A method and system for identifying and adaptively ablating tissue heterogeneity based on intraoperative iceball morphology inversion

Abstract

The application discloses a kind of based on intraoperative ice ball morphology inversion's tissue heterogeneity identification and adaptive ablation method and system.The method obtains time sequence medical image data and freezing process parameters including ice ball region in the process of cryoablation, extracts ice ball space-time evolution characteristics, identifies high perfusion influence area, heat conduction abnormal area and potential undercooling risk area in target tissue, and generates adaptive ablation strategy including lengthening freezing time, directional compensation freezing, needle supplement suggestion based on identification result, and then executes and dynamically adjusts.The application constructs the closed-loop control link of "perception-decision-execution-feedback", improves the sufficiency and safety of cryoablation by improving intraoperative ice ball from geometric coverage display to the basis of adaptive ablation decision, and realizes differentiated and intelligent ablation for different tissue heterogeneity states.

Description

Technical Field

[0001] This invention relates to the fields of tumor cryoablation, combined cryoablation, medical image processing, and intelligent control of medical devices, and particularly to a method and system for tissue heterogeneity identification and adaptive ablation based on intraoperative ice puck morphology inversion. Background Technology

[0002] During cryoablation or combined cryoablation, the ice ball formed after the ablation probe acts on the target tissue can be observed in CT, MRI, or ultrasound images. Current clinical practice typically judges the adequacy of freezing based on whether the ice ball geometrically covers the lesion, and combines this with the operator's experience to decide whether to continue freezing, end freezing, switch to subsequent treatments, or need to perform a follow-up ablation.

[0003] However, existing technologies have at least the following problems: Current ablation decision-making methods primarily rely on the geometric coverage of the ice puck, failing to translate the spatiotemporal evolution information during the ice puck's formation into a basis for adjusting ablation parameters. The dynamic information exhibited by the ice puck during its formation, such as differences in directional expansion, boundary irregularities, and changes in expansion speed, contains rich information about tissue heterogeneity, but current technologies have failed to establish a mapping relationship between this information and specific ablation actions.

[0004] Local hyperperfusion, abnormal heat conduction, tissue structural heterogeneity, and vascular influence within the target tissue can significantly alter the ice ball expansion process, further affecting freezing adequacy and subsequent treatment efficacy. However, current technologies often lack mechanisms for differentiated ablation adjustments based on these heterogeneous states. For example, in liver tumor regions adjacent to large blood vessels, ice ball expansion is hindered by the vascular heat sink effect; if the same freezing parameters as surrounding tissues are used, this region is highly prone to insufficient freezing. Conversely, in fibrotic regions, reduced thermal conductivity leads to sluggish ice ball boundary advancement; blindly extending the freezing time may damage adjacent normal tissues.

[0005] Even if imaging reveals that the ice ball has spatially covered the lesion, there may still be issues such as insufficient local cooling, residual blood vessels, or inadequate directional freezing. Current technology struggles to address these potential risks intraoperatively. Surgeons often only discover incomplete ablation after surgery through imaging follow-up or pathological examination, missing the optimal opportunity for intraoperative remediation.

[0006] Current cryoablation or combined cryoablation procedures mostly use standardized preset parameters that cannot be dynamically updated based on real-time tissue heterogeneity observed during the procedure. This makes it difficult to balance ablation adequacy and safety in different tissue states. Intraoperative adjustments rely heavily on the physician's personal experience, lacking objective quantitative evidence and automated execution methods.

[0007] Therefore, there is an urgent need for a technical solution that can generate and dynamically adjust ablation strategies based on the tissue heterogeneity identification results of intraoperative hockey puck morphology inversion, so as to achieve adaptive ablation for different tissue heterogeneity states. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a method and system for tissue heterogeneity identification and adaptive ablation based on intraoperative hockey puck morphology inversion. This method identifies tissue heterogeneity states by extracting the spatiotemporal evolution characteristics of the hockey puck and combining them with freezing process parameters. Based on this, a corresponding adaptive ablation strategy is generated and executed, enabling differentiated processing of the ablation process for areas with high perfusion influence, abnormal heat conduction, and potential undercooling risk areas, forming a closed-loop control of "perception-decision-execution-feedback".

[0009] The core concept of this invention is to regard the intraoperative ice puck as a "spatiotemporal response signal" generated by the target tissue under controlled freezing stimulation, to identify the heterogeneous state of the tissue by analyzing the signal, and to further map the identification results into executable differentiated ablation actions, thereby constructing a closed-loop control link from "ice puck morphology observation" to "ablation parameter adjustment".

[0010] Specifically, under controlled excitation of known freezing process parameters (refrigerant flow rate, probe temperature, pipeline pressure, etc.), local high perfusion, low thermal conductivity, or structural heterogeneity of the target tissue will leave identifiable information in the spatiotemporal evolution characteristics of the ice hockey puck: high perfusion zones are characterized by a continuously smaller expansion radius in a specific direction and enhanced boundary fluctuations; abnormal heat conduction zones are characterized by increased boundary irregularity and nonlinear decay of expansion rate; and potentially undercooled zones are characterized by nominal coverage meeting standards but local dynamic anomalies. This invention first decodes this information into discrete tissue heterogeneity state classifications through multi-dimensional feature extraction and model inversion. Then, based on preset heterogeneity state-ablation action mapping rules or optimization models, it automatically generates and executes targeted adaptive ablation strategies, and continuously acquires new ice hockey puck images during strategy execution for effect evaluation and strategy correction.

[0011] This concept elevates cryoablation from an open-loop, experience-based operation to a closed-loop, adaptive control, enabling real-time response and precise compensation for tissue heterogeneity without requiring additional invasive testing.

[0012] The technical solution of the present invention is as follows: A method for tissue heterogeneity identification and adaptive ablation based on intraoperative hockey puck morphology inversion, comprising the following steps: S1. During the cryoablation of the target tissue, acquire time-series medical image data containing the ice ball region, and acquire the cryoablation process parameters corresponding to the cryoablation process; S2. Extract the spatiotemporal evolution features of ice hockey based on the time-series medical image data; S3. Based on the spatiotemporal evolution characteristics of the ice hockey puck and the freezing process parameters, identify the tissue heterogeneity state in the target tissue. The tissue heterogeneity state includes at least one of the following: high perfusion influence zone, abnormal heat conduction zone, and potential undercooling risk zone. S4. Generate an adaptive ablation strategy based on the identified tissue heterogeneity state, wherein the adaptive ablation strategy includes at least one of extending the freezing duration, implementing targeted compensatory freezing, and outputting supplementary needle suggestions; S5. Perform ablation according to the adaptive ablation strategy, and / or adjust the adaptive ablation strategy according to the updated spatiotemporal evolution characteristics of ice hockey.

[0013] In some embodiments, the freezing process parameters include at least one of refrigerant flow rate, freezing duration, probe temperature, and pipeline pressure; the spatiotemporal evolution characteristics of the ice puck include at least two of the following: ice puck boundary characteristics, directional expansion characteristics, and expansion dynamics characteristics; the ice puck boundary characteristics include at least one of the following: ice puck boundary position, ice puck volume, ice puck surface area, ice puck major-minor axis ratio, sphericity, boundary curvature, boundary irregularity, and boundary fractal dimension; the directional expansion characteristics include at least one of the following: ice puck expansion radius in different angular directions, expansion direction offset, symmetry index, and expansion difference in each direction; the expansion dynamics characteristics include at least one of the following: ice puck expansion velocity, expansion acceleration, expansion velocity change rate, time required to reach a preset radius, steady-state maintenance fluctuation characteristics, and retraction velocity.

[0014] In some embodiments, step S3 identifies the tissue heterogeneity state using a tissue heterogeneity identification model. This model establishes a correspondence between the spatiotemporal evolution characteristics of the ice hockey puck and the local perfusion influence state, local heat conduction state, and undercooling risk state of the target tissue. The tissue heterogeneity identification model is an inversion model based on biological heat transfer constraints, a machine learning model based on physical constraints, a finite element inverse problem solving model, an ensemble Kalman filter model, or a combination thereof. When identifying high perfusion influence zones in step S3, identification is based on at least one of the following characteristics: the ice hockey puck expansion radius is consistently smaller in a certain direction, the expansion speed is consistently lower than in other directions, and boundary fluctuations are enhanced. When identifying abnormal heat conduction zones, identification is based on at least one of the following characteristics: increased ice hockey puck boundary irregularity, abnormally concentrated boundary curvature, abnormally low expansion speed, and delayed expansion. When identifying potential undercooling risk zones, identification is based on at least one of the following characteristics: the ice hockey puck geometric coverage has met preset requirements but local expansion dynamics are abnormal, boundary complexity is abnormal, or directional expansion is obstructed.

[0015] In some embodiments, in step S4, the adaptive ablation strategy generated for the identified high perfusion impact area includes at least one of the following: extending the equivalent freezing duration of the corresponding region or direction, increasing the compensating freezing intensity of the corresponding region or direction, and delaying the freezing termination timing; the adaptive ablation strategy generated for the identified abnormal heat conduction area includes at least one of the following: extending the low temperature maintenance duration, adjusting the freezing termination threshold, and changing the timing of the cold-hot switching; the adaptive ablation strategy generated for the identified potential undercooling risk area includes at least one of the following: continuing freezing, implementing local compensating freezing, outputting a needle replacement suggestion, and adjusting the multi-needle synergistic timing; the adaptive ablation strategy generated in step S4 also includes a directional compensation strategy for the vascular impact risk direction, which includes at least one of the following: implementing compensating freezing for the vascular impact risk direction, adjusting the probe position, and outputting a needle redeployment suggestion.

[0016] In some embodiments, step S5 adjusts the adaptive ablation strategy based on the updated spatiotemporal evolution characteristics of the ice hockey puck, including re-identifying tissue heterogeneity and updating the freezing duration, compensating for freezing intensity, providing needle replacement suggestions, adjusting the timing of cold and hot switching, or adjusting the timing of multi-needle synergy. The time-series medical imaging data is CT image data, MRI image data, ultrasound image data, or a combination thereof, preferably CT image data. The needle replacement suggestions output in step S4 include at least one of needle replacement location, needle replacement direction, and needle replacement priority. The adjustment in step S5 is an online adjustment within a single freezing process or an iterative adjustment between multiple freezing cycles.

[0017] Another aspect of this invention discloses a tissue heterogeneity identification and adaptive ablation system based on intraoperative hockey puck morphology inversion, comprising: The data acquisition module is used to acquire time-series medical image data including the ice puck area and the corresponding freezing process parameters during the cryoablation of the target tissue; The feature extraction module is used to extract spatiotemporal evolution features of ice hockey based on the time-series medical image data; The heterogeneity identification module is used to identify the tissue heterogeneity state in the target tissue based on the spatiotemporal evolution characteristics of the ice hockey puck and the freezing process parameters. The tissue heterogeneity state includes at least one of the following: high perfusion influence zone, abnormal heat conduction zone, and potential undercooling risk zone. The strategy generation module is used to generate an adaptive ablation strategy based on the identified tissue heterogeneity state. The adaptive ablation strategy includes at least one of extending the freezing duration, implementing targeted compensatory freezing, and outputting supplementary needle suggestions. The execution and adjustment module is used to execute ablation according to the adaptive ablation strategy and / or adjust the adaptive ablation strategy according to the updated spatiotemporal evolution characteristics of hockey.

[0018] In some embodiments, the strategy generation module is further used to generate at least one of the following: a directional compensation strategy for the direction of vascular impact risk, a timing adjustment strategy for hot / cold switching, and a multi-needle synergistic timing adjustment strategy; the heterogeneity identification module integrates a tissue heterogeneity identification model, which is an inversion model based on biological heat transfer constraints, a machine learning model based on physical constraints, a finite element inverse problem solving model, an ensemble Kalman filter model, or a combination thereof; the execution and adjustment module is connected to the refrigerant supply and freezing control unit through a communication interface to adjust the refrigerant flow rate, probe temperature, freezing duration, or pipeline pressure in real time.

[0019] Compared with the prior art, the present invention has at least the following beneficial effects: 1. This invention elevates the intraoperative ice puck from a geometrically overlaid display object to a decision-making basis for adaptive ablation strategies, realizing a closed-loop transformation from "seeing the ice puck" to "using the ice puck to guide actions"; 2. This invention can identify high-perfusion-affected areas, abnormal heat conduction areas, and potential undercooling risk areas based on the spatiotemporal evolution characteristics of ice hockey, and generate differentiated ablation strategies for different heterogeneous states, breaking through the traditional "one-size-fits-all" ablation parameter setting mode; 3. This invention can provide compensatory treatment for problems such as local insufficient cooling, obstruction near blood vessels, and insufficient directional freezing. It improves the adequacy of ablation through directional compensatory freezing, supplementary needle suggestions, and multi-needle synergistic timing adjustment. 4. This invention supports online adjustments within a single freezing process and iterative adjustments between multiple freezing cycles, enabling the ablation strategy to evolve in real time with the evolution of the ice puck, thereby improving the flexibility and safety of the intraoperative strategy. 5. This invention can improve the individualization and intelligence of cryoablation or combined cryoablation without adding additional invasive detection, and can be directly embedded into existing intraoperative image navigation workflows. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the overall process of the method of the present invention; Figure 2 This is a schematic diagram of the ice hockey time-series medical image acquisition and feature extraction process in this invention; Figure 3 This is a schematic diagram of the tissue heterogeneity state identification process in this invention; Figure 4 This is a schematic diagram of the mapping relationship between "heterogeneous state-adaptive ablation strategy" in this invention; Figure 5 This is a schematic diagram of the adaptive ablation strategy execution and dynamic adjustment process in this invention; Figure 6 This is a block diagram of the system structure of the present invention. Detailed Implementation

[0021] The present invention will be further described in detail below with reference to specific embodiments. It should be understood that the following embodiments are only used to explain the present invention and are not intended to limit the scope of protection of the present invention. Example 1

[0022] like Figure 1 As shown, this embodiment provides a method for intraoperative tissue heterogeneity identification and adaptive cryocompensation based on CT images, forming a complete closed loop from image acquisition, feature extraction, heterogeneity identification, strategy generation to strategy execution and feedback adjustment.

[0023] 1. System Hardware Architecture The system includes at least: Cryoablation probes (single or multi-needle configuration); The refrigerant supply and refrigeration control unit has functions such as refrigerant flow regulation, probe temperature control, and pipeline pressure monitoring, and reserves external control interfaces (such as RS-232, Ethernet or DICOM interface). The CT image acquisition unit is equipped with rapid intraoperative scanning capability (preferably multi-slice spiral CT, slice thickness ≤2.5 mm, reconstruction interval ≤1.25 mm). The process parameter acquisition unit is used to acquire at least one of the following in real time: refrigerant flow rate, freezing duration, probe temperature, and pipeline pressure, with a sampling frequency of not less than 1 Hz; The data processing, heterogeneity identification, and strategy generation unit can be a standalone workstation or a computing module integrated into the imaging equipment, and it has GPU-accelerated computing capabilities to meet real-time requirements.

[0024] 2. Image Acquisition and Preprocessing After cryoablation begins, the CT image acquisition unit acquires images of the ice hockey puck area at preset time intervals, resulting in a time-series CT image sequence. For example... Figure 2 As shown, the time interval can be any value from 0.5 seconds to 30 seconds, preferably any value from 1 second to 10 seconds.

[0025] In the preferred scheme, the data acquisition interval is dynamically adjusted based on the puck's expansion rate: the total volume change rate Rv(t) is defined as Rv(t) = [V(t) - V(t-Δt)] / Δt. When Rv(t) > 5 cm³ / min, the acquisition interval is shortened to 1-3 seconds; when Rv(t) < 1 cm³ / min, the acquisition interval is extended to 10-30 seconds. This dynamic adjustment mechanism ensures information density during critical phases while reducing radiation exposure and equipment occupancy during non-critical phases.

[0026] The acquired raw CT data underwent the following preprocessing: Image registration: Based on rigid registration algorithms (such as mutual information registration), CT images at different times are aligned to the same coordinate system to eliminate displacement caused by the patient's breathing and slight changes in body position; Noise suppression: Anisotropic diffusion filtering or a denoising algorithm based on nonlocal means is used to suppress CT noise while preserving the sharpness of the hockey puck boundary; Region of Interest (ROI) clipping: Automatically clips the local area containing the puck and the surrounding safety boundary, centered on the probe position, reducing subsequent computation.

[0027] 3. Ice hockey puck segmentation and boundary extraction The data processing unit performs hockey puck segmentation on the preprocessed time-series CT images. The hockey puck appears as a low-density region in the CT images (CT value approximately -20 HU to -50 HU, significantly lower than the surrounding soft tissue). This implementation provides three feasible segmentation schemes: Solution A: Segmentation based on adaptive threshold The initial CT value threshold range was set to -60 HU to 0 HU. The optimal segmentation threshold T_otsu was automatically calculated within the ROI using the Otsu adaptive thresholding method. Morphological opening operations (kernel radius 1-2 pixels) were performed on the binarized result to remove noise, followed by morphological closing operations (kernel radius 2-3 pixels) to fill the small holes at the boundary, resulting in the hockey puck mask M_ice. The outer contour of M_ice was extracted as the hockey puck boundary Γ(t).

[0028] Option B: Segmentation based on region growing Using the probe tip coordinates P_tip as the seed point, a gray-level similarity threshold ΔH = 15 HU was set, and the morphological constraint was a maximum growth radius R_max = 50 mm (to prevent excessive growth outside the body). A 26-neighborhood 3D region growing algorithm was employed to iteratively merge voxels satisfying |H(x) - H_seed| < ΔH and ||x - P_tip|| < R_max, starting from the seed point, to obtain the ice puck region.

[0029] Solution C: Deep Learning-Based Segmentation The network employs a 3D U-Net or nnU-Net architecture. The input is CT volumetric data cropped from the ROI (voxel size uniformly resampled to 1 mm × 1 mm × 1 mm), and the output is a hockey puck probability map of the same size as the input. Training is performed using historical intraoperative CT images and expert-annotated hockey puck masks. The training loss function is a weighted sum of Dice Loss and cross-entropy loss. During inference, a binary mask is obtained by setting a threshold of 0.5 on the probability map.

[0030] After segmentation, the ice hockey puck boundary is smoothed (e.g., based on B-spline curve fitting or moving least squares method) and topologically corrected (removing isolated small regions and filling internal holes) to ensure the continuity and simple connectivity of the boundary Γ(t).

[0031] 4. Extraction of spatiotemporal evolution features of ice hockey like Figure 2 As shown, based on the extracted multi-time hockey puck boundary Γ(t), the following three types of features are calculated: (1) Characteristics of ice hockey boundary Ice puck volume V(t): summation of voxel volumes over the mask M_ice; Ice hockey puck surface area A(t): The total area of ​​the triangular facets is calculated after extracting the isosurfaces from the mask using the Marching Cubes algorithm; Major-minor axis ratio of the ice hockey puck: The ratio of the longest principal axis length L_max to the shortest principal axis length L_min is the major-minor axis ratio of the ice hockey puck mask in three dimensions. Sphericity ψ(t) = π^(1 / 3) · (6V(t))^(2 / 3) / A(t), with a value range of (0,1], the closer to 1, the closer to a sphere; Boundary irregularity: Calculate the ratio of the perimeter (or surface area) of the boundary to the perimeter (or surface area) of a standard circle (or sphere) of the same area (or volume); Boundary fractal dimension: Using box counting, the boundary is covered with different scales ε, and the linear slopes of log(N(ε)) and log(1 / ε) are fitted as the fractal dimension D_f.

[0032] (2) Directional expansion characteristics A three-dimensional polar coordinate system is established with the probe position P_tip as the origin. The azimuth angle θ∈[0,2π) and the zenith angle φ∈[0,π] are divided into N_θ and N_φ sectors respectively (preferably N_θ=16, N_φ=8). For each time t, the radius of ice ball expansion in each sector is calculated: R(θ_i, φ_j, t) = max{||x - P_tip|| : x ∈ Γ(t), x lies within sector (θ_i, φ_j)} Extended direction offset: d(t) = C(t) - P_tip, where C(t) is the coordinate of the hockey puck's center of mass, the offset direction is d(t) / |d(t)|, and the offset amount is |d(t)|; Symmetry index: For a selected axis of symmetry (such as the x-axis), S_x(t) = 1 - Σ|R(θ_i,φ_j,t) - R(π-θ_i,φ_j,t)| / ΣR(θ_i,φ_j,t); Expansion difference in each direction: D(t) = σ[R(θ_i,φ_j,t)] / μ[R(θ_i,φ_j,t)].

[0033] (3) Extended dynamic characteristics The velocity of the puck's spread is: v(θ_i,φ_j,t) = [R(θ_i,φ_j,t) - R(θ_i,φ_j,t-Δt)] / Δt; Extended acceleration: a(θ_i,φ_j,t) = [v(θ_i,φ_j,t) - v(θ_i,φ_j,t-Δt)] / Δt; Time required to reach the preset radius: t_0(θ_i,φ_j) = min{t : R(θ_i,φ_j,t) ≥ R_preset}; Steady-state maintenance fluctuation characteristics: During the freezing maintenance phase (when the global volume change rate |dV / dt| < 0.5 cm³ / min), the coefficient of variation of the radius in each direction CV = σ[R] / μ[R], and the dominant frequency component f_dom extracted by fast Fourier transform (FFT) are calculated; Retraction speed: After stopping cooling or switching to heating, v_retract(θ_i,φ_j) = -[R(t) -R(t-Δt)] / Δt.

[0034] 5. Identification of organizational heterogeneity like Figure 3 As shown, the extracted spatiotemporal evolution features of the ice hockey puck and the freezing process parameters are input into the tissue heterogeneity identification model to identify the heterogeneous state in the target tissue.

[0035] 5.1 Identification of High Irrigation Influence Zones The hyperperfusion-affected area is caused by the heat sink effect of blood flow, and its hockey "fingerprint" appearance is as follows: Judgment conditions (at least two conditions must be met): There exist at least three consecutive time points where the radius of expansion in a certain direction, R(θ,φ,t), is less than 60% of the global average radius at that time point. The directional expansion velocity v(θ,φ,t) remains below 50% of the global average velocity for at least 2 minutes. In the steady state phase, the coefficient of variation (CV) of the boundary in this direction is > 0.15, or the dominant frequency component f_dom is highly correlated with the heart rate (1-1.5Hz) (correlation coefficient > 0.6).

[0036] Identification output: Mark the sector direction or corresponding spatial region as a high perfusion influence area, and quantify the perfusion influence intensity I_perfusion = 1 - R_local / R_global.

[0037] 5.2 Identification of Anomalous Heat Conduction Zones Abnormal heat conduction areas are caused by localized abnormalities in tissue thermal conductivity (such as fibrosis, necrosis, and fatty infiltration), and their hockey "fingerprint" appearance is as follows: Judgment conditions (at least two conditions must be met): A local boundary fractal dimension D_f > 1.25 (two-dimensional section) or D_f > 2.3 (three-dimensional surface) indicates an irregular boundary height. The number of anomaly concentration points with an absolute value |κ| > 0.5 mm⁻¹ of local boundary curvature κ exceeds 20% of the total number of boundary points; The expansion rate exhibits a nonlinear decay: a(θ,φ,t) < -0.5 mm / min² and persists for more than 1 minute; Extended propulsion lag: t_0(θ,φ) > μ[t_0] + 2σ[t_0], meaning that the time to reach the preset radius is significantly later than the average level.

[0038] Identification output: Mark the corresponding spatial region as a heat conduction anomaly zone and quantify the heat conduction anomaly index I_conduction = |D_f - D_f,normal| / D_f,normal.

[0039] 5.3 Identification of Potential Low-Cooling Risk Zones Potential undercooling risk areas refer to areas that are nominally covered by hockey but may not actually reach effective freezing temperatures (such as below -40°C). Their hockey "fingerprint" characteristics are as follows: Judgment conditions (at least one condition must be met): The total volume of the ice hockey puck has reached the preset coverage volume (such as 1.5 times the volume of the lesion), but the expansion rate of a certain local area suddenly drops to below 0.2 mm / min when it approaches the target boundary; The local boundary complexity (measured by fractal dimension or irregularity) is abnormally high, but it does not conform to the typical spatial distribution pattern of the heat conduction anomaly zone; The directional expansion stalls or rebounds before reaching the preset radius (R(t+Δt) < R(t) and does not enter the retraction phase).

[0040] Identification output: Mark the corresponding area as a potential undercooling risk area and calculate the undercooling risk index I_cold = 1 - v_local / v_threshold, where v_threshold is the empirical threshold of the minimum expansion rate required to achieve effective freezing.

[0041] 6. Adaptive Ablation Strategy Generation like Figure 4As shown, the strategy generation module generates adaptive ablation strategies based on the identified heterogeneous states, using predefined mapping rules or optimization models.

[0042] 6.1 Strategies for High Irrigation Influence Zones For regions where the perfusion effect intensity I_perfusion > 0.4: Extend the equivalent freezing duration: Δt_add = k1 × I_perfusion × t_base, where k1 is an empirical coefficient (preferably 0.5-1.5), and t_base is the current freezing time. For example, if freezing has been in progress for 10 minutes, I_perfusion = 0.6, and k1 = 1.0, then it is recommended to extend the duration by 6 minutes. Increase the compensating freezing intensity: Increase the corresponding probe (multi-needle scenario) or the overall refrigerant flow by 20%-50%, or further reduce the probe target temperature from -150℃ to -180℃ (if the equipment supports it). Postpone the termination of cryotherapy: Modify the original termination conditions (such as the ice ball covering the lesion periphery by 5 mm) to cover the periphery by 10 mm, or require an additional 3-5 mm increase in the radius of extension in the high-risk direction.

[0043] 6.2 Strategy for Heat Conduction Anomalies For regions where the thermal conductivity anomaly index I_conduction > 0.3: Extend the duration of cryogenic maintenance: After the ice ball reaches the target volume, maintain the cryogenic temperature for an additional period of Δt_maintain = k2×I_conduction × t_base, where k2 is preferably 0.3-0.8; Adjust the freezing end threshold: appropriately lower the temperature or volume threshold for freezing end to ensure that the effective freezing temperature is reached inside the abnormal heat conduction zone; Change the timing of switching between hot and cold ablation: In combined hot and cold ablation, if the area with abnormal heat conduction is located at the edge of the lesion, the timing of switching to heating can be delayed to ensure that the abnormal area is fully frozen.

[0044] 6.3 Strategies for Potential Cold Weather Risk Zones For regions with a cold risk index I_cold > 0.5: Continue freezing: Extend the global or local freezing time until the expansion dynamics of the risk area return to normal or the maximum safe time limit is reached; Implement localized compensation freezing: In a multi-needle system, activate the auxiliary probe adjacent to the risk area for coordinated freezing, or increase the refrigerant distribution ratio of the main probe in the corresponding direction; Output replacement needle suggestion: If I_cold is still higher than 0.3 after cryopreservation compensation, a replacement needle suggestion will be generated, including: Replacement needle location: The geometric centroid or the point of maximum temperature field gradient in the potential undercooling risk zone; Needle insertion direction: The direction vector from the current probe position to the center of gravity of the risk area, or the optimal needle insertion angle determined based on temperature field simulation; Needle replacement priority: sorted according to I_cold value, with the highest I_cold value having the highest priority (P1), and so on down.

[0045] 6.4 Directional Compensation Strategies for Vascular Impact Risks If a certain direction is identified as a direction of risk for vascular impact (an extreme manifestation of a high-perfusion-affected area): Targeted compensatory freezing: In multi-needle systems, a compensatory probe is added in the direction of risk, or the refrigerant distribution valve of the existing probe is adjusted to provide an additional 30%-70% refrigerant flow in the direction of risk. Adjust probe position: Calculate the distance and orientation of the current probe position to the nearest large blood vessel. It is recommended to move the probe 2-5 mm away from the blood vessel, or adjust the probe angle to bring the needle tip closer to the risk area. Recommendations for re-needling: If the vessel diameter is >3 mm and the distance from the lesion is <10 mm, it is recommended to use the "double needle clamping" needle placement method, that is, to place needles on both sides of the vessel to form a counter-frogging cryotherapy.

[0046] 7. Strategy Execution and Dynamic Adjustment like Figure 5 As shown, the execution and adjustment module sends the generated adaptive ablation strategy to the refrigerant supply and refrigeration control unit for execution via the communication interface.

[0047] 7.1 Online adjustment (within a single freezing process) During strategy execution, the CT image acquisition unit continuously acquires new hockey images, and the feature extraction module updates the spatiotemporal evolution characteristics of the hockey team in real time. Strategy re-evaluation is initiated when the newly acquired data meets the following triggering conditions: Triggering condition 1: Within 3-5 minutes after the compensation freeze is executed, the expansion speed in the target direction does not increase by more than 20%; Triggering condition 2: The newly identified heterogeneous state does not match the current execution strategy (e.g., the original strategy is for a high-perfusion area, but new data shows that the area has turned to normal expansion, while another area shows a new heat conduction anomaly). Triggering condition 3: The total volume of the ice puck or the radius in a certain direction unexpectedly shrinks (shrinkage speed > 1 mm / min and not during the planned heating phase).

[0048] The adjustment actions include: Recalculate the perfusion effect intensity I_perfusion and the thermal conduction anomaly index I_conduction; If I_perfusion decreases by more than 30%, gradually reduce the compensation intensity to avoid over-freezing; If a new low-cooling risk zone emerges, it will be added to the strategy queue and executed according to priority. If two consecutive adjustments are ineffective, the priority of the supplementary injection recommendation will be increased or a prompt for manual intervention will be issued.

[0049] 7.2 Iterative Adjustment (between multiple freeze cycles) For treatment regimens employing multiple cryo-warming cycles: After the first cycle is completed, the results of tissue heterogeneity identification and the actual ablation response for that cycle are saved. Before the second cycle begins, the strategy mapping parameters are corrected based on the data from the first cycle (such as adjusting the coefficients of k1 and k2). If a certain direction is a high-perfusion-affected area in the first round, stronger compensation will be prioritized for that direction in the second round, and the image acquisition interval will be shortened for intensive monitoring. If the risk area is eliminated after the first round of injections, the corresponding injection recommendation will be cancelled in the second round; if it is not eliminated, the injection depth will be increased or the angle will be adjusted. Example 2

[0050] The tissue heterogeneity identification model in step S3 can be implemented in the following specific ways: Employing a physical constraint-based deep learning model Construct a deep neural network (such as 3D ResNet or Transformer architecture), with the input being a time-series hockey puck feature tensor (containing multi-channel volume data of boundary, directionality, and dynamic features) and a freezing process parameter vector, and the output being a probability map of three heterogeneous states (probability P_p of high perfusion influence zone, probability P_c of heat conduction anomaly zone, and probability P_r of potential undercooling risk zone).

[0051] Introduce physical constraint terms into the loss function: L_total = L_dice + λ1L_ce + λ2L_physics Where L_dice is the Dice loss, L_ce is the cross-entropy loss, and L_physics is the physical residual loss, penalizing predictions that violate heat transfer laws (e.g., predicted high-perfusion areas should have lower local expansion rates). The training data uses intraoperative CT sequences from historical cases and corresponding postoperative pathological perfusion assessments (CD31 staining) and elastography results as the gold standard. Example 3

[0052] In multi-needle cryoablation scenarios, the strategy generation module needs to coordinate the refrigerant distribution and timing control of multiple probes.

[0053] 1. Multi-needle coordinated timing adjustment Assuming the system is equipped with N probes (N≥2), and the refrigerant flow rate of each probe can be adjusted independently: When a high-injection-affected area or a potential under-cooling risk area is identified: Calculate the distance d_i from each probe to the target area and the directional angle θ_i; Define the influence weight w_i = cos(θ_i) / (d_i + ε), where ε is a small constant to prevent division by zero; Refrigerant flow is redistributed according to weights: Qi_i_new = Qi_i_base + ΔQ_total × w_i / Σw_j, where ΔQ_total is the additional total flow that the system can provide; If the weight of a probe is too low (w_i < 0.1×max(w)) and there is no significant heterogeneity around the probe, its flow rate can be temporarily reduced to save refrigerant and reduce damage to normal tissue.

[0054] 2. Calculation of the position of the replacement stitch For potential areas at risk of undercooling that require re-needling, the re-needling location P_new is determined through the following steps: Calculate the geometric centroid of the risk region G = (1 / N_voxel) Σx_i, where x_i is the voxel coordinate within the risk region; Centered on G, search for the optimal needle insertion point within a radius r = 5-10 mm. Constraints include: avoiding large blood vessels (distance > 5 mm), avoiding bones, and ensuring no important organs are on the path. If multiple candidate points exist, select the point where the line connecting P_new and the nearest existing probe is perpendicular to the long axis of the blood vessel to maximize the hedging effect of cryotherapy.

[0055] 3. Calculation of stitch direction The direction of the replacement stitch, v_new, is determined by a combination of the following factors: Main direction: The direction from the skin entry point to P_new; Correction: Based on the direction of the temperature field gradient ▽T at P_new, the cold source at the needle tip is oriented towards the residual warm area; Safety constraints: The angle between the probe and the adjacent probe should be >15° to avoid contact with the probe body. Example 4

[0056] like Figure 6 As shown, the present invention also provides a tissue heterogeneity identification and adaptive ablation system based on intraoperative hockey puck morphology inversion.

[0057] The system includes: The data acquisition module includes at least one of a CT acquisition unit, an MRI acquisition unit, and an ultrasound acquisition unit, as well as a process parameter acquisition unit. The process parameter acquisition unit collects probe temperature, refrigerant flow rate, freezing duration, and pipeline pressure in real time via sensors. The sampled data is converted from analog to digital and then uploaded to the data processing unit via TCP / IP or serial communication.

[0058] Feature extraction module: Receives time-series medical image data, performs image registration, denoising, ROI cropping, hockey puck segmentation, and boundary extraction, and calculates boundary features, directional expansion features, and expansion dynamics features. This module can be accelerated using GPU parallel processing, with feature extraction time for a single frame CT image (512×512×100 voxels) controlled within 3-5 seconds.

[0059] Heterogeneity identification module: This module integrates an organized heterogeneity identification model, mapping feature vectors to heterogeneous state classifications. It supports multi-model switching, allowing users to choose between a physical inversion model (high accuracy, time-consuming, suitable for offline fine-grained assessment) or a machine learning model (fast speed, suitable for online real-time identification) depending on the clinical scenario.

[0060] The strategy generation module, based on a predefined rule base or optimization algorithm, transforms heterogeneous states into specific ablation action instructions. The rule base is stored in an extensible XML or JSON format, facilitating updates to the mapping relationships by clinical experts based on evidence-based medicine.

[0061] The execution and adjustment module connects to the refrigerant supply and refrigeration control unit via a standardized communication protocol (such as HL7 FHIR or a custom binary protocol) and issues commands for flow adjustment, temperature setting, and time setting. This module also receives feedback signals from the control unit (actual probe temperature, actual flow rate) for closed-loop verification. If the deviation between the actual parameters and the target parameters exceeds a safety threshold (e.g., temperature deviation > 10℃), an alarm is triggered and automatic adjustment is paused, awaiting manual confirmation.

[0062] The data flow between the modules is as follows: Data acquisition module - feature extraction module: raw image data (DICOM format) and process parameters (JSON format); Feature extraction module - heterogeneity identification module: feature vector (HDF5 or custom binary format); Heterogeneity identification module - policy generation module: heterogeneous state labels and confidence levels; Strategy Generation Module - Execution and Adjustment Module: Ablation Strategy Instruction Set; Execution and Adjustment Module - Data Acquisition Module: Triggers the synchronization signal for the next cycle of image acquisition.

[0063] It should be noted that the term "tissue heterogeneity state" in this specification refers to a state that has a differential effect on the ablation response of the target tissue, including high perfusion influence state, abnormal heat conduction state, potential undercooling risk state, and combinations thereof.

[0064] It should be noted that the "high perfusion influence zone" referred to in this instruction manual refers to the area or direction in which local blood flow significantly inhibits or disturbs the expansion of the ice ball.

[0065] It should be noted that the "abnormal heat conduction zone" referred to in this instruction manual refers to the area where the heat conduction capacity of the tissue body is significantly different from that of the surrounding area, thus causing abnormalities in the expansion process of the ice puck.

[0066] It should be noted that the “potential undercooling risk zone” referred to in this manual refers to an area that, although it meets the nominal geometric coverage requirements, may still not reach the preset effective freezing threshold based on the spatiotemporal evolution characteristics of ice hockey.

[0067] It should be noted that the "adaptive ablation strategy" referred to in this manual refers to a set of differentiated ablation actions generated based on the heterogeneity of the tissue, including adjustment of freezing duration, adjustment of compensating freezing intensity, supplementary needle recommendations, adjustment of timing for switching between hot and cold, adjustment of multi-needle synergistic timing, and combinations thereof.

[0068] It should be noted that the term "directional compensatory freezing" in this manual refers to enhanced freezing measures implemented in a specific spatial direction or local area to compensate for insufficient freezing caused by tissue heterogeneity in that direction or area.

[0069] Of course, the above are just typical examples of the present invention. In addition, the present invention may have many other specific embodiments. All technical solutions formed by equivalent substitution or equivalent transformation fall within the scope of protection claimed by the present invention.

Claims

1. A method for tissue heterogeneity identification and adaptive ablation based on intraoperative hockey puck morphology inversion, characterized in that, Includes the following steps: S1. During the cryoablation of the target tissue, acquire time-series medical image data containing the ice ball region, and acquire the cryoablation process parameters corresponding to the cryoablation process; S2. Extract the spatiotemporal evolution features of ice hockey based on the time-series medical image data; S3. Based on the spatiotemporal evolution characteristics of the ice hockey puck and the freezing process parameters, identify the tissue heterogeneity state in the target tissue. The tissue heterogeneity state includes at least one of the following: high perfusion influence zone, abnormal heat conduction zone, and potential undercooling risk zone. S4. Generate an adaptive ablation strategy based on the identified tissue heterogeneity state, wherein the adaptive ablation strategy includes at least one of extending the freezing duration, implementing targeted compensatory freezing, and outputting supplementary needle suggestions; S5. Perform ablation according to the adaptive ablation strategy, and / or adjust the adaptive ablation strategy according to the updated spatiotemporal evolution characteristics of ice hockey.

2. The tissue heterogeneity identification and adaptive ablation method according to claim 1, characterized in that, The freezing process parameters include at least one of refrigerant flow rate, freezing duration, probe temperature, and pipeline pressure; the spatiotemporal evolution characteristics of the ice puck include at least two of the following: ice puck boundary characteristics, directional expansion characteristics, and expansion dynamics characteristics; the ice puck boundary characteristics include at least one of the following: ice puck boundary position, ice puck volume, ice puck surface area, ice puck major-minor axis ratio, sphericity, boundary curvature, boundary irregularity, and boundary fractal dimension; the directional expansion characteristics include at least one of the following: ice puck expansion radius in different angular directions, expansion direction offset, symmetry index, and expansion difference in each direction; the expansion dynamics characteristics include at least one of the following: ice puck expansion velocity, expansion acceleration, expansion velocity change rate, time required to reach the preset radius, steady-state maintenance fluctuation characteristics, and retraction velocity.

3. The tissue heterogeneity identification and adaptive ablation method according to claim 1 or 2, characterized in that, In step S3, the tissue heterogeneity state is identified by a tissue heterogeneity identification model. The tissue heterogeneity identification model is used to establish the correspondence between the spatiotemporal evolution characteristics of the ice hockey puck and the local perfusion influence state, local heat conduction state, and undercooling risk state of the target tissue. The tissue heterogeneity identification model is an inversion model based on biological heat transfer constraints, a machine learning model based on physical constraints, a finite element inverse problem solving model, an ensemble Kalman filter model, or a combination thereof.

4. The tissue heterogeneity identification and adaptive ablation method according to claim 3, characterized in that, When identifying the high-perfusion-affected area in step S3, the identification is based on at least one of the following characteristics: the ice ball expansion radius in a certain direction is continuously smaller, the expansion speed is continuously lower than in other directions, and the boundary fluctuation is enhanced. When identifying abnormal heat conduction zones, identification is based on at least one of the following features: increased irregularity of ice puck boundary, abnormal concentration of boundary curvature, abnormal attenuation of expansion velocity, and sluggish expansion propulsion. When identifying potential undercooling risk areas, identification is based on at least one of the following characteristics: the puck's geometric coverage has met the preset requirements but the local expansion dynamics are abnormal, the boundary complexity is abnormal, or the directional expansion is blocked.

5. The tissue heterogeneity identification and adaptive ablation method according to claim 4, characterized in that, In step S4, the adaptive ablation strategy generated for the identified high perfusion impact area includes at least one of the following: extending the equivalent freezing duration of the corresponding area or direction, increasing the compensating freezing intensity of the corresponding area or direction, and delaying the freezing termination time; the adaptive ablation strategy generated for the identified abnormal heat conduction area includes at least one of the following: extending the low temperature maintenance duration, adjusting the freezing termination threshold, and changing the timing of the cold-hot switching; the adaptive ablation strategy generated for the identified potential undercooling risk area includes at least one of the following: continuing freezing, implementing local compensating freezing, outputting needle replacement suggestions, and adjusting the multi-needle synergistic timing; the adaptive ablation strategy generated in step S4 also includes a directional compensation strategy for the vascular impact risk direction, which includes at least one of the following: implementing compensating freezing for the vascular impact risk direction, adjusting the probe position, and outputting a re-needle placement suggestion.

6. The tissue heterogeneity identification and adaptive ablation method according to claim 5, characterized in that, In step S5, the adaptive ablation strategy is adjusted according to the updated spatiotemporal evolution characteristics of the ice hockey puck, including re-identifying the tissue heterogeneity state and updating the freezing duration, compensating for the freezing intensity, providing needle replacement suggestions, adjusting the timing of cold and hot switching, or adjusting the timing of multi-needle synergy. The time-series medical imaging data is CT image data, MRI image data, ultrasound image data, or a combination thereof, preferably CT image data. The needle replacement suggestions output in step S4 include at least one of needle replacement position, needle replacement direction, and needle replacement priority. The adjustment in step S5 is an online adjustment within a single freezing process or an iterative adjustment between multiple freezing cycles.

7. A tissue heterogeneity identification and adaptive ablation system based on intraoperative hockey puck morphology inversion, characterized in that, include: The data acquisition module is used to acquire time-series medical image data including the ice puck area and the corresponding freezing process parameters during the cryoablation of the target tissue; The feature extraction module is used to extract spatiotemporal evolution features of ice hockey based on the time-series medical image data; The heterogeneity identification module is used to identify the tissue heterogeneity state in the target tissue based on the spatiotemporal evolution characteristics of the ice hockey puck and the freezing process parameters. The tissue heterogeneity state includes at least one of the following: high perfusion influence zone, abnormal heat conduction zone, and potential undercooling risk zone. The strategy generation module is used to generate an adaptive ablation strategy based on the identified tissue heterogeneity state. The adaptive ablation strategy includes at least one of extending the freezing duration, implementing targeted compensatory freezing, and outputting supplementary needle suggestions. The execution and adjustment module is used to execute ablation according to the adaptive ablation strategy and / or adjust the adaptive ablation strategy according to the updated spatiotemporal evolution characteristics of hockey.

8. The tissue heterogeneity identification and adaptive ablation system according to claim 7, characterized in that, The strategy generation module is further used to generate at least one of the following: a directional compensation strategy for the direction of vascular impact risk, a timing adjustment strategy for switching between hot and cold, and a multi-needle synergistic timing adjustment strategy; the heterogeneity identification module integrates a tissue heterogeneity identification model, which is an inversion model based on biological heat transfer constraints, a machine learning model based on physical constraints, a finite element inverse problem solving model, an ensemble Kalman filter model, or a combination thereof; the execution and adjustment module is connected to the refrigerant supply and freezing control unit through a communication interface to adjust the refrigerant flow rate, probe temperature, freezing duration, or pipeline pressure in real time.

Images