Prediction method for treatment effect of gastric superficial early-stage neoplastic lesion

Abstract

The invention provides a prediction method for the effect of treating gastric superficial early-stage neoplastic lesions, and relates to the technical field of treatment effect prediction. Comprising the steps of performing initial image scanning on a to-be-treated patient to obtain a preoperative image set; performing parameter sequence extraction on the preoperative image set to obtain an initial parameter sequence set; obtaining an initial derivative parameter set according to the initial parameter sequence set; performing operative treatment on the to-be-tested patient according to the derivative parameter set to obtain the to-be-tested patient; performing secondary image scanning on the patient to be detected to obtain a postoperative image set; performing parameter sequence extraction on the postoperative image set to obtain a secondary parameter sequence set; obtaining a secondary derivative parameter set according to the secondary parameter sequence set; obtaining a first ROI map and a second ROI map according to the derivative parameter set and the secondary derivative parameter set; and inputting the first ROI map and the second ROI map into a treatment effect prediction model to obtain a prediction result. The problem of low treatment effect prediction precision in the prior art is solved.

Description

Technical Field

[0001] The invention relates to the technical field of therapeutic effect prediction, and in particular to a method for predicting the effect of treating superficial early-stage gastric tumor lesions. Background Art

[0002] Gastric cancer is one of the malignant tumors with high morbidity and mortality worldwide. Early detection and treatment are crucial to improving patient prognosis. Endoscopic therapy (such as endoscopic mucosal resection) and transarterial chemoembolization (TACE) are widely used in the treatment of superficial early gastric tumors. However, the prediction of treatment effect remains a major challenge in clinical practice, especially in terms of the complexity of tumor biological characteristics and treatment response.

[0003] At present, traditional imaging methods (such as CT and MRI) have certain limitations in evaluating the biological characteristics and treatment response of tumors. Especially in the early stages, imaging results often cannot fully reflect the heterogeneity and microstructural changes of tumors. In addition, existing treatment effect prediction methods usually rely on a single imaging parameter and lack a comprehensive evaluation of the overall characteristics of the tumor, which may lead to insufficient accuracy of the prediction results. Summary of the invention

[0004] In order to overcome the deficiencies of the prior art, the purpose of the present invention is to provide a method for predicting the effect of treating superficial early gastric neoplastic lesions. The present invention solves the problem of low accuracy of the effect of treating superficial early gastric neoplastic lesions in the prior art.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A method for predicting the effect of treating superficial early-stage gastric tumor lesions, comprising:

[0007] Perform an initial image scan on the patient to be treated to obtain a preoperative image set;

[0008] Extracting parameter sequences from the preoperative image set to obtain an initial parameter sequence set;

[0009] Obtaining an initial derived parameter set according to the initial parameter sequence set;

[0010] At intervals of a first preset number of days, performing surgical treatment on the patient to be tested according to the derived parameter set to obtain the patient to be tested;

[0011] After a second preset number of days, performing a second image scan on the patient to be tested to obtain a postoperative image set;

[0012] Performing parameter sequence extraction on the postoperative image set to obtain a secondary parameter sequence set;

[0013] Obtaining a secondary derivative parameter set according to the secondary parameter sequence set;

[0014] Obtain a first ROI map and a second ROI map according to the derived parameter set and the secondary derived parameter set;

[0015] Inputting the first ROI map and the second ROI map into a treatment effect prediction model to obtain a prediction result;

[0016] Wherein, the method for constructing the treatment effect prediction model is:

[0017] Obtain clinical information and pathological data of each patient;

[0018] Combining the clinical data and pathological data with the features of the initial parameter sequence set and the secondary parameter sequence set to obtain a first feature combination set and a second feature combination set;

[0019] Determine a feature variation set according to the first feature combination set and the second feature combination set;

[0020] Determine the correlation between the parameter features in the first feature combination set and the second feature combination set and the treatment effect using the Pearson correlation coefficient;

[0021] stratifying the features in the feature change set according to the correlation to obtain stratified features;

[0022] A treatment effect prediction model is constructed based on the stratified features and the random forest model.

[0023] Preferably, an initial image scan is performed on the patient to be treated to obtain a preoperative image set, including:

[0024] Performing an initial image scan on the patient to be treated using IVIM to obtain a first preoperative image subset;

[0025] Performing an initial image scan on the patient to be treated using T1WI to obtain a second preoperative image subset;

[0026] Performing an initial image scan on the patient to be treated using T2WI to obtain a third preoperative image subset;

[0027] Performing an initial image scan on the patient to be treated using LAVA to obtain a fourth preoperative image subset;

[0028] The first preoperative image subset, the second preoperative image subset, the third preoperative image subset and the fourth preoperative image subset are integrated to obtain a preoperative image set.

[0029] Preferably, the initial parameter sequence set includes:

[0030] Number of slices, slice thickness, slice spacing, field of view, matrix, echo time, repetition time, diffusion sensitivity coefficient, fat suppression and respiratory triggering mode.

[0031] Preferably, obtaining an initial derived parameter set according to the initial parameter sequence set includes:

[0032] Determine the quantitative range and size range of diffusion sensitivity coefficients;

[0033] Determine a fast diffusion coefficient, a slow diffusion coefficient, a perfusion fraction and a single exponential parameter according to a quantity threshold of a diffusion sensitivity coefficient and a size threshold of the diffusion sensitivity coefficient;

[0034] An initial set of derived parameters is obtained based on the fast diffusion coefficient, slow diffusion coefficient, perfusion fraction and monoexponential parameter.

[0035] Preferably, the determining of the fast diffusion coefficient, the slow diffusion coefficient, the perfusion fraction and the single exponential parameter according to the quantity threshold of the diffusion sensitivity coefficient and the size threshold of the diffusion sensitivity coefficient comprises:

[0036] Construct signal strength attenuation curve;

[0037] The signal intensity attenuation curve and different diffusion sensitivity coefficients are fitted to obtain the fast diffusion coefficient, slow diffusion coefficient, perfusion fraction and single exponential parameter.

[0038] Preferably, the expression of the signal strength attenuation curve is:

[0039] Sb / S0=(1-f)×exp(-bD)+f×exp(-bD*);

[0040] Among them, Sb represents the signal intensity in the voxel corresponding to the corresponding b value; S0 represents the signal intensity in the voxel when the b value is 0, D is the slow diffusion coefficient, D* is the fast diffusion coefficient, f is the perfusion fraction, ADC is a monoexponential parameter, and b is the diffusion sensitivity coefficient.

[0041] Preferably, the expression of the diffusion sensitivity coefficient is:

[0042] b=γ 2 G 2 δ 2 (Δ-δ / 3);

[0043] Among them, γ is the gyromagnetic ratio constant of the nucleus, G is the intensity of the diffusion gradient, δ is the duration of the diffusion gradient, and δ is the interval time between gradient pulses.

[0044] Preferably, the expression of the treatment effect prediction model is:

[0045]

[0046] Among them, β0 is the model constant term, β1, β2, β3, and β4 are the first, second, third, and fourth regression coefficients of the image-derived parameter changes, respectively, and C i is the ith clinical feature, P j is the jth pathological feature, and ε is the error term.

[0047] The present invention discloses the following technical effects:

[0048] The present invention provides a method for predicting the effect of treating superficial early-stage tumorous lesions of the stomach, comprising: performing an initial image scan on a patient to be treated to obtain a preoperative image set; performing parameter sequence extraction on the preoperative image set to obtain an initial parameter sequence set; obtaining an initial derived parameter set based on the initial parameter sequence set; performing surgical treatment on the patient to be tested based on the derived parameter set at an interval of a first preset number of days to obtain a patient to be tested; performing a secondary image scan on the patient to be tested at an interval of a second preset number of days to obtain a postoperative image set; performing parameter sequence extraction on the postoperative image set to obtain a secondary parameter sequence set; obtaining a secondary derived parameter set based on the secondary parameter sequence set; obtaining a first ROI map and a second ROI map based on the derived parameter set and the secondary derived parameter set; The first ROI map and the second ROI map are input into the treatment effect prediction model to obtain the prediction result; wherein, the construction method of the treatment effect prediction model is: obtaining the clinical information and pathological data of each patient; combining the clinical information and pathological data with the features of the initial parameter sequence set and its secondary parameter sequence set respectively to obtain the first feature combination set and the second feature combination set; determining the feature change set according to the first feature combination set and the second feature combination set; using the Pearson correlation coefficient to determine the correlation between the parameter features in the first feature combination set and the second feature combination set and the treatment effect; stratifying the features in the feature change set according to the correlation to obtain stratified features; constructing the treatment effect prediction model according to the stratified features and the random forest model. The present invention introduces the comparison of preoperative and postoperative parameters, calculates the change in parameters, integrates the clinical information and pathological data of the patient, and constructs a prediction model together with the imaging parameters. The accuracy of the treatment effect prediction is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0050] Figure 1A flow chart of a method for predicting the effect of treating superficial early gastric neoplastic lesions provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0052] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0053] like Figure 1 As shown, the present invention provides a method for predicting the effect of treating superficial early-stage gastric tumor lesions, comprising:

[0054] Step 100: Perform an initial image scan on the patient to be treated to obtain a preoperative image set;

[0055] Step 200: extracting parameter sequences from the preoperative image set to obtain an initial parameter sequence set;

[0056] Step 300: Obtain an initial derived parameter set according to the initial parameter sequence set;

[0057] Step 400: performing surgical treatment on the patient to be tested according to the derived parameter set at intervals of a first preset number of days to obtain the patient to be tested;

[0058] Step 500: performing a second image scan on the patient to be tested at an interval of a second preset number of days to obtain a postoperative image set;

[0059] Step 600: extracting parameter sequences from the postoperative image set to obtain a secondary parameter sequence set;

[0060] Step 700: Obtain a secondary derivative parameter set according to the secondary parameter sequence set;

[0061] Postoperative Parameter Set: The imaging parameters extracted from the postoperative image set, corresponding to the initial parameter set.

[0062] Step 800: Obtain a first ROI map and a second ROI map according to the derived parameter set and the secondary derived parameter set;

[0063] Step 900: inputting the first ROI map and the second ROI map into a treatment effect prediction model to obtain a prediction result;

[0064] Wherein, the method for constructing the treatment effect prediction model is:

[0065] Step 901: Obtain clinical information and pathological data of each patient;

[0066] Step 902: Combining the clinical data and pathological data with the features of the initial parameter sequence set and the secondary parameter sequence set to obtain a first feature combination set and a second feature combination set;

[0067] Specifically, Clinical Data: **Collect clinical information of each patient, including but not limited to: age, gender, symptoms (such as abdominal pain, indigestion, etc.), laboratory test results (such as blood routine, tumor marker levels), family history and lifestyle habits, etc. **Pathological Data: **Collect pathological information of each patient: tumor differentiation degree (high differentiation, moderate differentiation, poor differentiation)

[0068] The size and location of the tumor, the depth of invasion, lymph node metastasis, and other pathological features (such as vascular invasion and neural invasion).

[0069] Step 903: determining a feature variation set according to the first feature combination set and the second feature combination set;

[0070] Step 904: using the Pearson correlation coefficient to determine the correlation between the parameter features in the first feature combination set and the second feature combination set and the treatment effect;

[0071] Step 905: hierarchizing the features in the feature variation set according to the correlation to obtain hierarchical features;

[0072] Step 906: Construct a treatment effect prediction model based on the stratified features and the random forest model.

[0073] More specifically, construct the first feature combination set (Feature Combination Set 1):

[0074] Steps: Merge the clinical data and pathological data of each patient horizontally to generate a data table containing clinical and pathological information of all patients. Merge the imaging features in the initial parameter sequence set with the above data table. When merging, ensure that the patient identity matches correctly (such as by patient ID). Result: The first feature combination set is obtained, which contains the clinical data, pathological data and preoperative imaging features of each patient.

[0075] Construct the second feature combination set (Feature Combination Set 2): Repeat the above process and merge the postoperative imaging features in the secondary parameter sequence set with the patient's clinical data and pathological data. Results: The second feature combination set was obtained, which included the clinical data, pathological data and postoperative imaging features of each patient.

[0076] Furthermore, an initial image scan is performed on the patient to be treated to obtain a preoperative image set, including:

[0077] Performing an initial image scan on the patient to be treated using IVIM to obtain a first preoperative image subset;

[0078] Performing an initial image scan on the patient to be treated using T1WI to obtain a second preoperative image subset;

[0079] Performing an initial image scan on the patient to be treated using T2WI to obtain a third preoperative image subset;

[0080] Performing an initial image scan on the patient to be treated using LAVA to obtain a fourth preoperative image subset;

[0081] The first preoperative image subset, the second preoperative image subset, the third preoperative image subset and the fourth preoperative image subset are integrated to obtain a preoperative image set.

[0082] Specifically, all enrolled cases underwent abdominal MRI dynamic enhancement examination including IVIM-DWI sequence using 3.0T superconducting magnetic resonance imaging (GE Discovery MR750, GE Healthcare, Milwaukee, WI, USA). (1) A 32-channel body coil was used, with the xiphoid process as the center for positioning, and a respiratory gating hose was placed; (2) Since the postprandial and fasting states have a great impact on the blood volume of the liver, this study required fasting for 4 to 6 hours before IVIM-DWI examination; (3) Before scanning, professional magnetic resonance imaging staff conducted breathing training for patients (requiring patients to grasp the rhythm of even breathing and the timing of breath holding at the end of exhalation); (4) The scanning range required the upper edge to cover the top of the diaphragm and the lower edge to cover the lower edge of the liver.

[0083] The specific MRI scanning sequences are as follows: transverse axial fat-suppressed T2-weighted sequence (periodically rotated overlapping parallel lines with enhanced reconstructed fatsaturation T2 weighted imaging, PROPELLER FST2WI), breath-hold transverse axial fast three-dimensional spoiled gradient echo T1-weighted sequence (1ive racecquisition with volume acceleration flex, LAVA-Flex); IVIM-DWI uses a respiratory-triggered spin echo planar imaging sequence. In order to improve the accuracy of the rapid diffusion coefficient (D*), we use an optimized b-value distribution, including 6 low b values ​​(≤200 s / mm2) and 5 high b values ​​(>200 s / mm2), a total of 11 b values, and the b values ​​are (0, 10, 20, 40, 80, 200, 400, 600, 800, 1000, 1200 s / mm2). The total acquisition time of the scanning scheme was about 17 minutes; the IVIM sequence scanning time was 5 minutes and 30 seconds. The scanning parameters remained the same during the study. During the enhanced scanning, paramagnetic contrast agent injection (Magnevist, Bayer, Berlin, Germany) was performed using a high-pressure syringe. Arterial phase images were acquired about 18 seconds after contrast agent injection, portal venous phase images were acquired about 70 seconds after contrast agent injection, and delayed phase images were acquired about 180 seconds after contrast agent injection. The contrast agent dose was 0.1 mmol / Kg, and the contrast agent injection rate was 1.5-2 ml / s. The median time interval between the preoperative IVIM-DWI examination and the TACE treatment was 5.5 days (range: 1-16 days), and the median time interval between the TACE treatment and the postoperative follow-up was 42 days (range: 28-146 days).

[0084] Specifically, the initial parameter sequence set includes:

[0085] Number of slices, slice thickness, slice spacing, field of view, matrix, echo time, repetition time, diffusion sensitivity coefficient, fat suppression and respiratory triggering mode.

[0086] Furthermore, the specific treatment plan of TACE is as follows: First, the groin area is disinfected and draped as usual. After local anesthesia with 1% lidocaine, the right femoral artery is punctured using the modified Seldinger technique and a 5F arterial sheath is introduced. Subsequently, a 4F catheter (RH, Terumo Corporation, Japan) is inserted for celiac trunk-hepatic artery and superior mesenteric artery angiography. After cannulation to the common hepatic artery, a non-ionic contrast agent (Visipaque, 320 mg / mL; GE Healthcare, Princeton, NJ) is injected through a high-pressure syringe. After completing hepatic artery angiography, the tumor vascular staining and blood supply vascular path of primary large liver cancer can be preliminarily determined. Immediately afterwards, under the guidance of the DSA vascular path, a 2.6F Stride microcatheter (Progreat, Terumo Corporation, Tokyo, Japan) is super-selectively cannulated into the tumor blood supply branch artery. If the tumor vessels overlap or are not clearly displayed during the operation, rotational angiography or vertebral beam CT imaging is used to further clarify its blood supply. When the front part of the microcatheter is properly positioned and there is no reflux of contrast agent, embolization treatment is performed. In addition, since 30% to 50% of patients with large liver cancer have collateral blood supply, especially when the tumor is located in the front of the liver, adjacent to the diaphragm or invades the liver capsule, the incidence of collateral circulation blood supply is very high. Therefore, after the initial completion of embolization, it is necessary to observe whether the embolization effect is thorough; if the iodized oil embolization is not thorough, actively look for the possibility of ectopic hepatic artery or other collateral tumor blood supply arteries, such as the inferior phrenic artery, intercostal artery, gastroduodenal artery, left gastric artery, internal thoracic artery, adrenal artery, etc.; if there is collateral blood supply, the corresponding collateral blood supply arteries are embolized respectively. Precautions for the preparation of embolic drugs and materials before embolization: (1) Before surgery, the medication regimen is often determined based on the size of the tumor diameter, the patient's weight and the patient's general condition, including 2 to 4 chemotherapeutic drugs such as epirubicin (30-50 mg), fluorouracil (500-750 mg), pirarubicin hydrochloride (40-50 mg), platinum (30-60 mg), 5-FU (500-1000 mg), and calcium folinate (75-100 mg); (2) Before surgery, powdered chemotherapeutic drugs such as epirubicin and platinum are mixed with iodized oil (lipiodol; Laboratoire Guerbet, Roissy, France) to form an emulsion; the total amount of iodized oil in one treatment depends on the size of the tumor, but generally does not exceed 20 ml; (3) For patients with tumors with abundant blood supply, gelatin sponge particles (diameter 1 mm) or polyvinyl alcohol particles (PVA, 300-700 umol / L, Cook Medical, Bloomington, IN, USA) must be given for embolization.

[0087] Further, obtaining an initial derived parameter set according to the initial parameter sequence set includes:

[0088] Determine the quantitative range and size range of diffusion sensitivity coefficients;

[0089] Determine a fast diffusion coefficient, a slow diffusion coefficient, a perfusion fraction and a single exponential parameter according to a quantity threshold of a diffusion sensitivity coefficient and a size threshold of the diffusion sensitivity coefficient;

[0090] An initial set of derived parameters is obtained based on the fast diffusion coefficient, slow diffusion coefficient, perfusion fraction and monoexponential parameter.

[0091] Further, the determining of the fast diffusion coefficient, the slow diffusion coefficient, the perfusion fraction and the single exponential parameter according to the quantity threshold of the diffusion sensitivity coefficient and the size threshold of the diffusion sensitivity coefficient includes:

[0092] Construct signal strength attenuation curve;

[0093] The signal intensity attenuation curve and different diffusion sensitivity coefficients are fitted to obtain the fast diffusion coefficient, slow diffusion coefficient, perfusion fraction and single exponential parameter.

[0094] Furthermore, the expression of the signal strength attenuation curve is:

[0095] Sb / S0=(1-f)×exp(-bD)+f×exp(-bD*);

[0096] Among them, Sb represents the signal intensity in the voxel corresponding to the corresponding b value; S0 represents the signal intensity in the voxel when the b value is 0, D is the slow diffusion coefficient, D* is the fast diffusion coefficient, f is the perfusion fraction, ADC is a monoexponential parameter, and b is the diffusion sensitivity coefficient.

[0097] Furthermore, the expression of the diffusion sensitivity coefficient is:

[0098] b=γ 2 G 2 δ 2 (Δ-δ / 3);

[0099] Among them, γ is the gyromagnetic ratio constant of the nucleus, G is the intensity of the diffusion gradient, δ is the duration of the diffusion gradient, and δ is the interval time between gradient pulses.

[0100] Specifically, IVIM-DWI data of all b values ​​were input into the GEAW4.6 post-processing workstation, and the data were analyzed by MADC software (GE Medical System, Milwaukee, WI, USA), and IVIM derived parameters: ADC, D, D*, f were extracted synchronously to generate pseudo-color images of IVIM numbers. The IVIM biexponential fitting formula is: Sb / S0 = (1-f) × exp(-bD) + f × exp(-bD*), where Sb represents the signal intensity in the voxel corresponding to the corresponding b value; S0 represents the signal intensity in the voxel when the b value is 0. D is the slow diffusion coefficient, and the unit is ×10-3mm2 / s; D* is the fast diffusion coefficient, and the unit is ×10-3mm2 / s; f is the perfusion fraction, and the unit is %. The single exponential parameter ADC is analyzed and measured using a single exponential model, and its unit is ×10-3mm2 / s. In terms of image measurement, two senior attending physicians used ITK-SNAP software to select the b=1000 (s / mm2) image in IVIM-DWI to delineate the ROI of the tumor. In order to more accurately delineate the tumor boundary, the T2WI image and dynamic enhancement image in the patient's concurrent scanning sequence should be referred to during measurement. At the same time, in order to avoid measurement bias, the two observers used a double-blind and random method to manually delineate the tumor layer layer by layer on their respective workstations. When delineating, it should be noted that the delineated ROI should include all tumor components such as cystic changes, hemorrhage, necrosis, etc. in the tumor to obtain the ROI map of the entire tumor; if the two observers have questions about the tumor boundary, they need to consult the third chief physician.

[0101] After completing the above steps, the outlined ROI needs to be copied synchronously to other parameter images of IVIM, such as D, D*, f and ADC images. The whole tumor volume voxel of each IVIM parameter is automatically analyzed using Matlab software to generate the mean, minimum value (min), 5th digit, 10th digit, 25th digit, 50th digit, 75th digit, 90th digit and maximum value (max), as well as skewness and kurtosis, which are recorded in turn; the skewness and kurtosis values ​​can reflect the shape of the histogram, and can also analyze the distribution status of ADC value, D*, D value and f value.

[0102] Furthermore, the expression of the treatment effect prediction model is:

[0103]

[0104] Among them, β0 is the model constant term, β1, β2, β3, and β4 are the first, second, third, and fourth regression coefficients of the image-derived parameter changes, respectively, and C i is the ith clinical feature, Pj is the jth pathological feature, and ε is the error term.

[0105] Specifically, clinical characteristics: age (years) = 60, sex (male = 1, female = 0) = 1; pathological characteristics: tumor size (cm) = 2.5, degree of differentiation (high = 1, medium = 2, low = 3) = 2; known regression coefficient β values ​​(hypothesis): β0 = -2.5, β1 = 10 × 103, β2 = 8 × 103, β3 = 0.5 × 103, β4 = 20, β5 = 0.05 corresponding to age, β6 = -1.0 corresponding to sex, β7 = -0.8 (corresponding to tumor size), β8 = 1.5 (corresponding to degree of differentiation);

[0106] TE=-2.5+10×103×(0.25×10-3)+8×103×(0.15×10-3)+0.5×103×(5×10-3)+20 ×0.05+0.05×60-1.0×1-0.8×2.5+1.5×2=-2.5+2.5+1.2+2.5+1+3-1-2+3=4.7;

[0107] If TE is greater than a certain threshold (e.g., 0), then the treatment is predicted to be effective. In this example, TE=4.7>0,

[0108] Therefore, the treatment effect is predicted to be good.

[0109] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0110] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for predicting the effect of treating superficial early-stage gastric neoplastic lesions, characterized in that: include: Perform an initial image scan on the patient to be treated to obtain a preoperative image set; Extracting parameter sequences from the preoperative image set to obtain an initial parameter sequence set; Obtaining an initial derived parameter set according to the initial parameter sequence set; At intervals of a first preset number of days, performing surgical treatment on the patient to be tested according to the derived parameter set to obtain the patient to be tested; After a second preset number of days, performing a second image scan on the patient to be tested to obtain a postoperative image set; Performing parameter sequence extraction on the postoperative image set to obtain a secondary parameter sequence set; Obtaining a secondary derivative parameter set according to the secondary parameter sequence set; Obtain a first ROI map and a second ROI map according to the derived parameter set and the secondary derived parameter set; Inputting the first ROI map and the second ROI map into a treatment effect prediction model to obtain a prediction result; Wherein, the method for constructing the treatment effect prediction model is: Obtain clinical information and pathological data of each patient; Combining the clinical data and pathological data with the features of the initial parameter sequence set and the secondary parameter sequence set to obtain a first feature combination set and a second feature combination set; Determine a feature variation set according to the first feature combination set and the second feature combination set; Determine the correlation between the parameter features in the first feature combination set and the second feature combination set and the treatment effect using the Pearson correlation coefficient; stratifying the features in the feature change set according to the correlation to obtain stratified features; A treatment effect prediction model is constructed based on the stratified features and the random forest model.

2. A method for predicting the effect of treating superficial early gastric neoplastic lesions according to claim 1, characterized in that: Perform an initial image scan on the patient to be treated to obtain a preoperative image set, including: Performing an initial image scan on the patient to be treated using IVIM to obtain a first preoperative image subset; Performing an initial image scan on the patient to be treated using T1WI to obtain a second preoperative image subset; Performing an initial image scan on the patient to be treated using T2WI to obtain a third preoperative image subset; Performing an initial image scan on the patient to be treated using LAVA to obtain a fourth preoperative image subset; The first preoperative image subset, the second preoperative image subset, the third preoperative image subset and the fourth preoperative image subset are integrated to obtain a preoperative image set.

3. A method for predicting the effect of treating superficial early gastric neoplastic lesions according to claim 2, characterized in that: The initial parameter sequence set includes: Number of slices, slice thickness, slice spacing, field of view, matrix, echo time, repetition time, diffusion sensitivity coefficient, fat suppression and respiratory triggering mode.

4. A method for predicting the effect of treating superficial early gastric neoplastic lesions according to claim 3, characterized in that: Obtaining an initial derived parameter set according to the initial parameter sequence set includes: Determine the quantitative range and size range of diffusion sensitivity coefficients; Determine a fast diffusion coefficient, a slow diffusion coefficient, a perfusion fraction and a single exponential parameter according to a quantity threshold of a diffusion sensitivity coefficient and a size threshold of the diffusion sensitivity coefficient; An initial set of derived parameters is obtained based on the fast diffusion coefficient, slow diffusion coefficient, perfusion fraction and monoexponential parameter.

5. A method for predicting the effect of treating superficial early-stage gastric neoplastic lesions according to claim 4, characterized in that: The method of determining the fast diffusion coefficient, the slow diffusion coefficient, the perfusion fraction and the single exponential parameter according to the quantity threshold of the diffusion sensitivity coefficient and the size threshold of the diffusion sensitivity coefficient comprises: Construct signal strength attenuation curve; The signal intensity attenuation curve and different diffusion sensitivity coefficients are fitted to obtain the fast diffusion coefficient, slow diffusion coefficient, perfusion fraction and single exponential parameter.

6. A method for predicting the effect of treating superficial early-stage gastric tumor lesions according to claim 5, characterized in that: The expression of the signal strength attenuation curve is: Sb / S0=(1-f)×exp(-bD)+f×exp(-bD*); Among them, Sb represents the signal intensity in the voxel corresponding to the corresponding b value; S0 represents the signal intensity in the voxel when the b value is 0, D is the slow diffusion coefficient, D* is the fast diffusion coefficient, f is the perfusion fraction, ADC is a monoexponential parameter, and b is the diffusion sensitivity coefficient.

7. A method for predicting the effect of treating superficial early-stage gastric neoplastic lesions according to claim 6, characterized in that: The expression of the diffusion sensitivity coefficient is: b=c 2 G 2 d 2 (D-d / 3); Among them, γ is the gyromagnetic ratio constant of the nucleus, G is the intensity of the diffusion gradient, δ is the duration of the diffusion gradient, and δ is the interval time between gradient pulses.

8. A method for predicting the effect of treating superficial early-stage gastric neoplastic lesions according to claim 7, characterized in that: The expression of the treatment effect prediction model is: Among them, β0 is the model constant term, β1, β2, β3, and β4 are the first, second, third, and fourth regression coefficients of the image-derived parameter changes, respectively, and C i is the ith clinical feature, P j is the jth pathological feature, and ε is the error term.

Images