Ablation evaluation method, ablation estimation model generation method and optimization method

Abstract

The invention provides an ablation evaluation method, an ablation estimation model generation method and an ablation estimation model optimization method. The ablation evaluation method comprises the following steps: acquiring a medical image of a patient after ablation; the ablated medical image is an enhanced magnetic resonance image; performing threshold segmentation on the ablated medical image, and extracting to obtain a coarse segmentation boundary of an ablation area; and carrying out smoothing processing on the coarse segmentation boundary to obtain an optimized segmentation boundary. According to the method, the separation boundary of the ablation area and the non-ablation area of the ablation scene is marked by using the contrast agent, so that the boundary of the ablation area in the image is more obvious and continuous, and the smooth segmentation boundary is quickly and efficiently obtained by performing threshold segmentation and smoothing processing on the enhanced medical image after ablation; the boundary can be used for calculating the ablation volume and judging the ablation state, and information support is provided for risk prevention and control and subsequent treatment.

Description

Technical Field

[0001] The present invention relates to the field of medical device technology, and in particular to an ablation evaluation method, an ablation prediction model generation method, and an optimization method. Background Art

[0002] Tissue ablation is the process of using energy to cause tissue cell apoptosis to eliminate specific target areas, such as ultrasound ablation, microwave ablation, radiofrequency ablation, laser ablation, high-voltage pulsed electric field ablation, cryoablation, etc.

[0003] At present, there are many areas that need to be improved in the information support for tissue ablation: 1. In some ablation processes, the temperature map of the target area is collected and input into the ablation prediction model to estimate the current real-time ablation boundary, providing information support for the doctor's ablation operation. However, there is an error between the estimated ablation boundary and the actual ablation boundary, which can easily lead to misjudgment and often requires remedial measures; 2. At present, the boundary between the ablation area and the non-ablation area in the images collected after ablation is not clear enough and cannot be identified. There are also frequent discontinuous boundaries, which brings troubles to the assessment of the ablation status.

[0004] In order to solve or partially solve the defects of the prior art that the ablation prediction model has poor prediction accuracy, the evaluation after ablation is not comprehensive, and the evaluation is difficult, the present invention provides an ablation evaluation method, an ablation prediction model generating method, and an optimization method. Summary of the invention

[0005] The present invention provides an ablation evaluation method, an ablation prediction model generation method and an optimization method, which are used to solve the defects of the prior art that the prediction accuracy of the ablation prediction model is poor, the evaluation after ablation is not comprehensive, and the evaluation is difficult.

[0006] The present invention provides an ablation assessment method, comprising:

[0007] Acquiring a medical image of the patient after ablation; the medical image after ablation is an enhanced magnetic resonance image;

[0008] Performing threshold segmentation on the medical image after ablation to extract a rough segmentation boundary of the ablation area;

[0009] The rough segmentation boundary is smoothed to obtain an optimized segmentation boundary.

[0010] Further, the threshold segmentation of the medical image after ablation to extract the rough segmentation boundary of the ablation area includes: extracting a closed boundary of high signal as the rough segmentation boundary of the ablation area, wherein the closed boundary means that the total length of the boundary accounts for no less than 50% of the fitted closed perimeter, or the total area of ​​the boundary accounts for no less than 50% of the fitted closed area. Preferably, it can be no less than 60%, and more preferably, it can be no less than 70%, 75%, 80%, 85%, or 90%, etc.

[0011] Optionally, the optimized segmentation boundary includes an inner boundary and an outer boundary.

[0012] Optionally, the medical image after ablation is a two-dimensional enhanced magnetic resonance image or a three-dimensional enhanced magnetic resonance image.

[0013] Optionally, performing threshold segmentation on the medical image after ablation to extract a rough segmentation boundary of the ablation area includes:

[0014] Determining an adaptive threshold of the medical image after ablation based on statistics;

[0015] The ablated medical image is segmented according to the adaptive threshold to obtain the rough segmentation boundary.

[0016] Furthermore, the determining the adaptive threshold of the medical image after ablation based on statistics includes:

[0017] Generating a grayscale distribution histogram of the medical image after ablation;

[0018] According to the grayscale distribution histogram, a threshold value that maximizes the grayscale distribution difference on both sides of the threshold value is determined as the adaptive threshold value of the medical image after ablation.

[0019] Optionally, after segmenting the ablated medical image according to the adaptive threshold to obtain the rough segmentation boundary, the method further includes:

[0020] Detecting the closure of the rough segmentation boundary;

[0021] If the rough segmentation boundary is not closed, a new threshold is adjusted based on the adaptive threshold for segmentation until a closed new rough segmentation boundary is obtained through segmentation.

[0022] Optionally, performing threshold segmentation on the medical image after ablation to extract a rough segmentation boundary of the ablation area includes:

[0023] determining a region of interest in the medical image after ablation;

[0024] The rough segmentation boundary is obtained by extracting the region of interest in the medical image after ablation.

[0025] Optionally, determining the region of interest in the medical image after ablation includes:

[0026] Registering the medical image after ablation with the medical image before operation to determine the target area in the medical image after ablation;

[0027] The region of interest to be segmented is determined according to the target region in the ablated medical image.

[0028] Optionally, the smoothing of the rough segmentation boundary to obtain an optimized segmentation boundary includes:

[0029] Performing morphological processing on the rough segmentation boundary to obtain the smoothed optimized segmentation boundary;

[0030] The morphological processing includes at least one of the following: erosion, dilation, opening operation, and closing operation.

[0031] Optionally, the method further includes: calculating the ablation volume according to the optimized segmentation boundary.

[0032] Furthermore, the method also includes: evaluating the ablation effect in combination with the target area of ​​the patient before ablation.

[0033] Furthermore, the step of evaluating the ablation effect in combination with the target area of ​​the patient before ablation includes:

[0034] Obtain medical images of patients before ablation;

[0035] Extracting features from the target area peripheral region in the medical image before ablation, and extracting features from the segmentation boundary peripheral region in the medical image after ablation, and generating a deformation correction matrix according to the extracted features;

[0036] The target area morphology in the medical image before ablation is corrected according to the deformation correction matrix to obtain the simulated target area morphology in the medical image after ablation, and then the ablation effect is evaluated in combination with the segmentation boundary.

[0037] Optionally, the method further comprises:

[0038] The segmentation boundary is compared with the estimated boundary calculated by using the ablation prediction model during the ablation process, and the parameters in the ablation prediction model are optimized.

[0039] Furthermore, the method further comprises:

[0040] The optimized parameters are associated with the patient's pathological information and stored so that in the subsequent ablation process, the corresponding parameters can be determined according to the specific patient's pathological information, which is used to more accurately calculate the estimated boundary during the ablation process.

[0041] The present invention also provides an ablation prediction model optimization method, comprising:

[0042] Acquire the patient's ablation process data;

[0043] Obtaining medical images collected after ablation of the patient, and extracting segmentation boundaries;

[0044] Inputting the patient ablation process data into an ablation prediction model, calculating an estimated boundary, and optimizing the parameters of the ablation prediction model so that the estimated boundary approaches the segmentation boundary;

[0045] The parameters of the ablation prediction model when the predicted boundary is closest to the segmentation boundary are determined as optimized parameters, and are stored in association with the pathological information of the patient.

[0046] The present invention also provides a method for generating an ablation prediction model, comprising:

[0047] According to the pathological information of the current patient, the target pathological information with the highest similarity is matched from the database; wherein the patient pathological information and the parameters of the corresponding ablation prediction model are stored in association in the database, and the parameters of each ablation prediction model are obtained based on the ablation process data of the corresponding patient and the post-ablation segmentation boundary optimization;

[0048] Acquire target parameters of the ablation prediction model corresponding to the target pathology information from the database;

[0049] A target ablation prediction model is generated according to the target parameters, for performing ablation prediction on the current patient.

[0050] The present invention also provides a method for generating an ablation prediction model, comprising:

[0051] According to the pathological information of the current patient, a number of target pathological information with high similarity is matched from a database; wherein the patient pathological information and the parameters of the corresponding ablation prediction model are stored in association in the database, and the parameters of each ablation prediction model are obtained based on the ablation process data of the corresponding patient and the post-ablation segmentation boundary optimization;

[0052] According to the target pathological information and the corresponding ablation prediction model parameters, a parameter generation model is used to generate target parameters corresponding to the current patient pathological information;

[0053] A target ablation prediction model is generated according to the target parameters, for performing ablation prediction on the current patient.

[0054] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of any of the aforementioned ablation assessment methods are implemented, or the steps of any of the aforementioned ablation prediction model optimization methods are implemented, or the steps of any of the aforementioned ablation prediction model generation methods are implemented, or the steps of any of the aforementioned ablation prediction model generation methods are implemented.

[0055] The present invention further provides a laser ablation system, characterized in that it comprises: the electronic device described in any one of the above items, as well as a laser module and a display module.

[0056] The present invention provides an ablation assessment method, an ablation prediction model generation method, and an optimization method, which have at least the following beneficial effects:

[0057] 1. For ablation scenarios, contrast agents are creatively used to mark the boundary between the ablation area and the non-ablation area. The consistency between the segmentation boundary shown in the enhanced image and the actual ablation boundary is confirmed through experiments. Compared with the prior art, the ablation boundary in the image of the present invention is more obvious, the contrast is higher, and the continuity of the ablation boundary is also significantly improved, which helps to more accurately determine the ablation area.

[0058] 2. In some implementations, two-dimensional enhanced magnetic resonance images are obtained to evaluate the ablation effect, which reduces the amount of data collected and saves patient expenses. In some implementations, three-dimensional enhanced magnetic resonance images are obtained to provide a more comprehensive three-dimensional segmentation and evaluation of the ablation area, and can also be combined with preoperative and intraoperative images to evaluate the ablation effect.

[0059] 3. Through adaptive threshold segmentation, the rough segmentation boundary of the ablation area is segmented efficiently and automatically, avoiding the segmentation error caused by the overall brightness difference of the image. By smoothing the rough segmentation boundary, the display effect of the segmentation boundary is optimized, and the optimized boundary is closer to the real ablation area boundary.

[0060] 4. By focusing on the region of interest, only the rough segmentation boundary is extracted from the region of interest, which reduces the amount of data processing and improves the segmentation accuracy. Especially for adaptive threshold segmentation, it reduces the interference of other irrelevant areas on the adaptive threshold segmentation process and improves the segmentation accuracy.

[0061] 5. Use the target area's peripheral area to correct the target area's morphology, and reduce the deviation of the ablation evaluation process caused by the change of the target area's morphology during the ablation process.

[0062] 6. Use the images after ablation to extract the segmentation boundary as the gold standard of the ablation boundary, input the ablation process data into the ablation prediction model to estimate the ablation boundary, and continuously adjust the ablation prediction model to make the estimated ablation boundary close to the segmented boundary. The ablation prediction model is retrospectively optimized using the more credible segmentation boundary determined after ablation, which improves the ablation prediction effect of the model. The optimized coefficients are also stored in the database by associating them with the patient's pathological information, providing a reference for selecting the ablation prediction model for new patients in the future, and determining the ablation prediction model that is more suitable for the patient.

[0063] 7. For new patients, a "dedicated ablation prediction model" that is more suitable for the patient is matched / generated from the database based on the patient's pathological information before the ablation operation, which greatly improves the accuracy of ablation prediction during the ablation process. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0065] Figure 1 It is a flowchart of an ablation evaluation method provided by the present invention;

[0066] Figure 2 is an example of a conventional MR image acquired after ablation;

[0067] Figure 3 is an example of an enhanced magnetic resonance image acquired after ablation;

[0068] Figure 4 These are enhanced magnetic resonance images and brain tissue slices after ablation of animal brain tissue;

[0069] Figure 5 It is a flowchart of an ablation prediction model optimization method provided by the present invention;

[0070] Figure 6 It is a flowchart of a method for generating an ablation prediction model provided by the present invention;

[0071] Figure 7 It is a flowchart of another ablation prediction model generation method provided by the present invention. DETAILED DESCRIPTION

[0072] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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.

[0073] Combine the following Figure 1-Figure 7 Describe the ablation evaluation method, ablation prediction model generation method, and optimization method of the present invention, Figure 1 is a flow chart of an ablation evaluation method provided by the present invention, such as Figure 1 As shown, the method includes:

[0074] S11, obtaining medical images of the patient after ablation; the medical images after ablation are enhanced magnetic resonance images;

[0075] S12, performing threshold segmentation on the medical image after ablation to extract a rough segmentation boundary of the ablation area;

[0076] S13, smoothing the rough segmentation boundary to obtain an optimized segmentation boundary.

[0077] Specifically, for tissue ablation, such as ultrasound ablation, microwave ablation, radiofrequency ablation, laser ablation, high-voltage pulsed electric field ablation, cryoablation, etc., the evaluation of ablation is divided into monitoring and estimation during the ablation process and actual evaluation after ablation. Compared with monitoring and estimation during the ablation process, there is sufficient time to collect high-quality images after ablation. The enhanced medical images collected after the patient's ablation have higher resolution and wider coverage, which helps to accurately evaluate the ablation status. The current "post-ablation evaluation" usually collects conventional magnetic resonance images, CT, ultrasound and other images for evaluation, for example Figure 2 The magnetic resonance T2 images shown have low differentiation between the ablation area and the surrounding tissues, but high similarity. The differentiation between the ablation area and the non-ablation area is poor, and the separation boundary is often discontinuous, especially when there is a large area of ​​edema around the ablation area. The "separation boundary" is more discontinuous. Since the separation boundary is fuzzy and discontinuous, doctors usually draw a voxel-level boundary based on personal experience, which has a large error.

[0078] The present invention preferably uses magnetic resonance images (i.e., enhanced magnetic resonance images) collected after contrast agents are used for the post-ablation medical images collected from patients. Generally, enhanced magnetic resonance images are used to enhance the development effect of structures such as blood vessels, thyroid gland, and liver. Different from this, the present invention is based on the following assumptions: 1. Under normal conditions, due to the existence of a "blood-brain barrier" between blood and brain tissue, contrast agents cannot enter brain tissue. However, for the specific scenario of brain ablation, the ablation operation destroys the capillaries in and around the ablation area, and the contrast agent can break through the "blood-brain barrier" through the destroyed capillaries and be released to the surrounding area of ​​the ablation area; 2. After ablation, the metabolic activity near the ablation area is relatively high, and the concentration of contrast agents accumulated nearby is also relatively high. At the same time, the cells in the ablation area lose their activity, and the contrast agent cannot enter them. As a whole, the contrast agent accumulates at the boundary separating the ablation area and the non-ablation area, and is displayed as a circle of high-signal boundaries in the collected post-ablation medical images, and the high-signal boundaries are highly overlapped with the boundaries of the ablation area. The present invention is based on one of the above assumptions or the combined effect of two assumptions and utilizes enhanced magnetic resonance imaging to "mark" the boundary of the ablation area. Figure 3 That is, an example of enhanced T1 image after ablation is shown. It can be seen that an obvious high signal closed boundary appears in the figure, which can be used as the boundary of the ablation area. The inventors also verified the above hypothesis through animal experiments. Figure 4 For example, Figure 4 The upper middle image is the enhanced T1 image after ablation of animal brain tissue. Figure 4 The lower middle part is a tissue section at the corresponding position after ablation of animal brain tissue. After staining, the range of ablated tissue can be displayed. The inventors have concluded through multiple experimental comparisons that the high signal boundary in the enhanced magnetic resonance image is consistent with the boundary of the ablation area in the actual tissue, and can be used as a criterion for judging the ablation boundary.

[0079] Of course, the enhanced MRI image can also be an equivalent enhanced MRI image, for example, a T2 image is acquired after ablation, and a T1 enhanced image is generated based on it. For another example, a T1 and CT image are acquired after ablation, and a T1 enhanced image is generated by combining the T1 image and the CT image to enhance the development effect of a specific structure in the image.

[0080] On this basis, threshold segmentation is performed on the enhanced medical images of patients after ablation, which can accurately, simply and quickly extract high-signal boundaries with better closure and higher contrast as the rough segmentation boundaries of the ablation area. On the basis of the rough segmentation boundary, the rough segmentation boundary is further smoothed to optimize the display effect. The optimized segmentation boundary is closer to the real ablation area boundary and can be used to calculate the ablation area / volume and judge the ablation status (for example, transitional ablation / under-ablation), providing information support for risk prevention and control and subsequent treatment.

[0081] In this embodiment, contrast agents are used to mark the boundary between the ablation area and the non-ablation area in the ablation scene, so that the boundary of the ablation area in the image is more obvious, the contrast is higher, and the continuity and closure of the boundary are significantly improved. By performing threshold segmentation and smoothing on the enhanced medical image after ablation, a smooth segmentation boundary is accurately, quickly and efficiently obtained. The boundary can be used to calculate the ablation area / volume and judge the ablation status, providing information support for risk prevention and control and subsequent treatment.

[0082] Based on any embodiment, in one embodiment, the optimized segmentation boundary includes an inner boundary and an outer boundary.

[0083] Specifically, in enhanced magnetic resonance imaging, since contrast agents accumulate near the ablation boundary, they appear as high-signal boundaries with a certain width on the medical image, including an inner boundary adjacent to the target area and an outer boundary away from the target area.

[0084] Based on any of the embodiments, in one embodiment, the medical image after ablation is a two-dimensional enhanced magnetic resonance image.

[0085] In another embodiment, the post-ablation medical image is a three-dimensional enhanced magnetic resonance image.

[0086] Specifically, the acquisition of two-dimensional enhanced magnetic resonance images can only understand the ablation status of the corresponding acquisition section, while the target area is three-dimensional, and the ablation status cannot be fully evaluated based on the two-dimensional enhanced magnetic resonance images alone. On the other hand, if it is to be compared with the ablation scheme planned before surgery and the ablation range estimated during surgery, it is also necessary to ensure that the image acquisition position of each stage is the same section, which increases the difficulty of acquiring images and makes it difficult to combine preoperative and intraoperative images to evaluate the ablation effect. In this embodiment, three-dimensional enhanced magnetic resonance images are acquired (that is, enhanced magnetic resonance images are acquired at multiple continuous faults), and the extracted segmentation boundaries are also three-dimensional, so the ablation evaluation effect is more comprehensive. On the other hand, if it is to be compared with the ablation scheme planned before surgery and the ablation range estimated during surgery, the corresponding two-dimensional section can be determined after registration, and the corresponding two-dimensional section can be intercepted in the three-dimensional enhanced magnetic resonance image, so as to jointly evaluate the ablation effect at the same fault position.

[0087] Based on any embodiment, in some embodiments, S12 includes: extracting a closed boundary of high signal as a rough segmentation boundary of the ablation area, wherein the closed boundary means that the total length of the segmentation boundary accounts for no less than 50% of the fitted closed perimeter, or the total area of ​​the segmentation boundary accounts for no less than 50% of the fitted closed area.

[0088] Specifically, the present invention extracts the high-signal closed boundary in the enhanced image after ablation of the patient as the ablation boundary based on the consistency between the high-signal boundary and the ablation boundary, thereby improving the accuracy of determining the ablation boundary. Although the present invention improves the contrast of the ablation boundary and the closure of the boundary, and can extract a 100% closed boundary in most cases, for a small number of patients (for example, older patients with lower metabolic rates), the above-mentioned discontinuous boundaries may still occur, but the continuity of the boundaries is still improved, and the "closed" ratio is not less than 50%. Preferably, it can be no less than 60%, and more preferably, it can be no less than 70%, 75%, 80%, 85%, or 90%, etc.

[0089] For two-dimensional enhanced magnetic resonance images, a closed boundary means that the total length of the segmented boundary accounts for no less than 50% of the fitted closed perimeter. For example, if the boundary includes one or more segments, the total length of each segment boundary is calculated, and the missing segments are connected by straight lines, splines, etc. to obtain a fitted closed boundary, and its perimeter is calculated, or the closed boundary is fitted in combination with image information. For example, if there are multiple crescent-shaped boundaries, and the signal strength of the multiple crescent-shaped boundaries is inconsistent, the image can be divided into blocks according to the multiple boundary segments, and the boundaries are determined by adaptive thresholds respectively, and then connected into a closed boundary in combination with the shapes of the multiple boundary segments, and the length of the multiple boundary segments and the perimeter of the closed boundary are calculated. In short, the above total length accounts for no less than 50% of the perimeter. For three-dimensional enhanced magnetic resonance images, a closed boundary means that the total area of ​​the segmented boundary accounts for no less than 50% of the fitted closed area. For example, if the boundary includes one or more curved surfaces, the total area of ​​each curved surface is calculated, and the missing curved surface is smoothly supplemented to calculate its fitted closed area. The above total area accounts for no less than 50% of the fitted closed area. Preferably, it may be not less than 60%, more preferably, it may be not less than 70%, 75%, 80%, 85%, or 90%, etc.

[0090] Furthermore, since the boundary has a certain "thickness", the above perimeter is calculated after extracting the center line of the segmentation boundary with thickness, and the above area is calculated after extracting the center surface of the segmentation boundary with thickness.

[0091] Based on any embodiment, in one embodiment, S12 includes:

[0092] S121, determining an adaptive threshold of the medical image after ablation based on statistics;

[0093] S122 . Segment the ablated medical image according to the adaptive threshold value to obtain a rough segmentation boundary.

[0094] Specifically, the adaptive threshold can flexibly determine the appropriate threshold according to the specific image, thereby improving the adaptability and accuracy of the algorithm, especially for different images with large differences in overall brightness, the ablation area can be accurately segmented.

[0095] Based on the previous embodiment, in one embodiment, S121 includes:

[0096] Generate a grayscale distribution histogram of the medical image after ablation;

[0097] According to the grayscale distribution histogram, the threshold that maximizes the grayscale distribution difference on both sides of the threshold is determined as the adaptive threshold of the medical image after ablation.

[0098] Specifically, the grayscale distribution histogram of the medical image is first statistically analyzed, and then a threshold to be determined is assumed to be determined, and the distribution of each grayscale on the left side of the threshold (less than the threshold) and the grayscale distribution on the right side of the threshold (greater than the threshold) are determined respectively, so that the greater the difference in grayscale distribution on both sides of the threshold, the better the segmentation effect of the threshold. The threshold with the largest difference in grayscale distribution on both sides of the threshold is determined as the adaptive threshold of the medical image after ablation, which can better segment the high signal boundary.

[0099] Based on the previous embodiment, in one embodiment, after the medical image after ablation is segmented according to the adaptive threshold value to obtain a rough segmentation boundary, the method further includes:

[0100] Check the closure of the rough segmentation boundary;

[0101] If the coarse segmentation boundary is not closed, a new threshold is adjusted based on the adaptive threshold for segmentation until a closed new coarse segmentation boundary is obtained.

[0102] Specifically, the method of the present invention can improve the contrast of the ablation area boundary in the medical image, and greatly improve the continuity and closure of the segmentation boundary. However, for a very small number of patients (for example, those who are older and have a lower metabolic rate), the above-mentioned "segmentation boundary" may still be discontinuous. In this regard, after detecting the discontinuity of the rough segmentation boundary, the adaptive threshold is adjusted to obtain a new threshold and the threshold segmentation is performed again until the segmentation obtains a closed rough segmentation boundary. For example, the new threshold is determined by reducing 1 (i.e., 1 grayscale level) each time on the basis of the adaptive threshold until the segmentation obtains a closed new rough segmentation boundary, and for example, it is reduced by 2, 3, and so on each time.

[0103] Based on any of the above embodiments, in one embodiment, threshold segmentation is performed on the medical image after ablation to extract a rough segmentation boundary of the ablation area, including:

[0104] Determine regions of interest in medical images after ablation;

[0105] The region of interest in the ablated medical image is extracted to obtain a rough segmentation boundary.

[0106] Specifically, by focusing on the region of interest and extracting only the coarse segmentation boundary of the region of interest, the amount of data processing can be reduced and the segmentation accuracy can be improved, especially for adaptive threshold segmentation, which reduces the interference of other irrelevant areas on the "adaptive threshold" and improves the segmentation accuracy.

[0107] Based on the previous embodiment, in one embodiment, determining a region of interest in a medical image after ablation includes:

[0108] Registering the medical image after ablation with the medical image before ablation to determine the target area in the medical image after ablation;

[0109] The region of interest to be segmented is determined according to the target region in the medical image after ablation.

[0110] Specifically, the lesion area is segmented in the medical image before ablation, and the target area to be ablated is determined. By aligning the medical image after ablation with the medical image before ablation, the target area in the medical image before ablation can be mapped to the medical image after ablation. On this basis, the region of interest can be determined, and the segmentation boundary can be extracted more accurately and efficiently.

[0111] It should also be noted that the "lesion area" in the present invention is the range of lesions determined based on medical images collected before ablation, such as tumors and epilepsy; the "target area" refers to the area where the ablation operation needs to be performed, which is usually not completely consistent with the lesion area. For example, a certain range is expanded on the basis of the determined tumor range as the target area to ensure that the tumor is completely treated and the possibility of recurrence is reduced; the "ablation area" refers to the area where the ablation operation causes tissue degeneration and necrosis. Due to equipment errors and operating errors, the "ablation area" usually does not necessarily completely overlap with the target area; the "region of interest" is a certain range covering the target area. For example, with the "target area" as the center, a range twice the area of ​​the target area is determined as the "region of interest". For another example, with the "target area" as the center, a range three times the diameter of the target area is determined as the "region of interest".

[0112] Based on any embodiment, in one embodiment, S13 includes:

[0113] Perform morphological processing on the rough segmentation boundary to obtain a smoothed segmentation boundary;

[0114] The morphological processing includes at least one of the following: erosion, dilation, opening operation, and closing operation.

[0115] Specifically, the corrosion process can be understood as using a structural element with a certain volume and shape to move inside the area to be processed, and all possible positions that the origin of the structural element passes through during the movement constitute the result after corrosion. Each time "corrosion" is performed, the area to be processed is "reduced". Similarly, the expansion process can be understood as using a structural element with a certain volume and shape to move its origin within the area to be processed, and all possible positions that the structural element passes through during the movement constitute the result after expansion. The opening operation is corrosion first and then expansion, which can play a "burr removal" effect. The closing operation is expansion first and then corrosion, which can play a "gap filling" role. In this embodiment, by performing morphological processing on the area corresponding to the rough segmentation boundary, the segmentation boundary is smoother, the display effect is better, and it is closer to the real ablation area boundary. Further, combined with the embodiment of collecting three-dimensional enhanced magnetic resonance images, the above morphological processing is a three-dimensional morphological processing, so that the optimized segmentation boundary has a "three-dimensional smoothing" effect, and there is no "burr" at the transition between layers.

[0116] Based on any embodiment, in one embodiment, the method further includes: calculating an ablation area or an ablation volume according to the optimized segmentation boundary.

[0117] Specifically, the aforementioned segmentation boundary is the boundary of the ablation area determined based on the medical image. The segmentation boundary is similar to a shell with thickness, which covers the ablation area. According to the "spherical shell", the area of ​​the ablation area at the acquisition section of the two-dimensional enhanced magnetic resonance image can be calculated or the volume of the ablation area in the three-dimensional enhanced magnetic resonance image can be calculated. Furthermore, since the segmentation boundary has a certain thickness, the voxel-level boundary can be further determined, which is convenient for accurately calculating the ablation area / volume. In a preferred embodiment, the voxel-level simulated ablation boundary is determined in the segmentation boundary according to the ablated tissue type. For example, if a tumor is ablated, the inner boundary of the segmentation boundary is used as the voxel-level simulated ablation boundary for calculating the ablation volume; for another example, if an epileptic zone is ablated, the outer boundary of the segmentation boundary is used as the voxel-level simulated ablation boundary for calculating the ablation volume; for another example, if a nodule is ablated, the voxel-level simulated ablation boundary is determined between the inner and outer boundaries of the segmentation boundary according to the distance ratio corresponding to the "nodule", for example, the preset distance ratio corresponding to the "nodule" is "distance from the inner boundary: distance from the outer boundary = 4:1", and the voxel-level simulated ablation boundary is determined according to the distance ratio. It can be understood that the process of determining the voxel-level simulated ablation boundary in the aforementioned "tumor" and "epilepsy" cases can be understood as determining the voxel-level simulated ablation boundary according to a special preset distance ratio.

[0118] This embodiment determines the simulated ablation boundary at the voxel level in a differentiated and standardized manner according to tissue types, and calculates the ablation area or ablation volume, so that doctors can intuitively understand the ablation amount.

[0119] Further, based on the previous embodiment, in one embodiment, the method further includes: evaluating the ablation effect in combination with the target area of ​​the patient before ablation.

[0120] Specifically, the ablation area / volume is determined in the previous embodiment, and the ablation effect can be roughly evaluated in combination with the area / volume of the target area before ablation. That is, the ablation effect is evaluated based on the overlap between the area delineated by the ablation boundary at the voxel level and the target area. For example, the ratio of the number of overlapping voxels to the number of voxels in the target area is calculated as the ablation completion degree. For another example, the ratio of the number of overlapping voxels to the "sum of the number of voxels in the target area and the segmented boundary area" is calculated as the ablation accuracy. For another example, the ratio of the number of over-ablated voxels outside the target area to the number of voxels in the target area is calculated as the over-ablation ratio, and so on.

[0121] This embodiment accurately and standardizedly evaluates the ablation effect by combining the target area before ablation, providing information support for the doctor's subsequent treatment.

[0122] Based on the above embodiments, in one embodiment, evaluating the ablation effect in combination with the target area of ​​the patient before ablation includes:

[0123] Obtain medical images of patients before ablation;

[0124] Extracting features from the peripheral area of ​​the target area in the medical image before ablation, and extracting features from the peripheral area of ​​the segmentation boundary in the medical image after ablation, and generating a deformation correction matrix according to the features extracted above;

[0125] The target area morphology in the medical image before ablation is corrected according to the deformation correction matrix to obtain the simulated target area morphology in the medical image after ablation, and then the ablation effect is evaluated in combination with the segmentation boundary.

[0126] Specifically, in the process of using energy to ablate the target area, the morphology of the target area may change, such as the shrinkage of the tumor due to heat, or the expansion of tissue fluid in the local area due to vaporization. In this embodiment, the images before and after ablation are registered after feature extraction to generate a deformation correction matrix of the target area during the ablation process, which can correct the morphological changes of the target area and more accurately evaluate the ablation state. The above-mentioned extracted features, such as vascular branch points, brain sulci, organ boundaries, etc. around the target area, can be understood that the above-mentioned registration is a non-rigid registration, which can more accurately correct the morphology of the target area.

[0127] Based on any embodiment, in one embodiment, the method further includes:

[0128] The segmented boundary is compared with the estimated boundary calculated by using the ablation prediction model during the ablation process, and the parameters in the ablation prediction model are optimized.

[0129] Specifically, the temperature of the tissue and the duration of the temperature determine the ablation state of the tissue (i.e., whether degeneration and necrosis occur). The ablation prediction model can be an Arrhenius model, a CEM43 model, etc. The ablation prediction model can calculate the ablation state of each voxel according to the patient temperature map of the entire ablation stage. Some parameters in the model will affect the ablation results. The parameters of the ablation prediction model are corrected using the segmentation boundaries determined based on the medical images after ablation so that the calculation results are as close as possible to the segmentation boundaries determined after ablation. It can be understood that the ablation boundaries estimated by the updated model are closer to the above-mentioned segmentation boundaries determined based on the medical images after ablation.

[0130] In one example, the Arrhenius model is used, and the specific formula of the model is:

[0131]

[0132] Among them, Ω is the ablation judgment index, which is used to judge the ablation status of the voxel in combination with the threshold, A is the pre-exponential factor, and E a is the activation energy, T is the absolute temperature, and R is the gas constant. a It can be used as a parameter to be optimized.

[0133] In another example, the CEM43 model is used, and the specific formula of the model is:

[0134]

[0135] Among them, R is a parameter related to cell and tissue characteristics, t is the time in minutes, and T t is the temperature in degrees Celsius. The above R can be used as a parameter to be optimized.

[0136] In another example, the model used is:

[0137]

[0138] Among them, A(t) is the ablation judgment index, which is used to judge the ablation status of the voxel in combination with the threshold, B(m) is a function related to the tissue type, γ represents the nuclear gyromagnetic ratio, α is the temperature coefficient of the shielding constant, B0 is the main magnetic field intensity, TE is the echo time, φ(T) is the phase of the current image (after heating), and φ0 is the phase of the reference image (before heating). B(m) and α can be used as parameters to be optimized.

[0139] In this embodiment, the parameters in the ablation prediction model are optimized to improve the accuracy of the ablation prediction model.

[0140] Based on the previous embodiment, in one embodiment, the method further includes:

[0141] The optimized parameters are associated with the patient's pathological information and stored so that in the subsequent ablation process, the corresponding parameters can be determined according to the specific patient's pathological information, which is used to more accurately calculate the estimated boundary during the ablation process.

[0142] Specifically, the patient's pathological information, such as the patient's age, gender, disease type, target tissue type, etc., is associated with the tissue composition of the target area, which will affect the process of heat conduction and tissue degeneration and necrosis. Storing the optimized parameters in association with the patient's pathological information can provide a reference for selecting ablation formula parameters for new patients in the future, determine an ablation prediction model that is more suitable for the patient, and improve the accuracy of ablation prediction.

[0143] The present invention further provides an electronic device. The electronic device described below and the ablation evaluation method described above can refer to each other.

[0144] The present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of any of the aforementioned ablation evaluation methods when executing the program. That is, the aforementioned ablation evaluation method runs in the electronic device in the form of a computer program, providing information support for doctors to evaluate ablation status.

[0145] An ablation prediction model optimization method provided by the present invention is described below. The ablation prediction model optimization method described below and the ablation evaluation method described above can refer to each other.

[0146] Figure 5 is a flow chart of an ablation prediction model optimization method provided by the present invention, such as Figure 5 As shown, the present invention provides an ablation prediction model optimization method comprising:

[0147] S21, obtaining ablation process data of the patient;

[0148] S22, obtaining medical images collected after ablation of the patient, and extracting segmentation boundaries;

[0149] S23, inputting the patient's ablation process data into the ablation prediction model, calculating the estimated boundary, and optimizing the parameters of the ablation prediction model so that the estimated boundary approaches the segmentation boundary;

[0150] S24. Determine the parameters of the ablation prediction model when the predicted boundary is closest to the segmentation boundary as optimized parameters, and store them in association with the pathological information of the patient.

[0151] Specifically, the ablation process data here refers to the monitoring data collected during the entire ablation stage, such as all magnetic resonance temperature images continuously collected and generated for the target area during ultrasonic ablation and laser ablation. The temperature image is used to calculate the ablation state of each voxel in combination with the ablation prediction model. After ablation, there is sufficient time to collect medical images of the patient, and images with a wider spatial coverage and higher resolution can be collected to accurately determine the ablation boundary. For example, after ultrasonic ablation and laser ablation, enhanced magnetic resonance tissue images are collected at continuous faults covering the target area. The resolution of the image is high, so that the segmented boundary is closer to the actual ablation area boundary.

[0152] After obtaining the ablation process data and the segmentation boundary after ablation, the ablation process data is input into the ablation prediction model, and the model will output an estimated ablation boundary. By gradually adjusting and optimizing the parameters of the ablation prediction model, the estimated ablation boundary output by it is close to the above segmentation boundary. It can be understood that the closer the estimated boundary is to the segmentation boundary, the better the effect of the ablation prediction model. The ablation prediction model corresponding to the "estimated boundary closest to the segmentation boundary" is determined as the optimized ablation prediction model. It can be understood that the "closest" here can be the "closest" result screened out under a certain optimization number limit, or it can be the "closest" result determined by stopping the iteration after the optimization increment of the evaluation index is less than the threshold after multiple iterations of optimization. On the other hand, we believe that the patient's pathological information is associated with the tissue composition of the target area, which will affect the heat conduction and tissue degeneration and necrosis process. Based on this understanding, this embodiment associates the optimized coefficient with the patient's pathological information and stores it in a database. The database can be continuously supplemented and updated, and can provide a reference for the subsequent selection of ablation prediction models for new patients, determine the ablation prediction model that is more suitable for the patient, and improve the accuracy of ablation prediction. The above-mentioned patient pathological information includes patient age, patient gender, disease type, target area tissue type, etc.

[0153] This embodiment extracts segmentation boundaries based on medical images after ablation as the gold standard of ablation boundaries, inputs ablation process data into the ablation prediction model to estimate the ablation boundaries, and continuously adjusts the ablation prediction model to make the estimated ablation boundaries close to the segmented boundaries. The ablation prediction model is retrospectively optimized using the more credible segmentation boundaries determined after ablation, thereby improving the ablation prediction effect of the model. The optimized coefficients are also stored in association with the patient's pathological information to provide a reference for subsequent selection of ablation prediction models for new patients, thereby determining an ablation prediction model that is more suitable for the patient.

[0154] Furthermore, step S22 may refer to any of the aforementioned “ablation evaluation methods” to obtain the segmentation boundary after ablation.

[0155] Based on any of the foregoing embodiments, in one embodiment, in step S23, the proximity between the estimated boundary and the segmentation boundary is evaluated in the following manner:

[0156] The number of voxels within the estimated boundary and the number of voxels within the segmentation boundary are calculated respectively, and the proximity between the estimated boundary and the segmentation boundary is evaluated by combining the number of overlapping voxels within the estimated boundary and the segmentation boundary.

[0157] Specifically, assuming that the number of voxels within the estimated boundary is a, the number of voxels within the segmentation boundary is b, and the number of overlapping voxels within the estimated boundary and the segmentation boundary is c. In one example, the closeness between the estimated boundary and the segmentation boundary can be evaluated by c / max(a,b), and the larger the index is, the closer the closer the boundary is; in another example, the closeness between the estimated boundary and the segmentation boundary can be evaluated by 2c / (a+b), and the larger the index is, the closer the closer the boundary is; equivalent deformations can also be performed based on the above examples, such as multiplying by a certain coefficient, adding a certain constant, taking the inverse, and so on.

[0158] Based on the above embodiment, in one embodiment, in step S23, the proximity between the estimated boundary and the segmentation boundary is evaluated in the following manner:

[0159] The number of voxels that belong exclusively to the estimated boundary and the number of voxels that belong exclusively to the segmentation boundary are calculated, and the sum of the two is used to evaluate the closeness between the estimated boundary and the segmentation boundary.

[0160] It can be understood that “the number of voxels that belong exclusively to the estimated boundary” refers to the number of voxels that belong to the estimated boundary but not to the segmentation boundary, and “the number of voxels that belong exclusively to the segmentation boundary” refers to the number of voxels that belong to the segmentation boundary but not to the estimated boundary. The smaller the sum of the two, the closer the estimated boundary is to the segmentation boundary. Of course, equivalent deformations can be made based on this example, such as multiplying by a certain coefficient, adding a certain constant, taking the inverse, etc. For another example, the number of voxels m that belong exclusively to the estimated boundary, the number of voxels n that belong exclusively to the segmentation, and the number of voxels N within the segmentation boundary are calculated, and the closeness of the estimated boundary and the segmentation boundary is evaluated in the form of a percentage by calculating (m+n) / N.

[0161] The present invention further provides an electronic device. The electronic device described below and the ablation prediction model optimization method described above can refer to each other.

[0162] The present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of any of the above-mentioned ablation model optimization methods when executing the program. That is, the above-mentioned ablation prediction model optimization method runs in the electronic device in the form of a computer program, and after completing tissue ablation, the ablation prediction model is retrospectively optimized in combination with the collected medical images.

[0163] A method for generating an ablation prediction model provided by the present invention is described below. The ablation prediction model generating method described below and the ablation evaluation method and the ablation prediction model optimization method described above can refer to each other.

[0164] like Figure 6 As shown, the present invention provides a method for generating an ablation prediction model, comprising:

[0165] S31, matching the target pathological information with the highest similarity from the database according to the pathological information of the current patient; wherein the database stores the patient pathological information and the parameters of the corresponding ablation prediction model in association, and the parameters of each ablation prediction model are obtained according to the ablation process data of the corresponding patient and the post-ablation segmentation boundary optimization;

[0166] S32, obtaining target parameters of the ablation prediction model corresponding to the target pathology information from a database;

[0167] S33. Generate a target ablation estimation model according to the target parameters, for performing ablation estimation on the current patient.

[0168] Specifically, in this embodiment, a database is pre-constructed, and the database stores the pathological information of many patients and the optimized ablation prediction model parameters suitable for the patients. After obtaining the pathological information of the current patient, the target pathological information with the highest similarity is matched from the database, and the target parameters corresponding to the target pathological information are substituted into the standard ablation prediction model to generate a target ablation prediction model for ablation prediction of the current patient. In short, it is to find the patient with the most similar pathological information, and use the optimized ablation prediction model of the patient to perform ablation prediction on the current patient.

[0169] This embodiment uses medical imaging data of many patients after ablation to retrospectively optimize the ablation prediction model, and constructs a database to store the patient's pathological information in association with the parameters of the ablation prediction model. In the subsequent treatment of specific patients, the most appropriate and accurate ablation prediction model can be determined from the database, greatly improving the accuracy of ablation prediction during the ablation process.

[0170] Furthermore, the parameter optimization process of the ablation prediction model for each patient in the database may refer to any of the “ablation prediction model optimization methods” provided above.

[0171] Furthermore, the above S31 includes:

[0172] Generate a pathology vector for each patient's pathology information in the database;

[0173] Generate a pathology vector to be matched according to the pathology information of the current patient;

[0174] The pathology vector having the smallest angle with the pathology vector to be matched is determined among the pathology vectors in the database, and its corresponding pathology information is determined as the target pathology information.

[0175] Specifically, the pathological information of the patient, such as the patient's age, patient's gender, disease type, target tissue type, etc., can be generated by generating a pathological information vector, and the target pathological information closest to the current patient's pathological information can be determined more accurately and efficiently through vector angle matching. Furthermore, each attribute in the pathological information is normalized, for example, the age of 15 to 100 years old is scaled to the interval [0,1], male gender is set to 0, female gender is set to 1, epilepsy is set to 0, and tumor is set to 1. Through normalization, the pathological information of different attributes is closer in order of magnitude, and the matching and differentiation effect of the calculated angle is better.

[0176] The present invention further provides an electronic device. The electronic device described below and the ablation prediction model generation method described above can refer to each other.

[0177] The present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the program, the steps of any of the aforementioned ablation prediction model generation methods are implemented. That is, the aforementioned ablation prediction model generation method is run in the electronic device in the form of a computer program, and before performing the ablation operation, the method is used to determine an ablation prediction model suitable for the current patient with high prediction accuracy.

[0178] Another ablation prediction model generation method provided by the present invention is described below. The ablation prediction model generation method described below and the ablation evaluation method, ablation prediction model optimization method, and ablation prediction model generation method described above can correspond to each other.

[0179] Reference Figure 7 Another ablation prediction model generation method provided by the present invention includes:

[0180] S41, according to the pathological information of the current patient, matching several target pathological information with high similarity from the database; wherein the patient pathological information and the parameters of the corresponding ablation prediction model are stored in association in the database, and the parameters of each ablation prediction model are obtained based on the ablation process data of the corresponding patient and the post-ablation segmentation boundary optimization;

[0181] S42, generating target parameters corresponding to the current patient's pathological information using a parameter generation model according to a number of target pathological information and the corresponding parameters of the ablation prediction model;

[0182] S43. Generate a target ablation estimation model according to the target parameters, for performing ablation estimation on the current patient.

[0183] Specifically, in this embodiment, a database is pre-constructed, and the database stores the pathological information of many patients and the optimized ablation prediction model parameters suitable for the patients. The specific optimization process can refer to any of the "ablation prediction model optimization methods" provided in the previous text. After obtaining the pathological information of the current patient, a number of target pathological information with a high degree of similarity is matched from the database. According to the optimized ablation prediction model parameters corresponding to these target pathological information, the model parameters corresponding to the pathological information of the current patient are generated, and the target parameters are substituted into the standard ablation prediction model to generate a target ablation prediction model for ablation prediction of the current patient during the ablation process. It can be understood that, unlike the ablation prediction model generation method described above, the target parameters generated in this implementation are not necessarily parameters already in the database. In this embodiment, by matching a number of patients with similar pathological information to the current patient in the database, the model parameters corresponding to the pathological information of the current patient are fitted using the parameter generation model based on the "pathological information-model parameters" of these patients. The model parameters are more in line with the actual condition of the current patient and can better predict the ablation state during the ablation process.

[0184] The parameter generation model is, for example, an interpolation model. The interpolation process of the interpolation model is illustrated below through a more detailed embodiment.

[0185] Assume that the current patient's pathological information is (A i , B i , C i , D i ), and matched 3 patients with the closest pathological information in the database:

[0186] Patient 1’s pathological information (A1, B1, C1, D1), and its optimized ablation prediction model parameters are (a1, b1, c1);

[0187] Patient 2’s pathological information (A2, B2, C2, D2), and its optimized ablation prediction model parameters are (a2, b2, c2);

[0188] Patient 3’s pathological information (A3, B3, C3, D3) and its optimized ablation prediction model parameters are (a3, b3, c3).

[0189] You can use parameter a x is considered as a function of pathological information (A, B, C, D), and “the pathological information of the current patient - parameter a i " is a point on the function, while "pathological information of patient 1 - parameter a1", "pathological information of patient 2 - parameter a2", and "pathological information of patient 3 - parameter a3" can be regarded as three discrete points of the function and are located at "pathological information of the current patient - parameter a i Therefore, the function value of the current patient can be determined by multidimensional interpolation based on these three discrete points, that is, the parameter a can be determined by interpolation. i Similarly, other target parameters b of the current patient can also be determined i 、c i .

[0190] It is understandable that in the above example, the pathological information has four attributes, and four-dimensional interpolation is required to determine its target parameters one by one. Of course, the number of attributes of the pathological information can also be 2, 3, 5 or more, and interpolation of other number dimensions is required.

[0191] Another example of a parameter generation model is a deep learning model. Specifically, the database stores the "pathological information-model parameters" of many patients in association. Training samples can be made based on the "pathological information-model parameters" in the database to train a "deep model that can fit the ablation prediction model parameters". The process of making training samples is as follows: for a certain pathological information, several target pathological information with high similarity are matched from the database, and these target pathological information are used as samples. The model parameters corresponding to the pathological information are used as corresponding annotations to form a training sample. By referring to this process, more training samples can be generated for training to obtain a parameter generation model.

[0192] Furthermore, the parameter optimization process of the ablation prediction model for each patient in the aforementioned database may refer to any of the “ablation prediction model optimization methods” provided above.

[0193] The present invention further provides an electronic device. The electronic device described below and the ablation prediction model generation method described above can refer to each other.

[0194] The present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of any of the above-mentioned ablation prediction model generation methods are implemented. That is, the above-mentioned ablation prediction model generation method is run in the electronic device in the form of a computer program, and before performing the ablation operation, the method is used to generate an ablation prediction model suitable for the current patient with high prediction accuracy.

[0195] The present invention also provides a laser ablation system, which includes any of the electronic devices mentioned above and can execute the corresponding methods. The system also includes a laser module and a display module. The laser module can output laser to perform ablation operations on the ablation area, and the display module can display the collected patient medical images, segmentation results and other contents. Furthermore, the display module also interacts with the operator to obtain the instructions input by the operator.

[0196] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0197] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An ablation assessment method, characterized in that: include: Acquiring a medical image of the patient after ablation; the medical image after ablation is an enhanced magnetic resonance image; Performing threshold segmentation on the medical image after ablation to extract a rough segmentation boundary of the ablation area; The rough segmentation boundary is smoothed to obtain an optimized segmentation boundary.

2. The ablation assessment method according to claim 1, characterized in that: The optimized segmentation boundary includes an inner boundary and an outer boundary.

3. The ablation assessment method according to claim 1, characterized in that: The medical image after ablation is a two-dimensional enhanced magnetic resonance image or a three-dimensional enhanced magnetic resonance image.

4. The ablation assessment method according to claim 1, characterized in that: The step of performing threshold segmentation on the ablated medical image to extract a rough segmentation boundary of the ablation area includes: Determining an adaptive threshold of the medical image after ablation based on statistics; The ablated medical image is segmented according to the adaptive threshold to obtain the rough segmentation boundary.

5. The ablation assessment method according to claim 4, characterized in that: The step of determining the adaptive threshold of the medical image after ablation based on statistics comprises: Generating a grayscale distribution histogram of the medical image after ablation; According to the grayscale distribution histogram, a threshold value that maximizes the grayscale distribution difference on both sides of the threshold value is determined as the adaptive threshold value of the medical image after ablation.

6. The ablation assessment method according to claim 5, characterized in that: After segmenting the ablated medical image according to the adaptive threshold to obtain the rough segmentation boundary, the method further includes: Detecting the closure of the rough segmentation boundary; If the rough segmentation boundary is not closed, a new threshold is adjusted based on the adaptive threshold for segmentation until a closed new rough segmentation boundary is obtained through segmentation.

7. The ablation assessment method according to claim 1, characterized in that: The step of performing threshold segmentation on the ablated medical image to extract a rough segmentation boundary of the ablation area includes: determining a region of interest in the medical image after ablation; The rough segmentation boundary is obtained by extracting the region of interest in the medical image after ablation.

8. The ablation assessment method according to claim 7, characterized in that: The determining of the region of interest in the medical image after ablation comprises: Registering the medical image after ablation with the medical image before operation to determine the target area in the medical image after ablation; The region of interest to be segmented is determined according to the target region in the ablated medical image.

9. The ablation assessment method according to claim 1, characterized in that: The step of smoothing the rough segmentation boundary to obtain an optimized segmentation boundary includes: Performing morphological processing on the rough segmentation boundary to obtain the smoothed optimized segmentation boundary; The morphological processing includes at least one of the following: erosion, dilation, opening operation, and closing operation.

10. The ablation assessment method according to claim 1, characterized in that: The method further includes: calculating an ablation volume according to the optimized segmentation boundary.

11. The ablation assessment method according to claim 1, characterized in that: The method also includes: evaluating the ablation effect in combination with the target area of ​​the patient before ablation.

12. The ablation assessment method according to claim 11, characterized in that: The ablation effect is evaluated in combination with the target area of ​​the patient before ablation, including: Obtain medical images of patients before ablation; Extracting features from the target area peripheral region in the medical image before ablation, and extracting features from the segmentation boundary peripheral region in the medical image after ablation, and generating a deformation correction matrix according to the extracted features; The target area morphology in the medical image before ablation is corrected according to the deformation correction matrix to obtain the simulated target area morphology in the medical image after ablation, and then the ablation effect is evaluated in combination with the segmentation boundary.

13. The ablation assessment method according to claim 1, characterized in that: The method also includes: The segmentation boundary is compared with the estimated boundary calculated by using the ablation prediction model during the ablation process, and the parameters in the ablation prediction model are optimized.

14. The ablation assessment method according to claim 13, characterized in that: The method also includes: The optimized parameters are associated with the patient's pathological information and stored so that in the subsequent ablation process, the corresponding parameters can be determined according to the specific patient's pathological information, which is used to more accurately calculate the estimated boundary during the ablation process.

15. A method for optimizing an ablation prediction model, characterized in that: include: Acquire the patient's ablation process data; Obtaining medical images collected after ablation of the patient, and extracting segmentation boundaries; Inputting the patient ablation process data into an ablation prediction model, calculating an estimated boundary, and optimizing the parameters of the ablation prediction model so that the estimated boundary approaches the segmentation boundary; The parameters of the ablation prediction model when the predicted boundary is closest to the segmentation boundary are determined as optimized parameters, and are stored in association with the pathological information of the patient.

16. A method for generating an ablation prediction model, characterized in that: include: According to the pathological information of the current patient, the target pathological information with the highest similarity is matched from the database; wherein the patient pathological information and the parameters of the corresponding ablation prediction model are stored in association in the database, and the parameters of each ablation prediction model are obtained based on the ablation process data of the corresponding patient and the post-ablation segmentation boundary optimization; Acquire target parameters of the ablation prediction model corresponding to the target pathology information from the database; A target ablation prediction model is generated according to the target parameters, for performing ablation prediction on the current patient.

17. A method for generating an ablation prediction model, characterized in that: include: According to the pathological information of the current patient, a number of target pathological information with high similarity is matched from a database; wherein the patient pathological information and the parameters of the corresponding ablation prediction model are stored in association in the database, and the parameters of each ablation prediction model are obtained based on the ablation process data of the corresponding patient and the post-ablation segmentation boundary optimization; According to the target pathological information and the corresponding ablation prediction model parameters, a parameter generation model is used to generate target parameters corresponding to the current patient pathological information; A target ablation prediction model is generated according to the target parameters, for performing ablation prediction on the current patient.

18. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, it implements the steps of the ablation evaluation method as described in any one of claims 1 to 14, or implements the steps of the ablation prediction model optimization method as described in claim 15, or implements the steps of the ablation prediction model generation method as described in claim 16, or implements the steps of the ablation prediction model generation method as described in claim 17.

19. A laser ablation system, characterized in that: include: The electronic device of claim 18, and the laser module and the display module.

Images