, and delineates a region of interest (all gastric cancer focus regions) ROI (Region of Interest) on the portal vein image, and the output format is nii;
s4, identifying the marked image features; the indexes of the image features comprise Eccentricity Eccenttricity, range extension, gray level co-occurrence matrix_correlation information measurement GLCM_IMC and gray level co-occurrence matrix_maximum probability GLCM_maximum probability, wherein the Eccentricity Eccenttricity is an image morphological feature variable and is used for reflecting original stomach cancer focus morphological information in CT images; the range extension is an image morphological characteristic variable and is used for reflecting the size of the area range of the original gastric cancer focus in the CT image; the gray level co-occurrence matrix-correlation information measure GLCM-IMC is an image based on gray level texture characteristic variables and is used for reflecting the correlation of pixels of an area image where a gastric cancer focus is located in a CT image in each vector direction; the gray level co-occurrence matrix-maximum probability GLCM-maximum probability is a characteristic variable based on gray level texture of an image which presents local image characteristics on visual perception level on a CT image of a stomach cancer focus, is used for reflecting the spatial change of the pixel intensity level of an image in the area where the stomach cancer focus is located in the CT image, has stronger noise reduction capability, and can stably reflect the image information covered by specific focus tissues.
S5, extracting three groups of characteristic data sets from the identified image characteristics, wherein the first group of characteristic data sets comprise a plurality of intensity characteristic data, the second group of characteristic data sets comprise a plurality of morphological characteristic data, and the third group of characteristic data sets comprise a plurality of gray texture characteristic data; the intensity characteristic data are used for reflecting image intensity information of gastric cancer tumor focus on a CT image; the morphological characteristic data are used for reflecting morphological information of gastric cancer tumor focus; the gray texture characteristic data are used for reflecting voxel space distribution intensity level information of gastric cancer focus on a CT image and presenting characteristic information of the surface and the interior of tissue corresponding to the gastric cancer focus; specifically, the ROI image is imported into computer software Matlab (2016 a), the image features of the region of interest are extracted by using the automatic extraction function of the Matlab (2016 a) software, 292 feature data are extracted from each patient image, and the three types of features include intensity feature data, morphological feature data and gray texture feature data;
s6, carrying out regression analysis and scoring on the characteristic data obtained in the step S5;
and S7, carrying out qualitative analysis and prediction result output on the score value obtained in the step S6.
Further, S6 includes the following steps:
s61, screening a plurality of characteristic variables related to gastric cancer peritoneal transfer from the characteristic data by using a LASSO COX regression calculation model, and obtaining regression coefficients corresponding to the characteristic variables one by one;
s62, obtaining a scoring value by using a scoring calculation model; the formula of the scoring calculation model is Rad-score=Σ (regression coefficient characteristic variable);
specifically, as the image feature information quantity of the corresponding enhanced CT image is huge, a large number of irrelevant variables are not mixed, in order to reduce the variable dimension of a model and screen a series of variables with the highest correlation degree with gastric cancer peritoneal metastasis, the model establishment is more accurate, variable screening is carried out through a LASSO COX regression calculation model, and the regression coefficient of the relevant variables is calculated, wherein the regression model operates through a glmnet function in a glmnet function package in an R language 3.3.1 version, and the regression coefficient of the relevant variables is obtained through function operation; the formula of the scoring calculation model is specifically: rad-score= (-1.429359 e-03) eccentricity+ (1.232216 e-02) extension- (9.887834 e-02) glcm_imc+ (8.977322 e-02) glcm_maximumprobability.
Further, S7 includes the following steps:
s71, matching the statistical data of whether peritoneal metastasis occurs in each patient with corresponding scoring values;
s72, analyzing the matching result and generating a cut-off value which is used as a judging threshold value of the probability of peritoneal metastasis of the patient;
s73, comparing the score value of each patient with the cut-off value, and taking the comparison result as a judgment result with high or low illness probability;
specifically, according to the magnitude of the GC Rad-score value, the specific individuals of the image information are respectively subjected to qualitative predictive analysis, the specific qualitative analysis process is as follows, the actual peritoneal transfer condition (whether or not) of each patient in the training group is matched with the specific Rad-score value corresponding to the patient in the training group by using MedCalc V12.7 software, the cut-off value is calculated and obtained to be 0.0002 by using MedCalc V12.7 software, when the patient is more than or equal to 0.0002, the patient is qualified to have peritoneal transfer, when the patient is less than 0.0002, the patient is qualified to have no peritoneal transfer, and a reliable predictive analysis result is provided before operation.
Further, the LASSO COX regression calculation model is generated by a glmcet function package in the computer programming tool R language.
Further, in S71, a MedCalc medical statistics analysis software is called to match the statistics of whether peritoneal metastasis occurs to each patient with the corresponding scoring values;
in S72, the MedCalc medical statistics analysis software is invoked to analyze the matching result and generate a cut-off value, which is used as a judgment threshold for the probability of occurrence of peritoneal metastasis in the patient.
Further, in S73, when the score value of the patient is < the cutoff value, it is determined that the probability of occurrence of peritoneal metastasis of the patient is low; when the scoring value of the patient is more than or equal to the cutoff value, the probability of the patient to generate peritoneal metastasis is judged to be high. As shown in fig. 2-4, using the subject operating characteristic curve (ROC curve), the area under the ROC curve (AUC) was calculated, the accuracy of GC Rad-score in predicting peritoneal metastasis was examined, the higher the AUC value, the higher the accuracy of prediction was indicated, the higher the AUC of Rad-score was found to be in both training and validation groups and superior to TNM staging, and in multi-factor COX analysis, GC Rad-score could be used as a factor for independently predicting peritoneal metastasis from gastric cancer (P < 0.001); as shown in table 1, PM (+) indicates positive peritoneal transfer, PM (-) indicates negative peritoneal transfer, cN stage indicates clinical N-stage, cT stage indicates clinical T-stage; in the training group, the internal verification group and the external verification group, AUC values of the Rad-score are higher than those of the cN stage and the cT stage, which shows that the Rad-score is used as a prediction factor of peritoneal metastasis, and the effect is superior to that of the clinical single TNM stage prediction of peritoneal metastasis); GC Rad-score can be effectively used for prediction of peritoneal metastasis of gastric cancer. In fig. 2-4, the tracking method: training groups; internal validation cohort: an internal authentication group; external validation cohort: an external validation group; RS: rad-score: scoring the image group; PM (+): positive peritoneal metastasis; PM (-): peritoneal metastasis is negative; true positive rate: correct prediction rate.
Table 1:
multifactorial analysis is related to peritoneal metastasis of gastric cancer.
External validation group:
the working principle of the invention is as follows:
the invention utilizes the characteristic information which is not developed and utilized in the CT image and contains disease characteristics to assist a clinician in predicting the disease state of a patient, thereby determining whether to perform surgery or not and assisting in preparing an accurate treatment scheme; the invention uses reliable regression calculation model and grading calculation model to make the flow from data input to result output simple and reliable, and has specific judgment mark in the operation process, so that the operation process is simple, visual and easy to repeat, and the common technicians can complete the operation; the invention utilizes a plurality of image feature combination technologies to form a preoperative noninvasive detection technology with more diagnostic significance, performs qualitative predictive analysis on specific individuals, and provides reliable prediction and analysis results for preoperative noninvasive diagnosis.
The foregoing examples illustrate only a few embodiments of the invention and are described in detail herein without thereby limiting the scope of the invention. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the invention, which are all within the scope of the invention. Accordingly, the scope of protection of the present invention is to be determined by the appended claims.
Claims (9)
1. A prediction system for peritoneal metastasis from gastric cancer based on enhanced CT imaging histology, the prediction system comprising: the device comprises an enhanced CT image input module, an enhanced CT image preprocessing module, a feature extraction module, a data processing module and a prediction and result output module; the enhanced CT image input module is used for inputting enhanced CT images for detecting peritoneal metastasis of gastric cancer; the enhanced CT image preprocessing module is used for carrying out imaging processing on the enhanced CT image and identifying the image characteristics which are manually selected and marked; the feature extraction module is used for extracting three groups of feature data sets from the identified image features, wherein the first group of feature data sets comprise a plurality of intensity feature data, the second group of feature data sets comprise a plurality of morphological feature data, and the third group of feature data sets comprise a plurality of gray texture feature data; the intensity characteristic data are used for reflecting image intensity information of gastric cancer tumor focus on a CT image; the morphological characteristic data are used for reflecting morphological information of gastric cancer tumor focus; the gray texture characteristic data are used for reflecting voxel space distribution intensity level information of gastric cancer focus on a CT image and presenting characteristic information of the surface and the interior of tissue corresponding to the gastric cancer focus; the data processing module is used for carrying out regression analysis and scoring on each characteristic data and transmitting the scoring numerical value to the prediction and result output module; the prediction and result output module performs qualitative analysis and prediction result output on the obtained scoring value;
the data processing module comprises a characteristic variable screening sub-module, a variable coefficient acquisition sub-module and a scoring calculation sub-module; the characteristic variable screening submodule screens out a plurality of characteristic variables related to gastric cancer peritoneal transfer from the characteristic data through a LASSO COX regression calculation model; the variable coefficient acquisition submodule acquires regression coefficients corresponding to the characteristic variables one by one through the LASSO COX regression calculation model; the scoring computation submodule obtains a scoring value through a scoring computation model; the formula of the scoring calculation model is Rad-score=Σ (regression coefficient characteristic variable); the characteristic variables are Eccentricity Eccenttricity, range Extent, gray level co-occurrence matrix correlation information measurement GLCM_IMC and gray level co-occurrence matrix maximum probability GLCM_MaximProbability; the Eccentricity eccenttricity is an image morphological feature variable and is used for reflecting original stomach cancer focus morphological information in a CT image; the range extension is an image morphological characteristic variable and is used for reflecting the size of the area range of the original gastric cancer focus in the CT image; the gray level co-occurrence matrix-correlation information measure GLCM-IMC is an image based on gray level texture characteristic variables and is used for reflecting the correlation of pixels of an area image where a gastric cancer focus is located in a CT image in each vector direction; the gray level co-occurrence matrix-maximum probability GLCM-maximum probability is an image based on gray texture characteristic variables and is used for reflecting the spatial change of the pixel intensity level of an image in the area where the gastric cancer focus is located in the CT image and stably reflecting the image information covered by a specific focus tissue;
the prediction and result output module comprises a patient score matching sub-module, a cut-off value acquisition sub-module, a probability classification sub-module and a result output sub-module; the patient score matching sub-module matches the statistical data of whether the peritoneal transfer occurs to each patient with the corresponding score value, and transmits the matching result to the truncated value acquisition sub-module; the truncated value acquisition submodule analyzes the matching result and generates a truncated value which is used as a judging threshold value of the probability of peritoneal metastasis of the patient; the probability classification submodule is used for comparing the score value of each patient with the cut-off value and taking the comparison result as a judgment result with high or low illness probability.
2. The CT imaging-histology-based prediction system of gastric cancer peritoneal metastasis of claim 1, wherein the LASSO COX regression calculation model is generated by a glrnet function package in computer programming tool R language.
3. The prediction system of gastric cancer peritoneal metastasis based on enhanced CT imaging histology according to claim 1, wherein the patient score matching submodule invokes the MedCalc medical statistics analysis software to match the statistics of whether peritoneal metastasis occurs for each patient with the corresponding score value and transmits the matching result to the cut-off value acquisition submodule; and the cut-off value acquisition submodule calls MedCalc medical statistics analysis software to analyze the matching result and generate a cut-off value which is used as a judging threshold value of the probability of peritoneal metastasis of the patient.
4. The prediction system for gastric cancer peritoneal metastasis based on enhanced CT imaging histology according to claim 3, wherein when the patient's score value < cutoff value, then the patient's probability of developing peritoneal metastasis is low; when the scoring value of the patient is larger than or equal to the cutoff value, the probability of peritoneal metastasis of the patient is high.
5. The prediction method of the prediction system for gastric cancer peritoneal metastasis based on enhanced CT imaging histology according to claim 1, wherein the prediction method comprises the steps of:
s1, inputting an enhanced CT image for detecting peritoneal metastasis of gastric cancer;
s2, performing imaging processing on the input enhanced CT image;
s3, manually selecting the image obtained in the S2, and manually marking the image characteristics of the selected image;
s4, identifying the marked image features;
s5, extracting three groups of characteristic data sets from the identified image characteristics, wherein the first group of characteristic data sets comprise a plurality of intensity characteristic data, the second group of characteristic data sets comprise a plurality of morphological characteristic data, and the third group of characteristic data sets comprise a plurality of gray texture characteristic data; the intensity characteristic data are used for reflecting image intensity information of gastric cancer tumor focus on a CT image; the morphological characteristic data are used for reflecting morphological information of gastric cancer tumor focus; the gray texture characteristic data are used for reflecting voxel space distribution intensity level information of gastric cancer focus on a CT image and presenting characteristic information of the surface and the interior of tissue corresponding to the gastric cancer focus;
s6, carrying out regression analysis and scoring on the characteristic data obtained in the step S5;
and S7, carrying out qualitative analysis and prediction result output on the score value obtained in the step S6.
6. The prediction method according to claim 5, wherein S6 comprises the steps of:
s61, screening a plurality of characteristic variables related to gastric cancer peritoneal transfer from the characteristic data by using a LASSO COX regression calculation model, and obtaining regression coefficients corresponding to the characteristic variables one by one;
s62, obtaining a scoring value by using a scoring calculation model; the formula of the score calculation model is Rad-score=Σ (regression coefficient characteristic variable).
7. The prediction method according to claim 6, wherein S7 comprises the steps of:
s71, matching the statistical data of whether peritoneal metastasis occurs in each patient with corresponding scoring values;
s72, analyzing the matching result and generating a cut-off value which is used as a judging threshold value of the probability of peritoneal metastasis of the patient;
and S73, comparing the score value of each patient with the cut-off value, and taking the comparison result as a judgment result with high or low illness probability.
8. The method of claim 7, wherein the LASSO COX regression calculation model is generated by a glrnet function package in the computer programming tool R language.
9. The method according to claim 8, wherein in S71, a MedCalc medical statistics analysis software is called to match the statistics of whether peritoneal metastasis occurs in each patient with the corresponding scoring values;
in S72, the medical statistical analysis software of the MedCalc is called to analyze the matching result and generate a cut-off value, and the cut-off value is used as a judging threshold value of the probability of peritoneal metastasis of the patient;
in S73, when the patient' S score value < the cutoff value, it is determined that the patient has a low probability of developing peritoneal metastasis; when the scoring value of the patient is more than or equal to the cutoff value, the probability of the patient to generate peritoneal metastasis is judged to be high.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010455571.XA CN112071418B (en) | 2020-05-26 | 2020-05-26 | Gastric cancer peritoneal metastasis prediction system and method based on enhanced CT image histology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010455571.XA CN112071418B (en) | 2020-05-26 | 2020-05-26 | Gastric cancer peritoneal metastasis prediction system and method based on enhanced CT image histology |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112071418A CN112071418A (en) | 2020-12-11 |
| CN112071418B true CN112071418B (en) | 2023-08-01 |
Family
ID=73656157
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010455571.XA Active CN112071418B (en) | 2020-05-26 | 2020-05-26 | Gastric cancer peritoneal metastasis prediction system and method based on enhanced CT image histology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112071418B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115440386B (en) * | 2022-09-30 | 2023-06-20 | 中国医学科学院北京协和医院 | Method and equipment for predicting immune treatment effect of advanced cancer patient based on weighted multi-focus image histology characteristics |
| CN116862858A (en) * | 2023-07-04 | 2023-10-10 | 浙江大学 | Prediction method and system for curative effect of gastric cancer treatment based on image histology |
| CN116687353B (en) * | 2023-08-01 | 2023-12-19 | 宁波杜比医疗科技有限公司 | New adjuvant chemotherapy curative effect evaluation system, equipment and medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107368695A (en) * | 2017-08-24 | 2017-11-21 | 南方医科大学南方医院 | The construction method of GISTs malignant potential disaggregated model based on radiation group |
| CN109259780A (en) * | 2018-07-24 | 2019-01-25 | 南方医科大学南方医院 | The aided assessment system and method that gastric cancer prognosis and chemotherapy based on enhancing CT images group benefit |
-
2020
- 2020-05-26 CN CN202010455571.XA patent/CN112071418B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107368695A (en) * | 2017-08-24 | 2017-11-21 | 南方医科大学南方医院 | The construction method of GISTs malignant potential disaggregated model based on radiation group |
| CN109259780A (en) * | 2018-07-24 | 2019-01-25 | 南方医科大学南方医院 | The aided assessment system and method that gastric cancer prognosis and chemotherapy based on enhancing CT images group benefit |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112071418A (en) | 2020-12-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110060774B (en) | Thyroid nodule identification method based on generative confrontation network | |
| Wang et al. | Using deep convolutional neural networks for multi-classification of thyroid tumor by histopathology: a large-scale pilot study | |
| WO2018120942A1 (en) | System and method for automatically detecting lesions in medical image by means of multi-model fusion | |
| CN112071418B (en) | Gastric cancer peritoneal metastasis prediction system and method based on enhanced CT image histology | |
| US20210219944A1 (en) | Systems and Media for Automatically Diagnosing Thyroid Nodules | |
| CN112086197B (en) | Breast nodule detection method and system based on ultrasonic medicine | |
| CN112088394A (en) | Computerized classification of biological tissue | |
| WO2012154216A1 (en) | Diagnosis support system providing guidance to a user by automated retrieval of similar cancer images with user feedback | |
| CN114782307A (en) | Enhanced CT image colorectal cancer staging auxiliary diagnosis system based on deep learning | |
| Bai et al. | Automatic segmentation of cervical region in colposcopic images using K-means | |
| CN113436150A (en) | Construction method of ultrasound imaging omics model for lymph node metastasis risk prediction | |
| US20180053297A1 (en) | Methods and Apparatuses for Detection of Abnormalities in Low-Contrast Images | |
| US20150065868A1 (en) | System, method, and computer accessible medium for volumetric texture analysis for computer aided detection and diagnosis of polyps | |
| Kaur et al. | Computer-aided diagnosis of renal lesions in CT images: a comprehensive survey and future prospects | |
| Kumar et al. | A Novel Approach for Breast Cancer Detection by Mammograms | |
| CN116630680B (en) | Dual-mode image classification method and system combining X-ray photography and ultrasound | |
| US20160100789A1 (en) | Computer-aided diagnosis system and computer-aided diagnosis method | |
| WO2023198166A1 (en) | Image detection method, system and device, and storage medium | |
| CN114764855A (en) | Intelligent cystoscope tumor segmentation method, device and equipment based on deep learning | |
| CN117275677A (en) | Method for effectively identifying benign and malignant breast ultrasound image tumor | |
| Ahmed et al. | Detection of uterine fibroids in medical images using deep neural networks | |
| CN110738649A (en) | training method of Faster RCNN network for automatic identification of stomach cancer enhanced CT images | |
| CN114926486B (en) | Thyroid ultrasound image intelligent segmentation method based on multi-level improvement | |
| CN113838020B (en) | Lesion area quantification method based on molybdenum target image | |
| Arnold et al. | Indistinct frame detection in colonoscopy videos |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |