Radiotherapy path planning method based on surface profile monitoring

Abstract

The invention discloses a radiotherapy path planning method based on surface profile monitoring, particularly relates to the field of medical image data processing, and is used for solving the problems of low reusability and low intelligent degree of existing medical radiotherapy path planning. The method comprises the following steps: constructing a distribution prediction model of radiotherapy incidence points by collecting marked target region characteristic parameters and incidence point distribution in a historical case, completing three-dimensional segmentation of a target region and an organ structure in a CT image of a current patient, and establishing a unified voxel coordinate system; constructing an incidence point array covering the surface of the target region according to a prediction result, generating a candidate radiotherapy path set, and performing risk organ recognition on each path; further combining the spatial distance between the path trajectory and the risk organ contour surface to calculate the comprehensive risk score of the candidate path; and finally, performing dose simulation on the candidate paths according to a scoring sequence, outputting a radiotherapy path planning result meeting dose and safety requirements, and improving the precision and clinical applicability of path selection.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image data processing, and more particularly to a radiotherapy path planning method based on surface contour monitoring. Background Art

[0002] During high-precision radiotherapy, beam path planning is directly related to target coverage and avoidance of adjacent normal organs. This is particularly true in complex anatomical regions such as the head and neck, thorax, abdomen, and pelvis, where targets often border multiple high-risk organs, such as the optic nerve, spinal cord, lung tissue, mesentery, and bladder. Even the slightest deviation in the beam path can result in excessive non-target dose exposure. However, current preliminary selection of radiotherapy paths generally relies on the experience of physicists, with manual multi-angle simulations and tentative settings based on three-dimensional imaging structures. The lack of an automated preliminary selection guidance mechanism results in inefficient path generation, leading to significantly different delivery strategies for similar lesions in different patients. Consequently, preliminary path planning lacks consistency and reusability. Furthermore, current radiotherapy path simulations often employ an anatomically based occlusion determination strategy to screen candidate paths. This strategy uses the determination of whether the beam path crosses critical organs as the basis for path availability. However, this approach often uses binary occlusion and non-occlusion results as criteria, lacking detailed assessment of spatial proximity between the path and organs, traversal length, or risk level, making it difficult to achieve refined ranking and optimization of multiple path options. At the same time, the existing process relies heavily on manual specification or one-by-one simulation, and lacks an automated path optimization mechanism. This makes it difficult to quickly screen out high-risk paths and determine the optimal solution under multiple path alternatives, seriously restricting the efficiency and quality of path planning. Summary of the Invention

[0003] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a radiotherapy path planning method based on surface contour monitoring to solve the problems raised in the above-mentioned background technology.

[0004] To achieve the above object, the present invention provides the following technical solutions: A radiotherapy path planning method based on surface contour monitoring includes the following steps: S1: Collect target area characteristic parameters and radiotherapy injection point distribution of patients in historical radiotherapy cases, and establish a distribution prediction model of radiotherapy injection points corresponding to the target area; S2: Segment the patient's tumor area on CT 3D images and construct a voxel structure containing the target area and organ contours; S3: Input the current patient target area characteristic parameters into the distribution prediction model and output the injection point planning area; S4: Generate a set of candidate radiotherapy paths covering the target surface based on the entry point array within the entry point planning area, and mark the risk organs penetrated by each candidate path; S5: Based on the spatial distance between the candidate pathway and the contour surface of the risk organ and the type of risk organ, the risk score of the candidate pathway is calculated; S6: Perform dose distribution simulation verification on candidate pathways in order of risk scores and output the final radiotherapy pathway planning results.

[0005] In a preferred embodiment, in S1, the target area characteristic parameters and the distribution of radiotherapy injection points of patients in historical radiotherapy cases are collected, and a distribution prediction model of the target area corresponding to the radiotherapy injection points is established, which specifically includes: Obtain structurally annotated CT image data from historical radiotherapy cases, and extract the location, volume, and boundary curvature of the radiotherapy target as target feature parameters; Extract the actual radiotherapy entrance point position corresponding to each radiotherapy target area, establish a mapping sample group between the target area characteristic parameters and the corresponding entrance point, and perform normalization on the sample group; The gradient boosting tree was selected as the regression model, and supervised training was performed based on the target feature parameters and the injection point positions in the normalized sample group to establish a radiotherapy injection point distribution prediction model.

[0006] In a preferred embodiment, in S2, performing structural segmentation on the patient's tumor region CT three-dimensional image and constructing a voxel structure including the target region and organ contours specifically includes: Reconstructing the CT scan slice sequence of the patient's tumor area into a multi-layer three-dimensional volume image according to the acquisition order, and establishing a three-dimensional coordinate system based on the preset voxel physical space size as the mapping basis; Segment the target area and organ structure in the 3D volume image, identify the target area and organ contour through the grayscale distribution and connectivity of the CT scan slice, and generate the corresponding 3D voxel mask; The target area mask and the voxel mask of the organ contour are integrated into a structure mask set, and all voxels contained in the structure mask set are uniformly labeled with the name of the structure to which they belong.

[0007] In a preferred embodiment, identifying the target area and organ contour by grayscale distribution and connectivity of CT scan slices and generating the corresponding three-dimensional voxel mask specifically includes: Grayscale screening of all voxels in the 3D volume image is performed based on the CT grayscale value range of the patient's target area and organs in historical radiotherapy cases; The upper and lower layer connectivity of the filtered voxel set is determined according to the CT slice direction, and the voxel clusters with cross projections in the continuous layers are retained, while the discrete voxels with no connected isolated layers are eliminated. Perform spatial boundary closure judgment on connected voxel clusters to screen and construct structurally complete target areas and organ contour voxel masks.

[0008] In a preferred embodiment, in S3, the characteristic parameters of the current patient target area are input into the distribution prediction model, and the output of the injection point planning area specifically includes: Obtain the target area 3D voxel mask of the current patient and perform spatial registration to extract target area feature parameters; The extracted target area characteristic parameters are normalized according to the normalization rules in the distribution prediction model training process, and input into the distribution prediction model to output the matching radiotherapy entrance point prediction results; Based on the radiotherapy entrance point prediction results, a continuous uniform entrance point array is constructed on the outer surface of the skin at the corresponding part as the entrance point planning area.

[0009] In a preferred embodiment, in S4, a set of candidate radiotherapy paths covering the target surface is generated based on the array of entry points within the entry point planning area, and the risk organs penetrated by each candidate path are marked, including: Obtaining an array of incident points in the planned area of ​​the patient's skin outer surface, and marking the spatial coordinates of each incident point according to the three-dimensional coordinate system of the voxel mapping; Starting from each incident point in the array, ray extension is performed on the target surface. The incident paths that reach the target surface are screened and marked as candidate paths. The coordinates of the path starting point, the direction vector and the target intersection sequence are recorded. Perform overlap detection between the incident path trajectory and the structure mask set on the candidate path set, identify the organ structure passed by the incident path, mark the corresponding organ structure as a risk organ, and record the risk organ type.

[0010] In a preferred embodiment, in S5, the risk score calculation of the candidate path based on the spatial distance between the candidate path and the risk organ contour surface and the risk organ type specifically includes: Screening candidate paths that do not penetrate the risk organ in the candidate path set, and calculating the shortest spatial distance between the candidate path trajectory and the voxels of the risk organ structure contour surface in the three-dimensional coordinate system of the voxel mapping; The corresponding risk weight factor is set according to the radioresistance of each risk organ, and the spatial distance between each candidate path trajectory and all risk organs is combined with the corresponding risk weight factor to perform a weighted comprehensive calculation; The weighted comprehensive calculation results of all candidate paths are converted into risk score values ​​of uniform scale.

[0011] In a preferred embodiment, in S6, the dose distribution simulation verification is performed on the candidate paths in the order of risk scores, and the output of the final radiotherapy path planning result specifically includes: Sort the candidate paths in ascending order according to the risk score value to form a queue of candidate paths to be verified; Select candidate paths in the queue in turn, build a radiation beamline model according to the coordinates of the path starting point, direction vector and the corresponding target area intersection sequence, and initialize the beam simulation parameters; Beam energy attenuation calculation and dose deposition simulation verification are performed on the target mask, and the first qualified candidate path is selected from the verified candidate paths as the final radiotherapy path planning result.

[0012] The technical effects and advantages of the radiotherapy path planning method based on surface contour monitoring of the present invention are as follows: By constructing a predictive model between target characteristics and entry point distribution, the transformation of radiotherapy path planning from experience-driven to data-driven is achieved, which improves the accuracy and efficiency of path screening. The spatial relationship between the target and the risk organ is expressed in a unified coordinate system in combination with the three-dimensional voxel structure, providing structured support for path screening. By introducing the entry point planning area to limit the path generation range, the amount of invalid path calculation is reduced, and the path geometric constraints are strengthened. The risk annotation of the candidate path not only considers the type of penetrated organ, but also introduces spatial distance and organ biological properties to jointly construct a risk score, making the path evaluation results more clinically relevant. Finally, the dose distribution simulation is performed in combination with the risk score to screen the final path, effectively balancing treatment intensity and normal tissue protection, improving the safety and controllability of the radiotherapy path, and having the technical advantages of clear structure, reusable model, and physically feasible path output. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 Schematic diagram of a radiotherapy path planning method based on surface contour monitoring according to the present invention. DETAILED DESCRIPTION

[0014] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0015] Example 1 Figure 1 The present invention provides a radiotherapy path planning method based on surface contour monitoring, which includes the following steps: S1: Collect target area characteristic parameters and radiotherapy injection point distribution of patients in historical radiotherapy cases, and establish a distribution prediction model of radiotherapy injection points corresponding to the target area; S2: Segment the patient's tumor area on CT 3D images and construct a voxel structure containing the target area and organ contours; S3: Input the current patient target area characteristic parameters into the distribution prediction model and output the injection point planning area; S4: Generate a set of candidate radiotherapy paths covering the target surface based on the entry point array within the entry point planning area, and mark the risk organs penetrated by each candidate path; S5: Based on the spatial distance between the candidate pathway and the contour surface of the risk organ and the type of risk organ, the risk score of the candidate pathway is calculated; S6: Perform dose distribution simulation verification on candidate pathways in order of risk scores and output the final radiotherapy pathway planning results.

[0016] In S1, the target area characteristic parameters and radiotherapy entrance point distribution of patients in historical radiotherapy cases are collected, and a distribution prediction model of the target area corresponding to the radiotherapy entrance points is established.

[0017] We collected 3D CT imaging datasets from patients undergoing previous radiotherapy treatments, which had undergone structural segmentation and annotation. Each dataset contained the physician-verified tumor target structure and the corresponding treatment path planning records. The retrieved imaging data had been constructed using a DICOM 3D image volume, with slice spacing no greater than 2 mm and a uniform image matrix size of 512 × 512 × N, where N represents the number of slices scanned for the patient. This met the spatial accuracy requirements for subsequent structural feature extraction. For each dataset, we extracted the structural annotation information and extracted a voxel mask representing the tumor target within the image volume (see step S2 in the manual for details on the construction process). This mask indicated whether each voxel in the image volume belonged to tumor tissue. Spatial geometric processing was performed on the non-zero voxel coordinates within the mask structure, and the position of their geometric center within the 3D coordinate system was calculated (see step S2 in the manual for details on the coordinate system construction process). This was used as the target volume feature parameter. Furthermore, the number of non-zero voxels in the mask was counted and, combined with the actual spatial dimensions of each voxel (set to 2 mm³ by default), the actual tumor volume was calculated as the volume feature parameter. For boundary curvature extraction, surface reconstruction is first performed based on the voxel surface boundary to extract the closed surface model. Then, the average Gaussian curvature of each vertex is calculated on the three-dimensional mesh model, and the mean and variance of the overall curvature distribution are statistically calculated as the morphological complexity index. If there are structural burrs or non-closed structures in the curvature extraction area, the abnormal surface points are removed and bidirectional triangular mesh smoothing is performed to ensure that the boundary curvature extraction results are stable and usable. Finally, the position coordinates, volume values ​​and boundary curvature indicators are combined into a unified target area feature parameter vector as the input variable for subsequent modeling. All extraction results are stored in a unified structured data format to ensure that they can be read and compared in batches during the model training phase.

[0018] After extracting the characteristic parameters of each target volume, the actual radiotherapy pathway used for each patient was further obtained based on historical records, with a focus on extracting the radiotherapy entry point of the main pathway. This entry point is defined as the location where the treatment pathway first penetrates the patient's body surface and is recorded in the radiotherapy plan document as a body position annotation. To ensure data consistency, the entry point is registered to a three-dimensional coordinate system consistent with the CT image. This registration operation is performed based on the image origin positioning parameters and the coordinate transformation matrix, with an error control within 2 mm. For each sample, the extracted target volume characteristic parameters are combined with the corresponding entry point coordinates to form a set of input-output pairs, with the feature vector as the input and the three-dimensional position of the entry point as the target output. This is used to establish a supervised learning sample set. To construct the sample size, data from at least 500 patients with tumors in various locations were selected, covering multiple target volume types such as lung cancer, liver cancer, and brain tumors. Each sample type was labeled to facilitate subsequent analysis and generalization. To improve the numerical convergence efficiency of the model training phase and reduce the bias caused by differences in feature dimensions, all input and output variables in the sample set were normalized. Each dimension of the input vector is normalized using a min-max method, mapping the maximum and minimum values ​​of each dimension within the entire sample range to the interval [0, 1]. The output coordinates are normalized axis-by-axis based on the effective beam incidence range determined by the patient's body contour to prevent coordinate values ​​from falling into abnormal areas. All normalization parameters are fixed during the dataset construction phase to ensure consistency between training and prediction. This completes the construction of a standardized sample set of feature-target pairs.

[0019] Based on the constructed normalized sample set, a gradient boosting tree algorithm (GBT) was selected as the regression model to learn the nonlinear mapping between target feature parameters and the spatial coordinates of the actual impact point. The GBT model was chosen because it has high expressive power for small- to medium-scale structured data, can capture complex combinations of target morphological features and spatial impact preferences, and exhibits strong resistance to overfitting and ability to interpret feature importance. In the modeling architecture, three independent regressors were constructed to predict the x, y, and z coordinates of the impact point (for details, see step S2 in the instruction manual). This ensured that the fitting process for each dimension was independent and the gradient path was stable. During training, the input was the normalized target feature vector, and the output was the corresponding normalized impact point coordinate value. The mean squared error (MSE) metric was used as the loss function. The number of iterations was set to 300, the minimum number of leaf samples was set to 10, and the learning rate was fixed at 0.05. To avoid training anomalies caused by sample order or data bias, a 5-fold cross-validation mechanism was used during training, with stratified sampling by case number. The stability of the model's error convergence on the validation set was evaluated at each iteration. After the model converges, the output coordinates are denormalized according to the original normalization parameters to restore the actual entry point coordinates. The Euclidean distance is then compared with the true entry point coordinates as an evaluation metric for model prediction accuracy. Across the full test set, the proportion of prediction errors less than 5 mm reached over 82% (dynamically adjustable based on accuracy requirements), demonstrating the model's excellent spatial fitting capabilities. Ultimately, the model was named the entry point distribution prediction model and deployed. It accepts target feature input for any new patient during radiotherapy path planning and rapidly outputs the entry point planning region, providing a preliminary constraint range for candidate path screening and dose distribution simulation.

[0020] In S2, the patient's tumor area CT three-dimensional image is structurally segmented to construct a voxel structure containing the target area and organ contours.

[0021] The patient's CT scan image data was organized. The input data consisted of multiple consecutive 2D grayscale images in DICOM format, each corresponding to a single scan slice. The image sequence was ordered along the z-axis for hierarchical encoding. Each image had a uniform resolution of 512 × 512 pixels, with the pixel pitch clearly indicated in the image metadata. The slice spacing was set to no more than 2 mm in the acquisition protocol, and an equidistant scanning mode was used. (All CT image-related parameters were set to remain consistent with those in step S1.) The image sequence was sorted by slice number and stacked sequentially according to the image matrix size and slice spacing to form a 3D image volume. The data structure of this 3D image volume consisted of a 3D grayscale voxel array, with each voxel retaining the CT value at its location. To ensure a consistent spatial reference for subsequent structural analysis, path penetration assessment, and dose simulation, a 3D coordinate system for physical space mapping was established on the image volume. This coordinate system was centered at the pixel at slice 0 in the upper left corner of the image volume. The x- and y-axes along the rows and columns were defined, and the z-axis was defined perpendicular to the slice. Coordinate units were millimeters, consistent with CT parameters. The voxel spatial position index is directly mapped to a three-dimensional coordinate system (x, y, z), which is calculated using a linear mapping method. The x-direction increment is the pixel width, the y-direction increment is the pixel height, and the z-direction increment is the slice layer spacing. In this embodiment, the standard physical size of the voxel is set to 1mm×1mm×2mm, and the voxel resampling operation is performed on the original image to make the spatial resolution of the reconstructed image consistent in the three-dimensional direction, thereby ensuring the accuracy of the coordinate expression in the subsequent voxel-level structure annotation, ray path construction and dose mapping calculation. The final completed three-dimensional volume image and the corresponding coordinate system will serve as the spatial reference framework for downstream target area identification and structure segmentation processing.

[0022] Based on the establishment of a unified coordinate system for a three-dimensional image volume, segmentation of the tumor target volume and organ structures is performed. In this embodiment, the structural recognition process combines grayscale threshold-based initial screening with spatial connectivity enhancement. First, a statistical analysis of the CT grayscale values ​​within the image volume is performed. Known target tissues in the sample set typically exhibit density differences from adjacent tissues. For example, soft tissue tumors often exhibit a grayscale range between lung parenchyma and bone density. A preliminary screening threshold for target candidate regions is set. For example, voxel regions with grayscale values ​​between 40 and 80 (the specific threshold range is determined based on the grayscale values ​​of the target volume and organs in historical CT images) are extracted as a preliminary set of candidate voxels. Within-slice regional connectivity analysis is then performed on each candidate voxel set. Only voxel clusters that are closed within the slice layer and have structural continuity across multiple consecutive slices are retained. Continuity between slices is determined by setting a maximum connectivity gap within two layers in the Z-axis direction, requiring that the distance between the centroids of voxel regions in adjacent layers does not exceed 5 mm. For target areas with blurred edges or small structures, a three-dimensional morphological closing operation is further introduced to fill holes and repair pseudo-cracks in the segmentation results to ensure the spatial coherence and integrity of the target area structure. Organ structure recognition is performed using a matching method based on standard anatomical position templates. For example, high-frequency risk organs such as the spinal cord, heart, and lungs have relatively stable spatial position distributions between different patients. After positioning based on the thoracic bony reference line, the approximate center of the organ can be quickly located, and grayscale value threshold extraction and connectivity enhancement operations are performed in the central area. Ultimately, each identified structure is represented as a voxel mask, that is, a three-dimensional array consistent with the image volume dimension. Voxels with a median value of 1 in the mask represent voxels within the structure, and the remaining voxels are set to 0. The output results include a target area mask and multiple organ masks, each corresponding to a unique structure type.

[0023] After the voxel masks of each individual structure are generated, the spatial representations of all structures are unified into a structured mask set. In this embodiment, a three-dimensional volume image with an image size of 512×512×N is used as a unified reference dimension, and a structure label array is defined. The type is a three-dimensional integer array, and the initial value is 0, representing an unassigned state. Subsequently, a voxel-by-voxel scanning operation is performed on the target mask array. For every voxel with a mask value of 1, the voxel at the same position in the corresponding structure label array is assigned a preset number 1 and identified as a tumor target. Then, the organ structure mask is scanned sequentially and the structure label array is written in sequence according to the structure. For example, the heart is assigned number 2, the spinal cord is assigned number 3, the left lung is 4, and the right lung is 5. The numbering order has a fixed definition to prevent identification conflicts when structures overlap. During the writing process, for points where non-zero numbers already exist at the voxel position, such as the boundary overlapping area between organs, the structure number with a smaller number is retained and is considered a higher-risk structure. To ensure that structure labels correspond to clinical terminology, each structure number is permanently mapped to a specific name in the configuration file, such as number 2 for "heart" and number 3 for "spinal cord." Each voxel in the final structure mask set has a unique structural identifier, and the entire mask set forms a spatial structure annotation map that can be used for path penetration assessment, dose distribution mapping, and risk organ identification. This structure label map not only has spatial positioning capabilities but also possesses anatomical semantic expression capabilities, serving as a core reference structure during radiotherapy path planning and candidate path risk analysis.

[0024] In S3, the characteristic parameters of the current patient target area are input into the distribution prediction model, and the injection point planning area is output.

[0025] If the original acquisition direction of the patient's image is different from the direction used during the training of the distribution prediction model, such as the head and feet are reversed, or the left and right are flipped, a direction matrix transformation is performed based on DICOM positioning to ensure that the mask structure is consistent with the historical model sample in the spatial direction. After the registration is completed, the target feature parameters are extracted from the registered voxel set. The target feature parameters are obtained in the same way as the target feature parameters in step S1 of the manual. The target feature parameters include position coordinates, volume values ​​and boundary curvature indicators. The format is consistent with the historical training sample structure in step S1 of the manual, and is prepared as input data for the distribution prediction model. The extracted target feature parameters are standardized according to the normalization rules in the distribution prediction model training process. After the normalization is completed, the input is the standardized feature vector, and the output is the radiotherapy entrance point prediction result. The result is a three-dimensional spatial coordinate value, which indicates the spatial position of the optimal radiotherapy entrance point corresponding to the current patient's target area on the skin surface. The prediction result is the main path entrance starting point learned by the model after combining the historical experience path preference with the current target area spatial configuration. The model uses a gradient-boosted tree architecture. During training, independent regression paths are established for the x, y, and z coordinate dimensions, resulting in inference outputs of three continuous coordinate values. Once the predictions are generated, they are mapped back to the 3D coordinate system corresponding to the CT image, ensuring consistency with the image volume, structural mask, and other path planning data.

[0026] After obtaining the predicted coordinates of the radiotherapy entry point, an expansion process is constructed around the body surface location of these coordinates to define the entry point planning region, which forms the basis for subsequent candidate path generation. Specifically, the predicted coordinates are first located within the 3D image volume and determined to be within the image boundary. If they are close to the image boundary, the region center is adjusted to within 10 mm (or the 3D coordinate system is extended based on the voxel size) to ensure the complete generation of the subsequent array. The skin region is then searched and constructed by performing ray tracing in the outward normal direction from the predicted entry point. The first non-in-body voxel is identified as the skin surface point, and the surface segment at this point is intercepted as the construction basis. With this point as the center, an orthogonal coordinate region is constructed along the skin surface to form a 2D unwrapped planar region. The physical dimensions of this region are set to 40 mm × 40 mm and divided into a 10 × 10 grid. All grid intersections are defined as the candidate entry point array. The 3D spatial coordinates and normal direction vector of each entry point are recorded. This array point set is called the entry point planning region and represents all possible entry positions within an anatomically plausible region centered at the predicted entry point. During the path generation phase, all points in this array are considered potential path starting points for ray extension, expanding the candidate path set and improving coverage accuracy. By flattening the surface and then mapping the three-dimensional coordinates, a high-quality starting point set is constructed that both conforms to the skin surface morphology and exhibits spatial density uniformity, providing clear regional input for subsequent path optimization and structural obstacle avoidance.

[0027] In S4, a set of candidate radiotherapy paths covering the target surface is generated based on the entry point array within the entry point planning area, and the risk organs penetrated by each candidate path are marked.

[0028] After the entry point planning region is constructed, the entry point array within that region is assigned three-dimensional coordinates and structured output. This array is formed by uniformly dividing a two-dimensional grid, with each grid intersection representing an entry point. Since the entry points are located on the outer surface of the skin, accurate physical spatial location requires coordinate mapping using a three-dimensional voxel coordinate system established using the patient's CT image. First, the two-dimensional unfolded reference plane of the entry point planning region is obtained and reprojected onto the corresponding curved surface location on the outer surface of the skin in three-dimensional space. The three-dimensional coordinates of each entry point are calculated using the voxel index and the physical spacing of individual voxels. If the patient image was acquired with unequal spacing, the voxel coordinates of the corresponding axes are converted based on the pixel size of the corresponding axis. To ensure data consistency, the spatial coordinates of all entry points are recorded in a unified unit and stored as three-dimensional vectors. The entry point array ultimately forms a set of points containing coordinate vectors and surface normal vectors. The normal vectors are used to determine the direction of subsequent path extension without introducing subjective bias. A set of starting coordinates of radiotherapy injection points with physical spatial consistency based on the patient's actual skin surface is established, and subsequent path construction tasks are performed based on this point set. Based on the obtained injection point array, the path extension operation facing the target area is initiated point by point. Each injection point is equipped with its surface normal vector as the initial extension direction, so as to construct a ray trajectory that penetrates the voxel structure, which is in the form of a spatial straight line segment extending from the starting point of the skin surface to the internal space. In order to prevent the path from deviating from the target space, voxel-level spatial tracking is performed on each ray during the extension process. The method is to gradually calculate the voxel sequence that the ray passes through and detect whether it intersects with the target mask. Once the ray overlaps with the voxel in the target mask, it is considered that the ray has successfully penetrated the target structure, and the path is marked as a valid injection path and added to the candidate path set. At the same time, three key parameters of the path are recorded: the starting point of the path, that is, the coordinates of the incident point; the direction vector, which is derived from the surface normal but has been normalized; and the intersection sequence, that is, the continuous spatial position set formed by the path crossing the target structure voxels, reflecting the contact trajectory of the path within the target surface. Ray paths that fail to enter the target area are directly discarded and not recorded. After the above processing is completed, a set of candidate radiotherapy paths that truly penetrate the target area is formed. Each path has a complete structure, clear geometric information and incident direction, and can provide accurate spatial input data for subsequent risk structure detection and dose calculation.

[0029] After obtaining a set of candidate paths, the focus is on identifying the human structures penetrated by each path to determine whether any path passes through a critical organ. Specifically, the ray path from the starting point to the target intersection of each path in the candidate path set is traversed, and the corresponding continuous spatial position sequence in the voxel structure is progressively compared with the structure mask set. The structure mask set contains multiple labeling layers that record the voxel location distribution and name information of different organ structures. The rule for detecting overlap between paths and masks is: if the voxel location traversed by the path intersects with the mask voxel set of a specific organ, the path is considered to pass through that organ structure. To improve detection accuracy, each path performs a voxel-by-voxel intersection check, rather than just checking the starting and ending points. Furthermore, the target structure itself is excluded from the overlap check to avoid duplicate labeling of path self-penetrations. The name of each overlapping organ structure is extracted from the mask set and appended to the structure penetration list for that path. If the organ penetrated is a medically defined critical organ, such as the lungs, liver, heart, or kidneys, the path is marked as a path with an organ risk. Organ names are recorded as classification criteria in this step, providing semantic information for subsequent risk weight assignment and pathway prioritization. After this detection operation is completed, each candidate pathway has complete spatial geometry information and risk structure labels. The pathway set has been transformed from a purely geometric structure into a set of clinically interpretable functional pathways, meeting the prerequisites for pathway screening and dose regulation.

[0030] In S5, a risk score is calculated for the candidate path based on the spatial distance between the candidate path and the risk organ contour surface and the risk organ type.

[0031] After the candidate pathway set is constructed and the organ penetration information of each pathway is annotated, a subset of pathways that do not penetrate any risk organs is selected for distance evaluation. A candidate pathway that does not penetrate a risk organ refers to a pathway that does not overlap with any voxel regions defined as high-risk organs in the structural mask set. Such pathways are generally superior in spatial planning and have high clinical acceptability, but their proximity to risk organs requires further evaluation. To this end, the shortest spatial distance between each pathway and the voxel boundary surfaces of all risk organs is calculated on a per-organ basis in a 3D voxel coordinate system. Each pathway is defined by a starting point, an extension direction vector, and a sequence of target region intersection points. In space, it is represented as a straight line segment penetrating the target region. Distance measurements are performed based on the 3D voxel boundary voxel set corresponding to each pathway and each organ structure. The voxel boundary voxel set consists of all edge voxels adjacent to non-structural voxels in the organ mask, i.e., the contour surface of the organ. For each boundary voxel, the shortest spatial distance between its center point and the path trajectory is calculated, and a minimum value screening method is used to obtain the minimum distance value between each path and each organ. Because all distance calculations are performed in a unified three-dimensional voxel coordinate system and the voxel size is standardized, the obtained distances are physically consistent. The shortest spatial distance between each candidate path and each risk organ is stored in matrix form, providing complete input data for subsequent risk score calculations. This type of processing avoids the problem of discontinuity in path penetration detection and compensates for the defect of traditional radiotherapy path evaluation that over-relies on penetrating structures and ignores nearby hazards.

[0032] After constructing the shortest distance matrix, a comprehensive risk score is calculated for each candidate pathway based on a joint assessment mechanism combining organ radiosensitivity and pathway geometric distance. First, based on clinical data and a knowledge base of radiation biology, the radioresistance levels of common critical organs are defined. A fixed risk weight factor is assigned to each at-risk organ accordingly. This factor reflects the organ's tolerance to radiation damage, with the weaker the organ's cell regeneration capacity, the higher the corresponding weight factor. For example, critical structures such as the spinal cord, optic nerve, and kidneys are assigned a higher risk factor (e.g., 1.0), while organs with relatively strong regenerative capacity, such as the liver and lungs, are assigned lower factors (e.g., 0.3-0.5). These risk factors are pre-set by field professionals based on the patient's actual organ function and maintained uniformly throughout the entire process to ensure cross-comparability of the scoring results. A weighted distance score is calculated for each candidate pathway relative to all at-risk organs. Specifically, for each at-risk organ, a distance scoring function (e.g., inverse distance, logarithmic function, etc.) is applied to the shortest distance between the pathway and the organ. This is then multiplied by the corresponding organ's risk weight factor to determine the risk contribution of the pathway to that organ. The contribution values ​​of all risk organs are accumulated to form the total risk score for the pathway. To ensure a consistent output scale for different pathway scores, all raw score results are normalized to fall between a fixed interval (e.g., 0 and 1), ultimately outputting a standardized risk score. A higher score indicates a pathway that is closer to highly sensitive organs and has lower overall radiation tolerance, thus increasing clinical risk. Conversely, a lower score indicates a safer pathway.

[0033] In S6, the dose distribution simulation is performed on the candidate paths in the order of risk scores, and the final radiotherapy path planning results are output.

[0034] After calculating the risk scores of candidate paths and obtaining standardized scores based on a unified scale, a sorting operation is performed based on the scores. The candidate paths are stored in ascending order as a linear path queue to construct a candidate path queue for verification, which serves as the input source for the subsequent beam simulation verification module. This path queue logically represents the order of path selection from theoretically safest to relatively high-risk. Its core purpose is to prioritize the optimal path with the goal of minimizing potential secondary damage.

[0035] In the sorted candidate path queue, candidate path entries are read sequentially, starting from the first position. Each entry contains three key geometric parameters: the 3D coordinates of the path's starting point, the path's direction vector, and the sequence of intersection points formed by the path's intersection with the target voxel boundary as it penetrates the target volume. These parameters fully describe the radiation beam's propagation behavior and injection pattern in 3D space. Based on these three parameters, a geometric model of the radiation beam is constructed, serving as the path foundation for subsequent dose simulation and energy deposition. When constructing the beamline model, the path's starting point serves as the simulated particle source, the direction vector defines the beamline's direction, and the intersection point sequence defines the boundaries of the target energy deposition region. During the simulation parameter initialization phase, the radiation source energy level (e.g., 6MV or 10MV X-rays), beamline aperture size (e.g., with a 1cm safety margin around the target volume boundary), energy propagation spacing (e.g., one dose unit per millimeter), and dose decay model (e.g., exponential decay function or Monte Carlo approximation) are set. All parameters are defined in real physical units and based on industry standards. After each initialization is completed, a simulated beam model of the path is constructed, and a mapping relationship from the incident point to each energy distribution unit in the target area is established in the three-dimensional voxel structure. All path construction processes are static processes, which are not affected by the deformation of the target area and do not consider the offset caused by dynamic breathing, and are highly controllable. After completing the beamline model construction and parameter initialization, the beam energy propagation and dose deposition simulation operations are started based on the current candidate path. First, the penetration area is traversed voxel by voxel along the beam direction, and the beamline unit energy intensity is updated in real time according to the energy attenuation model. An approximate physical model is used to simulate the energy loss behavior of X-rays in different tissues, and the residual dose intensity in each voxel is calculated based on the path traversal length and tissue density, and the dose distribution map is accumulated to form a dose distribution map.

[0036] After the dose distribution map is constructed, it is mapped to the target mask area to evaluate the match between the actual deposited dose and the clinical prescription dose. The evaluation criteria include but are not limited to the following three items: whether the average dose in the target area reaches the preset prescription dose lower limit; whether there are obvious underdose voxels in the target area edge area; and whether there are overdose areas that overflow the target area boundary to non-target tissue. The evaluation criteria are set using quantitative thresholds. For example, at least 95% of the target area voxels must meet no less than 90% of the prescription dose. If the dose simulation results corresponding to the candidate path fully meet the above clinical dose coverage standards, the path is marked as verified and output as the final radiotherapy path, terminating the subsequent path verification process. If the current path does not meet the requirements, it automatically moves to the next path and repeats the above simulation and verification process. Through this strategy, the path with the lowest risk score is preferentially selected without compromising dose adaptability, forming the optimal "low risk + high dose matching" comprehensive path plan. This process avoids manual intervention and path trial and error, greatly improving path planning efficiency and clinical usability. At the same time, it provides the path selection process with a clear logical chain and evaluation criteria, which is in line with the engineering development trend of modern intelligent radiotherapy.

[0037] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters and thresholds in the formulas are set by technicians in this field according to actual conditions.

[0038] The above embodiments can be implemented in whole or in part via software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions or computer programs. When loaded or executed on a computer, the processes or functions described in the embodiments of this application are fully or partially performed. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired means (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server or data center that contains a collection of one or more available media. The available medium can be magnetic media (e.g., floppy disks, hard disks, tapes), optical media (e.g., DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0039] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0040] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and modules described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0041] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the modules is only a logical function division. In actual implementation, there may be other division methods, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0042] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, and may be located in one place or distributed across multiple network modules. Some or all of the modules may be selected to achieve the purpose of this embodiment according to actual needs.

[0043] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0044] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0045] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0046] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A radiotherapy path planning method based on surface contour monitoring, characterized in that: The steps include: S1: Collect target area characteristic parameters and radiotherapy injection point distribution of patients in historical radiotherapy cases, and establish a distribution prediction model of radiotherapy injection points corresponding to the target area; S2: Segment the patient's tumor area on CT 3D images and construct a voxel structure containing the target area and organ contours; S3: Input the current patient target area characteristic parameters into the distribution prediction model and output the injection point planning area; S4: Generate a set of candidate radiotherapy paths covering the target surface based on the entry point array within the entry point planning area, and mark the risk organs penetrated by each candidate path; S5: Based on the spatial distance between the candidate pathway and the contour surface of the risk organ and the type of risk organ, the risk score of the candidate pathway is calculated; S6: Perform dose distribution simulation verification on candidate pathways in order of risk scores and output the final radiotherapy pathway planning results.

2. A radiotherapy path planning method based on surface contour monitoring according to claim 1, characterized in that: In S1, the target area characteristic parameters and radiotherapy injection point distribution of patients in historical radiotherapy cases are collected, and a distribution prediction model of radiotherapy injection points corresponding to the target area is established. Specifically, the following steps are involved: Obtain structurally annotated CT image data from historical radiotherapy cases, and extract the location, volume, and boundary curvature of the radiotherapy target as target feature parameters; Extract the actual radiotherapy entrance point position corresponding to each radiotherapy target area, establish a mapping sample group between the target area characteristic parameters and the corresponding entrance point, and perform normalization on the sample group; The gradient boosting tree was selected as the regression model, and supervised training was performed based on the target feature parameters and the injection point positions in the normalized sample group to establish a radiotherapy injection point distribution prediction model.

3. The radiotherapy path planning method based on surface contour monitoring according to claim 1, characterized in that: In S2, the patient's tumor area CT 3D image is segmented to construct a voxel structure containing the target area and organ contours. Specifically, the following steps are performed: Reconstructing the CT scan slice sequence of the patient's tumor area into a multi-layer three-dimensional volume image according to the acquisition order, and establishing a three-dimensional coordinate system based on the preset voxel physical space size as the mapping basis; Segment the target area and organ structure in the 3D volume image, identify the target area and organ contour through the grayscale distribution and connectivity of the CT scan slice, and generate the corresponding 3D voxel mask; The target area mask and the voxel mask of the organ contour are integrated into a structure mask set, and all voxels contained in the structure mask set are uniformly labeled with the name of the structure to which they belong.

4. The radiotherapy path planning method based on surface contour monitoring according to claim 3, characterized in that: The identifying of the target area and the organ contour by the grayscale distribution and the connectivity of the CT scan slices and generating the corresponding three-dimensional voxel mask specifically includes: Grayscale screening of all voxels in the 3D volume image is performed based on the CT grayscale value range of the patient's target area and organs in historical radiotherapy cases; The upper and lower layer connectivity of the filtered voxel set is determined according to the CT slice direction, and the voxel clusters with cross projections in the continuous layers are retained, while the discrete voxels with no connected isolated layers are eliminated. Perform spatial boundary closure judgment on connected voxel clusters to screen and construct structurally complete target areas and organ contour voxel masks.

5. The radiotherapy path planning method based on surface contour monitoring according to claim 1, characterized in that: In S3, the characteristic parameters of the current patient target area are input into the distribution prediction model, and the output entry point planning area specifically includes: Obtain the target area 3D voxel mask of the current patient and perform spatial registration to extract target area feature parameters; The extracted target area characteristic parameters are normalized according to the normalization rules in the distribution prediction model training process, and input into the distribution prediction model to output the matching radiotherapy entrance point prediction results; Based on the radiotherapy entrance point prediction results, a continuous uniform entrance point array is constructed on the outer surface of the skin at the corresponding part as the entrance point planning area.

6. The radiotherapy path planning method based on surface contour monitoring according to claim 1, characterized in that: In S4, based on the entry point array within the entry point planning area, a set of candidate radiotherapy paths covering the target surface is generated, and the risk organs penetrated by each candidate path are marked, including: Obtaining an array of incident points in the planned area of ​​the patient's skin outer surface, and marking the spatial coordinates of each incident point according to the three-dimensional coordinate system of the voxel mapping; Starting from each incident point in the array, ray extension is performed on the target surface. The incident paths that reach the target surface are screened and marked as candidate paths. The coordinates of the path starting point, the direction vector and the target intersection sequence are recorded. Perform overlap detection between the incident path trajectory and the structure mask set on the candidate path set, identify the organ structure passed by the incident path, mark the corresponding organ structure as a risk organ, and record the risk organ type.

7. The radiotherapy path planning method based on surface contour monitoring according to claim 1, characterized in that: In S5, based on the spatial distance between the candidate path and the contour surface of the risk organ and the type of risk organ, the risk score of the candidate path is calculated, specifically including: Screening candidate paths that do not penetrate the risk organ in the candidate path set, and calculating the shortest spatial distance between the candidate path trajectory and the voxels of the risk organ structure contour surface in the three-dimensional coordinate system of the voxel mapping; The corresponding risk weight factor is set according to the radioresistance of each risk organ, and the spatial distance between each candidate path trajectory and all risk organs is combined with the corresponding risk weight factor to perform a weighted comprehensive calculation; The weighted comprehensive calculation results of all candidate paths are converted into risk score values ​​of uniform scale.

8. The radiotherapy path planning method based on surface contour monitoring according to claim 1, characterized in that: In S6, the dose distribution simulation is performed on the candidate pathways in the order of risk scores, and the final radiotherapy pathway planning results are output, including: Sort the candidate paths in ascending order according to the risk score value to form a queue of candidate paths to be verified; Select candidate paths in the queue in turn, build a radiation beamline model according to the coordinates of the path starting point, direction vector and the corresponding target area intersection sequence, and initialize the beam simulation parameters; Beam energy attenuation calculation and dose deposition simulation verification are performed on the target mask, and the first qualified candidate path is selected from the verified candidate paths as the final radiotherapy path planning result.

Images