Efficient tumor interventional ablation accurate positioning system
Through the combination of multimodal image fusion and deep learning models, combined with multi-probe strategy and real-time feedback control system, the problem of insufficient tumor localization and ablation accuracy in the existing technology is solved, and more efficient and accurate tumor ablation effect is achieved.
Patent Information
- Application Number
- CN202510061915.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-15
- Publication Date
- 2025-05-13
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing tumor interventional ablation technology has problems such as insufficient accuracy, insufficient single modal image information, and limited ablation range of ablation of a single probe in the tumor location and ablation process, resulting in unclear identification of tumor boundaries and incomplete ablation, which increases the risk of postoperative recurrence.
Multimodal image fusion technology is adopted, combined with CT, MRI and PET imaging, and the tumor boundary segmentation is automatically segmented using deep learning models, and the comprehensive ablation of the tumor area is achieved through multi-probe strategy and real-time feedback control system.
It improves the accuracy and comprehensiveness of the identification of tumor areas, ensures the coverage and effectiveness of tumor ablation, reduces the risk of postoperative recurrence, and optimizes the accuracy and safety of the ablation process.
Smart Images
Figure CN119970229A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of tumor ablation, and in particular to a high-efficiency tumor interventional ablation precise positioning system. Background Art
[0002] With the continuous advancement of medical technology, tumor interventional ablation technology has become an important means of treating patients with small tumors and early tumors. Interventional ablation technology includes radiofrequency ablation (RFA), microwave ablation (MWA), laser ablation, cryoablation, etc., which are increasingly favored by clinicians and patients because of their small trauma, fast recovery and significant treatment effect. However, the effectiveness of these technologies depends to a large extent on the precise positioning of the tumor. At present, tumor positioning mainly relies on imaging technologies such as ultrasound, CT and MRI. Ultrasound has the advantage of real-time imaging, but its image quality and depth penetration ability are limited, making it difficult to accurately identify deep tumors. Although CT imaging can provide high-resolution images, it is intermittent and the patient may be displaced during the examination, resulting in positioning errors. MRI is superior in soft tissue imaging, but its operation is complicated and requires high cooperation from patients. In addition, existing technologies usually fail to effectively integrate information from different imaging modes during tumor ablation, making it difficult for doctors to obtain a global view in actual operations, increasing the complexity and risk of the operation. The success of tumor ablation depends not only on the accurate positioning of the tumor itself, but also on the protection of surrounding important blood vessels, nerves and organs. Traditional positioning methods often cannot provide real-time feedback on changes in the tumor and its surrounding structures in a dynamically changing in vivo environment. The tumor's morphology, location, and volume often change during treatment, further increasing the difficulty of ablation.
[0003] However, the existing technology still has major deficiencies, such as:
[0004] In the existing tumor interventional ablation navigation system, usually only a single modality image (such as CT image) is relied upon to identify the tumor area. Although CT images can provide high-resolution anatomical structure information, they are insufficient in soft tissue contrast and functional information. Since a single modality image cannot fully reflect the various characteristics of the tumor, in actual operation, it is easy to have unclear tumor boundary identification and inaccurate segmentation results, which affects the accuracy of subsequent ablation operations. At the same time, in the existing tumor ablation process, ablation mainly relies on a single probe, and the ablation range is determined by preoperative planning. The ablation range of a single probe is limited. For larger or complex tumors, a single probe is difficult to completely cover the entire tumor area, and incomplete ablation is prone to occur, resulting in an increased risk of postoperative recurrence. Summary of the invention
[0005] The purpose of the present invention is to provide an efficient tumor interventional ablation precise positioning system to solve the problems raised in the above background technology.
[0006] To achieve the above object, the present invention provides the following technical solutions:
[0007] An efficient tumor interventional ablation precise positioning system comprises the following steps:
[0008] Step 1: CT image processing module;
[0009] Step 2: Processing of pixels with weak attributes;
[0010] Step 3: Machine Learning Module;
[0011] Step 4: Tumor ablation navigation module;
[0012] Step 5: Visualization module.
[0013] Preferably, the step 1: CT image processing module specifically includes:
[0014] Multimodal image fusion:
[0015] Obtain multimodal imaging data from patients, including CT, MRI, and PET images, and use multimodal image fusion algorithms to register and fuse different image data to ensure spatial consistency of different images;
[0016] Deep learning improves segmentation accuracy:
[0017] Use the trained deep learning model to automatically segment the fused image data, identify the tumor area, and post-process the segmentation results to obtain more accurate tumor boundaries.
[0018] Preferably, the step 2: processing pixels with weak attributes specifically includes:
[0019] Increase the weight factor:
[0020] In the segmentation similarity acquisition module of weak attribute pixels, a weight coefficient is introduced to perform weighted averaging according to the importance of different segmentation results.
[0021] Preferably, the step three: machine learning module specifically includes:
[0022] Comprehensive multi-dimensional features:
[0023] Multi-dimensional features of the tumor, including morphological features, texture features, and metabolic features, are extracted and input into an optimized machine learning model to classify the tumor malignancy grade.
[0024] Preferably, the step 4: tumor ablation navigation module specifically includes:
[0025] Multi-probe strategy:
[0026] According to the final tumor area location detected by the CT image processing module, the insertion path and location of the multiple probes are planned, and the multi-probe synchronous ablation technology is used to achieve comprehensive ablation of the tumor area through the synergistic effect of multiple thermal ablation needles;
[0027] Real-time feedback control:
[0028] A real-time feedback control system is configured to monitor the temperature changes around the ablation needle. Based on the temperature feedback data, the power and position of the ablation needle are dynamically adjusted to ensure that the ablation area reaches the expected temperature.
[0029] Preferably, the step 5: visualization module specifically includes:
[0030] 3D Visualization:
[0031] The final tumor area in the CT image processing module is 3D modeled using 3D modeling software. The malignant tumor grade obtained by the machine learning module and the ablation needle position and ablation area obtained by the tumor ablation navigation module are combined in the 3D model for real-time display.
[0032] Compared with the prior art, the present invention has the following beneficial effects:
[0033] 1. By fusing CT, MRI and PET images, more comprehensive tumor information is provided. Combining the advantages of each modality, such as CT anatomical structure information, MRI soft tissue contrast and PET metabolic activity data, the identification of tumor areas is more accurate and comprehensive. The U-Net deep learning model is used to automatically segment the fused images, which can effectively identify the precise boundaries of the tumor. Since the deep learning model has powerful image recognition capabilities, it can greatly improve the accuracy of tumor segmentation and reduce errors caused by human operations;
[0034] 2. Introduce a multi-probe strategy to plan the positions and insertion paths of multiple thermal ablation needles, and ensure that the ablation range can cover the entire tumor area through the synergistic effect of multiple probes. This not only increases the coverage of ablation, but also effectively avoids the problem of incomplete ablation and reduces the risk of postoperative recurrence;
[0035] 3. Configure a real-time feedback control system, use a temperature sensor to monitor the temperature changes around the ablation needle, and dynamically adjust the ablation parameters based on real-time data. This real-time feedback control system can ensure that the temperature in the ablation area reaches the expected level, avoiding excessive ablation that damages surrounding normal tissues or insufficient ablation that fails to completely eliminate the tumor. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION
[0037] 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.
[0038] See also Figure 1 , the present invention provides a technical solution:
[0039] An efficient tumor interventional ablation precise positioning system comprises the following steps:
[0040] Step 1: CT image processing module;
[0041] Step 2: Processing of pixels with weak attributes;
[0042] Step 3: Machine Learning Module:
[0043] Step 4: Tumor ablation navigation module;
[0044] Step 5: Visualization module.
[0045] Step 1: CT image processing module, including:
[0046] Multimodal image fusion:
[0047] Obtain multimodal imaging data from patients, including CT, MRI, and PET images, and use multimodal image fusion algorithms to register and fuse different image data to ensure spatial consistency of different images;
[0048] Deep learning improves segmentation accuracy:
[0049] Use the trained deep learning model to automatically segment the fused image data, identify the tumor area, and post-process the segmentation results to obtain more accurate tumor boundaries.
[0050] Collect CT, MRI and PET image data from patients, use the Mutual Information-based Registration algorithm for image registration, load CT images as reference images, load MRI and PET images, calculate the mutual information value of each pixel, adjust the position and rotation parameters of MRI and PET images to maximize the mutual information value, export the registered MRI and PET images, fuse the registered MRI and PET images with the CT images, and generate multimodal image data. Multimodal image fusion can provide more comprehensive tumor information, combining the advantages of CT, MRI and PET images to improve the accuracy and comprehensiveness of tumor identification;
[0051] The Mutual Information-based Registration algorithm is used to calculate the mutual information values between different images for registration. It can align image data of different modalities with higher precision, reduce registration errors, and improve the quality of subsequent image fusion. The Mutual Information-based Registration algorithm is used to calculate the mutual information values between different images for registration. It can align image data of different modalities with higher precision, reduce registration errors, and improve the quality of subsequent image fusion. The registered and fused multi-modal images can more clearly show the boundaries and surrounding tissues of the tumor, reduce misdiagnosis and missed diagnosis due to insufficient information from a single modality image, and improve the reliability of clinical diagnosis.
[0052] The U-Net model is used to automatically segment the tumor area. The pre-processed fused multimodal images are normalized and cut into pieces. The pre-processed images are input into the U-Net model to obtain the segmentation results output by the U-Net model. The segmentation results are post-processed and the morphological operations of dilation and corrosion are used to remove noise to obtain more accurate tumor boundaries. The deep learning model performs well in image segmentation tasks and can greatly improve the recognition accuracy of tumor areas.
[0053] Step 2: Processing of pixels with weak attributes, including:
[0054] Increase the weight factor:
[0055] In the segmentation similarity acquisition module of weak attribute pixels, a weight coefficient is introduced to perform weighted averaging according to the importance of different segmentation results.
[0056] Introduce a weight coefficient in the segmentation similarity acquisition module and define a weight coefficient α, which is used for weighted average of different segmentation results to calculate segmentation similarity. Use the formula segmentation similarity = α * Segmentation result 1+(1-α) * Segmentation result 2, calculate the segmentation similarity of each weak attribute pixel, adjust the value of α, optimize the weight coefficient according to the experimental results, and improve the recognition accuracy. By introducing the weight coefficient, the influence of different segmentation results on the final segmentation similarity can be more flexibly adjusted to improve the accuracy of weak attribute pixel recognition.
[0057] Define the weight coefficient α, which is used for the weighted average of different segmentation results: Introduce an adjustable weight coefficient α in the algorithm design. The coefficient ranges from 0 to 1. Set the initial value α. The initial value can be determined based on experience or small-scale experiments conducted in advance. Use the formula segmentation similarity = α * Segmentation result 1+(1-α) *Segmentation result 2, calculate the segmentation similarity of each weak attribute pixel, obtain multiple segmentation results generated by the deep learning model or other segmentation algorithms, such as segmentation result 1 and segmentation result 2, and for each pixel i, calculate its segmentation value in different segmentation results, using the formula: Segmentation similarity (i) = α * Segmentation result 1(i)+(1-α) * Segmentation result 2(i), calculate the segmentation similarity of each pixel, adjust the value of α, optimize the weight coefficient according to the experimental results, improve the recognition accuracy, conduct a series of experiments, adjust the value of α, observe the changes in the segmentation results, evaluate the segmentation effect under different α values through cross-validation or other evaluation indicators such as Dice coefficient and Jaccard coefficient, and select the α value that makes the best segmentation effect as the final weight coefficient.
[0058] By introducing the weight coefficient α, the contribution of different segmentation results to the final segmentation similarity can be flexibly adjusted, avoiding the impact of the inaccuracy of a single segmentation result on the overall result and improving the recognition accuracy of pixels with weak attributes. The introduction of the weight coefficient α makes the system more flexible and the α value can be adjusted according to different cases and imaging data to obtain the optimal segmentation effect with greater adaptability. By weighted averaging multiple segmentation results, the recognition accuracy of the tumor area can be improved, which is crucial for subsequent tumor ablation positioning and treatment plan design.
[0059] Step 3: Machine learning module, including:
[0060] Comprehensive multi-dimensional features:
[0061] Multi-dimensional features of the tumor, including morphological features, texture features, and metabolic features, are extracted and input into an optimized machine learning model to classify the tumor malignancy grade.
[0062] Extract multi-dimensional features of tumors: extract morphological features such as area and perimeter, use image processing algorithms to calculate the area and perimeter of the tumor area, extract texture features (such as grayscale co-occurrence matrix), calculate the grayscale co-occurrence matrix, extract contrast, uniformity and other features from it, extract metabolic features such as the SUV value of PET images, extract the SUV value of each tumor pixel from the PET image, calculate the average SUV, build a machine learning model and train it, use the extracted multi-dimensional features as input, use the random forest model for training, and after training, use the model to classify the tumor malignancy level of new data. The comprehensive multi-dimensional features can more comprehensively reflect the characteristics of the tumor and improve the accuracy of the machine learning model in judging the tumor malignancy level.
[0063] Use image processing algorithms to calculate the area and perimeter of the tumor region: Use image segmentation algorithms such as U-Net or other deep learning models to segment the tumor region, and use image processing libraries such as OpenCV or Scikit-image to calculate the area and perimeter of the segmented tumor region. Area: count the number of pixels in the tumor region, perimeter: calculate the length of the tumor region boundary;
[0064] Extract texture features such as gray-level co-occurrence matrix: Calculate the gray-level co-occurrence matrix and extract features such as contrast and uniformity from it. Convert the CT image data in the tumor area into a grayscale image. Use the gray-level co-occurrence matrix (GLCM) algorithm to calculate the texture features in the tumor area. Extract multiple texture features from the GLCM, such as contrast, uniformity, energy, and entropy.
[0065] Extract metabolic features (such as SUV values from PET images):
[0066] Extract the SUV value of each tumor pixel from the PET image and calculate the average SUV: Obtain PET image data from the fused multimodal image, extract the SUV value (Standardized Uptake Value) of each pixel in the tumor area, and calculate the average of all SUV values in the tumor area as the metabolic feature.
[0067] Build and train a machine learning model:
[0068] The extracted multi-dimensional features are used as input and trained using the random forest model:
[0069] Multi-dimensional features such as morphological features such as area, perimeter, texture features such as contrast, uniformity, metabolic features (such as average SUV) are integrated into feature vectors. Use Python's Scikit-learn library to build a random forest model, use feature vectors and corresponding tumor malignancy grade labels as training data, and train the model. After training, use the model to classify the tumor malignancy grade of new data: perform the same feature extraction process on the imaging data of new patients to generate feature vectors, input the feature vectors into the trained random forest model, perform prediction and classification of tumor malignancy grade, and output the classification results for reference by clinicians.
[0070] Step 4: Tumor ablation navigation module, including:
[0071] Multi-probe strategy:
[0072] According to the final tumor area location detected by the CT image processing module, the insertion path and location of the multiple probes are planned, and the multi-probe synchronous ablation technology is used to achieve comprehensive ablation of the tumor area through the synergistic effect of multiple thermal ablation needles;
[0073] Real-time feedback control:
[0074] A real-time feedback control system is configured to monitor the temperature changes around the ablation needle. Based on the temperature feedback data, the power and position of the ablation needle are dynamically adjusted to ensure that the ablation area reaches the expected temperature.
[0075] According to the final tumor area position detected by the CT image processing module, the insertion path and position of multiple probes are planned: the tumor area is 3D modeled using 3D modeling software, and the positions and insertion paths of multiple thermal ablation needles are planned in the 3D model to ensure that the entire tumor area is covered. Multiple thermal ablation needles are inserted at the same time, and the ablation equipment is started for synchronous ablation. The multi-probe strategy can cover a larger tumor area, avoid the problem that a single probe cannot completely ablate the tumor, and improve the ablation effect.
[0076] Configure a real-time feedback control system to monitor the temperature changes around the ablation needle: install temperature sensors around the ablation needle to collect temperature data in real time and transmit it to the control system. According to the temperature feedback data, dynamically adjust the power and position of the ablation needle, set the temperature threshold, and ensure that the ablation area reaches the expected temperature. When the temperature is lower than the threshold, increase the power of the ablation needle; when the temperature is higher than the threshold, reduce the power or adjust the position of the ablation needle. The real-time feedback control system can dynamically adjust the ablation parameters to ensure that the tumor tissue is completely ablated and avoid excessive or insufficient ablation.
[0077] According to the final tumor area position detected by the CT image processing module, the insertion path and position of the multi-probe are planned:
[0078] Use 3D modeling software to perform 3D modeling of the tumor area:
[0079] Import CT image data into 3D modeling software (such as 3D Slicer or Blender).
[0080] The image segmentation results are used to generate a three-dimensional model of the tumor area, which accurately reflects the shape, size and location of the tumor.
[0081] Plan the locations and insertion paths of multiple thermal ablation needles in the 3D model to ensure coverage of the entire tumor area:
[0082] Mark the center and boundary of the tumor in the three-dimensional model, design the positions of multiple thermal ablation needles so that they are evenly distributed in the tumor area, ensure that each needle can cover a certain range of tumor tissue, plan the insertion path of each thermal ablation needle to avoid puncturing important blood vessels and organs, ensure safety, verify the feasibility of the insertion path and position, and dynamically adjust the model to ensure the best coverage effect.
[0083] Perform multi-probe ablation:
[0084] Inserting multiple thermal ablation needles simultaneously:
[0085] According to the predetermined insertion path and position, use image guidance equipment (such as CT or ultrasound) to guide the insertion of the thermal ablation needle to ensure that each thermal ablation needle is accurately inserted into the predetermined position for position confirmation and fine-tuning.
[0086] Start the ablation device and perform synchronous ablation:
[0087] After confirming that all thermal ablation needles have been correctly inserted into place, start the thermal ablation device.
[0088] Multiple thermal ablation needles are activated simultaneously to perform synchronous ablation on the tumor area to ensure ablation efficiency and effect.
[0089] Configure a real-time feedback control system to monitor temperature changes around the ablation needle:
[0090] Installing a temperature sensor around the ablation needle:
[0091] A high-precision temperature sensor is installed around each thermal ablation needle to ensure real-time monitoring of temperature changes in the ablation area, real-time collection of temperature data, and transmission to the control system. The temperature sensor collects temperature data in real time and transmits it to the central control system via wired or wireless transmission. The control system displays and records temperature data in real time for dynamic monitoring.
[0092] Dynamically adjust the power and position of the ablation needle based on temperature feedback data:
[0093] Set the temperature threshold to ensure the ablation area reaches the desired temperature:
[0094] According to the clinical requirements of tumor ablation, a temperature threshold (such as 60℃-100℃) is set to ensure that the tumor tissue is effectively ablated. When the temperature is lower than the threshold, the power of the ablation needle is increased; when the temperature is higher than the threshold, the power is reduced or the position of the ablation needle is adjusted: the control system analyzes the temperature data in real time. When the temperature is lower than the set threshold, the power of the thermal ablation needle is automatically increased to ensure that the ablation temperature is reached. When the temperature exceeds the set threshold, the power of the thermal ablation needle is automatically reduced or the position of the ablation needle is fine-tuned to prevent excessive ablation from damaging normal tissue. Real-time monitoring and dynamic adjustment ensure the accuracy and safety of the ablation process.
[0095] The multi-probe strategy ensures that the tumor area is fully covered, improves ablation efficiency and effectiveness, and reduces the risk of residual tumor tissue. The real-time feedback control system improves the accuracy of the ablation process through temperature monitoring and dynamic adjustment, avoids over-ablation and under-ablation, and optimizes the ablation strategy. Accurate three-dimensional modeling and insertion path planning ensure the safety of the multi-probe ablation process and reduce damage to normal tissues and important structures. Real-time temperature feedback and power adjustment functions enable the ablation process to have dynamic adjustment capabilities, which can cope with temperature changes in the tumor area and ensure consistent ablation effects.
[0096] Step 5: Visualization module, including:
[0097] 3D Visualization:
[0098] The final tumor area in the CT image processing module is 3D modeled using 3D modeling software. The malignant tumor grade obtained by the machine learning module and the ablation needle position and ablation area obtained by the tumor ablation navigation module are combined in the 3D model for real-time display.
[0099] Use 3D modeling software to perform 3D modeling of the final tumor area in the image processing module: import CT image data, use image segmentation results to generate a 3D model of the tumor area, combine the malignant tumor grade obtained by the machine learning module, and the ablation needle position and ablation area obtained by the tumor ablation navigation module in the 3D model for real-time display: map the malignancy grade information to the 3D model (such as using different colors to represent different grades), superimpose the ablation needle position and ablation area in the 3D model to form a complete surgical navigation map. 3D visualization technology can provide more intuitive and comprehensive tumor ablation navigation information, helping doctors perform surgical operations more accurately.
[0100] Use 3D modeling software to perform 3D modeling of the final tumor area in the image processing module:
[0101] Import CT image data:
[0102] Use medical image processing software (such as 3D Slicer, Amira or Blender) to import CT image data and ensure that the CT image data has been stored and read in accordance with the DICOM standard to ensure the accuracy and integrity of the data.
[0103] Using the image segmentation results, generate a 3D model of the tumor area:
[0104] Call image segmentation algorithms (such as U-Net, etc.) to process CT image data, extract the tumor area, and use the segmentation results to generate a three-dimensional model of the tumor to ensure that the model accurately reflects the shape, size and position of the tumor. Surface reconstruction algorithms (such as Marching Cubes) can be used to construct a three-dimensional surface model.
[0105] The malignant tumor grade obtained by the machine learning module and the ablation needle position and ablation area obtained by the tumor ablation navigation module are combined in a three-dimensional model for real-time display:
[0106] Map the malignancy level information to the 3D model (e.g., use different colors to represent different levels):
[0107] The malignancy level information of each tumor area is obtained from the machine learning module, and the color mapping technology is used to display the information of different malignancy levels in different colors on the tumor 3D model. For example, low malignancy level is represented by green, medium malignancy level is represented by yellow, and high malignancy level is represented by red. This ensures that the color mapping has good differentiation, which facilitates doctors to quickly identify the malignancy of the tumor.
[0108] The ablation needle position and ablation area are superimposed on the 3D model to form a complete surgical navigation map:
[0109] The predetermined ablation needle position and ablation area information are obtained from the tumor ablation navigation module, the three-dimensional position and trajectory of the ablation needle are superimposed on the three-dimensional model, and a virtual ablation needle model is used for visualization to display the range of the ablation area. Different transparencies or color gradients can be used to represent the temperature distribution and ablation effect of the ablation area, achieving real-time interactive display and allowing doctors to rotate, scale and cut the three-dimensional model to view surgical navigation maps at different angles and sections.
[0110] Through 3D visualization technology, the tumor area, malignancy grade, ablation needle position and ablation area are integrated into a 3D model, providing an intuitive and accurate surgical navigation map to help doctors better plan and perform operations. The ablation needle position and ablation area are displayed in real time to ensure that the thermal ablation needle is accurately inserted into the predetermined position, reduce errors, and improve the accuracy and safety of the operation. The malignancy grade information is intuitively displayed on the 3D model to provide doctors with comprehensive tumor information, which is helpful for formulating personalized treatment plans and realizing real-time interactive 3D display. Doctors can view the surgical navigation map from different angles and sections, enhancing the flexibility and convenience of preoperative planning and intraoperative navigation.
[0111] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.
Claims
1. An efficient tumor interventional ablation precise positioning system, characterized by: The steps include: Step 1: CT image processing module; Step 2: Processing of pixels with weak attributes; Step 3: Machine Learning Module; Step 4: Tumor ablation navigation module; Step 5: Visualization module.
2. According to claim 1, a highly efficient tumor interventional ablation precise positioning system is characterized by: The step 1: CT image processing module specifically includes: Multimodal image fusion: Obtain multimodal imaging data from patients, including CT, MRI, and PET images, and use multimodal image fusion algorithms to register and fuse different image data to ensure spatial consistency of different images; Deep learning improves segmentation accuracy: Use the trained deep learning model to automatically segment the fused image data, identify the tumor area, and post-process the segmentation results to obtain more accurate tumor boundaries.
3. The high-efficiency tumor interventional ablation precise positioning system according to claim 2, characterized in that: The step 2: processing pixels with weak attributes, specifically includes: Increase the weight factor: In the segmentation similarity acquisition module of weak attribute pixels, a weight coefficient is introduced to perform weighted averaging according to the importance of different segmentation results.
4. The high-efficiency tumor interventional ablation precise positioning system according to claim 3, characterized in that: The step three: machine learning module specifically includes: Comprehensive multi-dimensional features: Multi-dimensional features of the tumor are extracted, including morphological features, texture features, and metabolic features, and these features are input into an optimized machine learning model to classify the tumor malignancy grade.
5. The high-efficiency tumor interventional ablation precise positioning system according to claim 1, characterized in that: The step 4: tumor ablation navigation module specifically includes: Multi-probe strategy: According to the final tumor area location detected by the CT image processing module, the insertion path and location of the multiple probes are planned, and the multi-probe synchronous ablation technology is used to achieve comprehensive ablation of the tumor area through the synergistic effect of multiple thermal ablation needles; Real-time feedback control: A real-time feedback control system is configured to monitor the temperature changes around the ablation needle. Based on the temperature feedback data, the power and position of the ablation needle are dynamically adjusted to ensure that the ablation area reaches the expected temperature.
6. The efficient tumor interventional ablation precise positioning system according to claim 1, characterized in that: The step 5: visualization module specifically includes: 3D Visualization: The final tumor area in the CT image processing module is 3D modeled using 3D modeling software. The malignant tumor grade obtained by the machine learning module and the ablation needle position and ablation area obtained by the tumor ablation navigation module are combined in the 3D model for real-time display.