CN117766108A

CN117766108A - Method for generating three-dimensional needle track model in DICOM image based on TPS report

Info

Publication number: CN117766108A
Application number: CN202311166706.0A
Authority: CN
Inventors: 姜冠群; 赵毅
Original assignee: Shandong Zhuoye Medical Technology Co ltd
Current assignee: Shandong Zhuoye Medical Technology Co ltd
Priority date: 2023-09-11
Filing date: 2023-09-11
Publication date: 2024-03-26

Abstract

The invention relates to the technical field of medical clinic, in particular to a method for generating a three-dimensional needle track model in a DICOM image based on a TPS report. The method comprises the following steps: performing data integration and data preprocessing of corresponding image data on the DICOM image data and the TPS report image data to generate standard image data; performing skin-layer image extraction processing of DICOM image data on the standard image data to generate skin-layer DICOM image data; performing three-dimensional space mapping and optimization of TPS report image needle track coordinates on standard image data to generate a needle track model; and reconstructing a three-dimensional needle model of the image by using the needle model and the DICOM image data of the skin layer, and performing reinforcement learning of the image needle model by using a deep reinforcement learning model to generate an optimized image needle model. According to the method, the three-dimensional needle track model is constructed more accurately by optimizing the reconstructed three-dimensional needle track model through the reinforcement learning model.

Description

Method for generating three-dimensional needle track model in DICOM image based on TPS report

Technical Field

The invention relates to the technical field of medical clinic, in particular to a method for generating a three-dimensional needle track model in a DICOM image based on a TPS report.

Background

The method for generating the three-dimensional needle track model in the DICOM image based on TPS (Template-based Planning System) report is used for improving the efficiency and accuracy of medical image processing. The information contained in the TPS report can automatically generate the needle track model, and by the method, a doctor can rapidly plan and simulate the needle track path on an image without manually drawing, so that the time is saved, the human error is reduced, and the treatment accuracy and safety are improved. However, the conventional generation of the three-dimensional needle track model in the DICOM image based on the TPS report only aligns the DICOM image with the image of the TPS report, and then the image of the TPS report may not be completely aligned with the DICOM image due to the artificial adjustment, so that the construction of the three-dimensional needle track model has deviation, resulting in inaccurate results and other problems.

Disclosure of Invention

Based on the above, the present invention provides a method for generating a three-dimensional needle track model in DICOM images based on TPS reports, so as to solve at least one of the above-mentioned technical problems.

To achieve the above object, a method for generating a three-dimensional needle track model in DICOM images based on TPS reports includes the steps of:

step S1: acquiring DICOM image data and TPS report image data in a medical database; performing data integration and data preprocessing of corresponding image data on the DICOM image data and the TPS report image data to generate standard image data;

Step S2: performing skin-layer image extraction processing of DICOM image data on the standard image data to generate skin-layer DICOM image data;

step S3: carrying out needle track coordinate extraction of TPS report images on the standard image data to generate needle track coordinate data; performing three-dimensional space mapping and optimization of needle track coordinates according to the needle track coordinate data to generate a needle track model;

step S4: performing space mapping alignment and optimization on the skin layer DICOM image data according to the needle track model to generate optimized DICOM image mapping data; reconstructing a three-dimensional needle track model of the image by using optimized DICOM image mapping data to generate an image needle track model;

step S5: and performing reinforcement learning on the image needle channel model by using the deep reinforcement learning model, so as to generate an optimized image needle channel model.

In medical image research, the accuracy and consistency of the data are very important, the corresponding relation of the medical image data can be established by integrating the DICOM image data and TPS report image data in the medical database, the standardization of the image quality can be carried out in the data preprocessing stage, and noise, artifacts and the like possibly existing are removed, so that the stability and reliability of subsequent processing are ensured, and the accuracy and efficiency of subsequent steps are improved. In medical image processing, it is generally necessary to locate and identify a specific tissue or structure, and to extract skin layer images from standard image data to obtain spatial position information of the skin layer, which is critical for subsequent needle track coordinate extraction and construction of a needle track model, and to extract the skin layer, so as to reduce the computational complexity in subsequent steps and reduce possible error sources. The TPS report contains the information of the needle track coordinates, the needle track coordinates of the TPS report image are extracted from the standard image data, the spatial position information of the needle track in the image can be obtained, the needle track coordinates are subjected to three-dimensional spatial mapping and optimization, and then the real needle track model is restored more accurately, so that the skin DICOM image data and the needle track model can be aligned more accurately in the subsequent steps, and possible positioning deviation is avoided. After the DICOM image data of the skin layer and the needle track model are subjected to space mapping alignment and optimization, more accurate DICOM image data is obtained, the reconstruction quality of the image needle track model is improved, artifacts or deviation caused by image alignment errors are reduced, and the accurate shape and position of the needle track model in an image can be better reserved by optimizing the DICOM image mapping data. The deep reinforcement learning is a powerful machine learning technology, the performance and accuracy of the model can be further optimized by reinforcement learning of the image needle channel model, the deep reinforcement learning model can perform feedback learning according to the output result and the expected result of the model, and therefore the performance of the needle channel model is continuously improved, and the robustness of the model can be improved through the process, so that the method can adapt to wider scenes. Therefore, the three-dimensional needle track model in the DICOM image generated based on the TPS report is not only the common needle track model constructed by aligning the DICOM image with the TPS report image, but also the needle track model is constructed by mapping the DICOM image to the TPS report image, and the needle track point positions are continuously optimized by utilizing the reinforcement learning model, so that the three-dimensional needle track model is constructed without deviation, and the result is more accurate.

Preferably, step S1 comprises the steps of:

step S11: acquiring DICOM image data and TPS report image data in a medical database;

step S12: performing data integration of corresponding image data on the DICOM image data and the TPS report image data to generate image data to be processed;

step S13: performing data cleaning treatment on the image data to be processed to generate cleaning image data;

step S14: performing image data noise reduction processing on the cleaning image data by using an image data noise reduction formula to generate noise reduction image data;

step S15: performing image contrast enhancement processing on the cleaning image data by utilizing histogram equalization to generate enhanced image data;

step S16: and carrying out standard image size adjustment on the noise reduction image data to generate standard image data.

The invention obtains DICOM image data and TPS report image data from a medical database. The DICOM image data comprises the structure and the anatomical information of the patient, the TPS report image data comprises the treatment plan information of the needle track, and the two data are combined together to form corresponding image data, so that a foundation is laid for the construction of a subsequent needle track model. The DICOM image data and TPS report image data are correspondingly integrated, data matching between the DICOM image data and TPS report image data is ensured, the integrated image data to be processed is a processed core data set, and structural information and treatment plan information are contained in the integrated image data, so that accurate input is provided for subsequent processing. Medical image data is generally affected by various factors, and the image data to be processed is subjected to data cleaning to remove unnecessary redundant data or error data in the data transmission process, so that the image data is cleaner and more reliable, which is helpful for reducing interference factors possibly introduced in subsequent processing and improving the quality and accuracy of the image data. Noise reduction of image data is an important step of image processing, and a noise reduction formula is used for processing the cleaning image data, so that noise in an image is reduced, and the definition and the visual effect of the image are improved. The noise-reduced image data is helpful for more accurately extracting features and information in subsequent steps. Histogram equalization is a common image enhancement technology, and can optimize contrast and brightness distribution of images, and through performing histogram equalization processing on the cleaned image data, visual effect and characteristic information of the images are enhanced, the enhanced image data is helpful for better displaying anatomical structures and needle track positions of patients, and richer information is provided for subsequent steps. The size and pixel resolution of the medical image data may be different depending on the equipment and the acquisition conditions, the noise reduction image data is subjected to standardized size adjustment, and the pixel size and resolution of the image are unified, so that the image deformation possibly caused by inconsistent sizes is eliminated, the accurate proportion of the image is maintained, and the stability and consistency of the subsequent steps are ensured.

Preferably, the image data denoising formula in step S14 is as follows:

where D (x, y) is represented by noise data with coordinates (x, y) in the image data after noise reduction, x is represented by an abscissa of the cleaning image data, y is represented by an ordinate of the cleaning image data, I (x, y) is represented by a pixel value with coordinates (x, y) in the cleaning image data, μ is represented by weight information of noise intensity, k is represented by filter intensity, F (x, y) is represented by an amount of noise data with coordinates (x, y) in the cleaning image data,expressed as smoothness for controlling gradient information, < >>The gradient is expressed as a gradient of pixel values of coordinates (x, y) in the cleaning image data, and τ is expressed as an abnormal adjustment value of a functional relation.

The invention utilizes an image data noise reduction formula which fully considers the abscissa x of the cleaning image data, the ordinate y of the cleaning image data, the pixel value I (x, y) with the coordinates of (x, y) in the cleaning image data, the weight information mu of noise intensity, the filtering intensity k, the noise data quantity F (x, y) with the coordinates of (x, y) in the cleaning image data and the smoothness degree for controlling gradient informationGradient +.f. of pixel value with coordinates (x, y) in the cleaning image data>And interactions between functions to form a functional relationship:

That is to say,the weight information of the noise intensity is expressed as a weight according to the noise ratio of the video, and when the noise intensity is relatively large, the weight information of the noise intensity is also larger; the filtering strength is used for balancing the trade-off between noise reduction and detail preservation, and the filtering treatment of noise is emphasized more by the larger filtering strength, but detail loss can be caused, and the proper filtering strength can reduce noise and keep image detail at the same time; the noise data amount with the coordinates (x, y) in the cleaning image data is a function related to the image data and is used for describing the distribution condition of noise in an image, and the noise can be estimated and noise reduced more accurately by analyzing the noise data amount with the coordinates (x, y) in the cleaning image data; cleaning gradients of pixel values with coordinates of (x, y) in image data for measuring edge characteristics of images, and maintaining definition of edge information in noise reduction process by introducing gradient informationA degree; the smoothness degree used for controlling the gradient information is used for adjusting gradient noise and ringing effect of the image. The noise level in the image data is reduced through the functional relation, so that the quality and the definition of the image are improved, the image data after noise reduction is more suitable for subsequent image processing and analysis, and the gradient information of the image data is considered, so that the details and the edge information of the image are kept in the noise reduction process, the details and the edge information are particularly important for reconstructing the three-dimensional needle model, and the constructed three-dimensional needle model is more accurate. And the abnormal adjustment value tau of the functional relation is utilized to adjust and correct the functional relation, so that the error influence caused by abnormal data or error items is reduced, noise data D (x, y) with coordinates of (x, y) in the image data after noise reduction is more accurately generated, and the accuracy and reliability of image data noise reduction processing of the cleaning image data are improved. Meanwhile, the weight information and the adjustment value in the formula can be adjusted according to actual conditions and are applied to different cleaning image data, so that the flexibility and applicability of the algorithm are improved.

Preferably, step S2 comprises the steps of:

step S21: performing DICOM image data extraction processing on the standard image data to generate optimized DICOM image data;

step S22: and performing skin-layer image extraction processing of the DICOM image on the optimized DICOM image data to generate skin-layer DICOM image data.

According to the invention, through carrying out DICOM image data extraction processing on standard image data, the interested region in the image, such as the position of the needle track and the surrounding tissue structure, can be obtained, the information irrelevant to the needle track can be removed in the extraction process, the information is focused on the key region, and the optimized DICOM image data is beneficial to reducing the calculation complexity of subsequent processing, and can accurately reflect the characteristics of the needle track and the surrounding tissue. In medical imaging, skin is an important reference structure in the needle track puncture process, and the position of a skin layer can be positioned and segmented by conducting skin layer image extraction processing on optimized DICOM image data, the skin layer DICOM image data is helpful for providing spatial information of skin, providing accurate position reference for subsequent needle track coordinate extraction and needle track model construction, reducing positioning deviation in needle track model construction, and improving the accuracy and visualization effect of the needle track model.

Preferably, step S3 comprises the steps of:

step S31: performing TPS report image data extraction on the standard image data to generate optimized TPS report image data;

step S32: performing needle track coordinate extraction of the TPS report image on the optimized TPS report image data to generate needle track coordinate data;

step S33: performing three-dimensional space mapping of needle track coordinates according to the needle track coordinate data to generate an initial needle track model;

step S34: and carrying out needle track relative position optimization verification of the initial needle track model on the initial needle track model to generate a needle track model.

According to the invention, TPS report image data extraction is carried out on standard image data to generate optimized TPS report image data in medical images, TPS report contains planning information of needle tracks including positions, directions, lengths and the like, and the optimized TPS report image data can be obtained by carrying out TPS report image data extraction on the standard image data, contains more accurate treatment planning information, and provides reliable basis for subsequent needle track coordinate extraction and three-dimensional space mapping. In the optimized TPS report image data, the extraction of needle track coordinates is performed for the needle track position information, the needle track coordinate data describes the spatial position of the needle track in the image, such as coordinate values in the anatomy of the patient, and the coordinate information is used for the subsequent three-dimensional space mapping to construct a needle track model. The needle track coordinate data is subjected to three-dimensional space mapping, two-dimensional coordinates of TPS report images are converted into actual positions in the three-dimensional space, and an initial needle track model, namely the approximate shape of the needle track in the anatomy structure of a patient, is generated through space mapping and is used as a basis for subsequent optimization verification. The initial needle track model is optimized and verified, the accuracy of the needle track model and the accuracy of the relative position are ensured, the verification process can be realized by comparing the optimized DICOM image data with the skin-layer DICOM image data, and the accuracy and the reliability of the needle track model are further improved by verifying and finding and correcting errors possibly existing.

Preferably, step S34 includes the steps of:

step S341: performing peripheral contour extraction of the image on the optimized TPS report image data by utilizing an edge detection technology to generate image peripheral contour data;

step S342: calculating the relative positions of the needle track coordinates and the peripheral contour according to the needle track coordinate data and the image peripheral contour data, and generating needle track optimization parameters;

step S343: and carrying out optimization verification on the needle track relative position of the initial needle track model according to the needle track optimization parameters to generate a needle track model.

According to the invention, the optimized TPS report image data is processed by utilizing the edge detection technology, the peripheral outline of the image is extracted, the edge detection technology can effectively identify the boundaries between different structures and tissues in the image, the peripheral outline data describes the overall shape and boundary information of the image, and the information is very important for the subsequent needle track model optimization verification. The relative position of the needle track and the peripheral outline are obtained by comparing and calculating the needle track coordinate data and the peripheral outline data of the image, and the relative position information is used for calculating needle track optimization parameters which describe the relative position and direction of the needle track in the image, such as the distance, angle and the like of the needle track and the skin layer. The needle optimization parameters will be used for subsequent optimization verification of the needle model. According to the needle track optimization parameters obtained through calculation, the initial needle track model is optimized and verified, the needle track model is better matched with the peripheral outline of the image and the needle track coordinate data through adjustment and optimization of the model, errors possibly introduced by the construction of the initial model can be eliminated in the optimization process, the accuracy and the visualization effect of the needle track model are improved, the needle track model after optimization has more accurate position and shape information, and a reliable data basis is provided for the three-dimensional needle track model reconstruction and the deep reinforcement learning of the subsequent image.

Preferably, step S4 comprises the steps of:

step S41: performing space mapping alignment on the skin layer DICOM image data according to the needle track model to generate DICOM image mapping data;

step S42: performing needle track deviation distance calculation on the DICOM image mapping data by using a needle track deviation distance calculation formula to generate a needle track deviation distance parameter;

step S43: performing mapping position deviation distance adjustment on the DICOM image mapping data according to the needle track deviation distance parameter to generate optimized DICOM image mapping data;

step S44: and reconstructing the three-dimensional needle track model of the image by using the optimized DICOM image mapping data to generate an image needle track model.

The invention uses the previously constructed needle track model to carry out space mapping alignment on the DICOM image data of the skin layer, ensures the consistency of the needle track model and the anatomical structure of a patient by aligning the needle track model with the DICOM image data, thereby accurately representing the position and the shape of the needle track in the body of the patient, and the alignment process is the basis of reconstructing the image needle track model and provides accurate reference for the subsequent calculation of the needle track deviation distance. The DICOM image mapping data is processed by using a needle track deviation distance calculation formula, a deviation distance parameter between a needle track and the image data is calculated, and the needle track deviation distance parameter describes the difference between a needle track model and the DICOM image data, such as the positioning deviation, the rotation angle and the like of the needle track. The calculation of the parameters provides necessary adjustment basis for the subsequent reconstruction of the image needle track model. According to the needle track deviation distance parameter, deviation adjustment of the mapping position is carried out on the DICOM image mapping data, and the position of the image needle track model is adjusted to enable the DICOM image mapping data to be better matched with the DICOM image data, so that optimal reconstruction of the image needle track model is achieved, the accuracy and precision of the needle track model can be improved in the adjustment process, and errors caused by the mapping deviation are reduced. And (3) reconstructing the three-dimensional needle track model of the image of the previous needle track model by utilizing the optimized DICOM image mapping data, mapping the needle track model into the optimized DICOM image mapping data through reconstruction, and generating a final image needle track model, wherein the needle track model is tightly combined with the optimized DICOM image mapping data in the reconstruction process, so that the needle track model can show the real position and shape in the image, and more accurate reference is provided for medical image processing and clinical application.

Preferably, step S41 comprises the steps of:

step S411: performing needle track position matching marking on the skin layer DICOM image data according to the needle track coordinate positions to generate marking positions of the skin layer DICOM image data;

step S412: and performing spatial mapping on the skin-layer DICOM image data according to the needle track model, and performing image mapping data alignment according to the mark position to generate DICOM image mapping data.

According to the invention, the needle track position matching marking is carried out on the DICOM image data of the skin layer according to the needle track coordinate data, and the region corresponding to the needle track position in the image can be determined by carrying out the needle track coordinate position matching marking on the image, and the marking process provides a key reference for the subsequent image space mapping alignment. According to the previously constructed needle track model, the skin layer DICOM image data is spatially mapped, the needle track model is aligned with the DICOM image data, and according to the needle track position marked before, the image is aligned with the mapping data, so that the needle track position and the shape in the image are matched with the needle track model, and the alignment process provides an accurate data basis for the subsequent image needle track model reconstruction.

Preferably, the lane offset distance calculation formula in step S42 is as follows:

Where K is a track deviation distance parameter, f is an initial adjustment value of the track position, c is a relative distance between the boundary of the DICOM image map data and the track coordinate position, ω is an angle of the puncture track, a is a boundary distance difference between the boundary of the DICOM image map data and the track model, b is a distance difference between the mark position of the DICOM image map data and the track coordinate position, p is a deviation angle between the mark position of the DICOM image map data and the track coordinate position, and δ is an abnormal adjustment value of the track deviation distance parameter.

The invention utilizes a needle track deviation distance calculation formula which fully considers the interaction relation among an initial adjustment value f of a needle track position, a relative distance c between a boundary of DICOM image mapping data and a needle track coordinate position, an angle omega of a puncture needle track, a boundary distance difference a between a boundary of DICOM image mapping data and a needle track model, a distance difference b between a mark position of DICOM image mapping data and the needle track coordinate position, a deviation angle p between the mark position of DICOM image mapping data and the needle track coordinate position and a function to form a functional relation:

That is to say,the initial adjustment value of the needle track position is used for determining an initial adjustment point of the needle track position, and the accurate position of the needle track position can be found by adjusting the initial adjustment value of the needle track position; the relative distance between the boundary of the DICOM image mapping data and the needle track coordinate position is used for considering the boundary condition of the DICOM image mapping data and factors influencing the needle track position; the angle of the puncture needle track is used for helping to measure the direction and the inclination degree of the needle track position in the DICOM image; the difference value of the boundary distance between the boundary of the DICOM image mapping data and the boundary of the needle track model is used for further controlling the position of the needle track model in the DICOM image; the distance difference between the marking position of the DICOM image mapping data and the needle track coordinate position is used for considering the deviation between the marking position of the DICOM image and the real needle track position; the deviation angle of the marking position of the DICOM image mapping data and the needle track coordinate position is used for helping to measure the deviation degree of the marking position. The needle track position can be optimized according to the initial adjustment value of the needle track position through the functional relation, so that the boundary of the needle track model and DICOM image data is more consistent, accurate positioning of the needle track model is facilitated, and the accuracy of the needle track position is improved; by considering the relative distance between the boundary of the DICOM image mapping data and the needle track position and the boundary of the DICOM image mapping data and the needle track model The distance difference value is used for determining the boundary condition of the DICOM image mapping data, so that the position of the needle track model in the DICOM image is better controlled; the distance difference and the deviation angle between the marking position of the DICOM image mapping data and the needle track coordinate position are used for considering the position difference between the marking position in the DICOM image and the needle track coordinate, so that the accuracy of the needle track model is further optimized, and the optimization and the accurate positioning of the needle track position are realized. And the functional relation is adjusted and corrected by utilizing the abnormal adjustment value delta of the needle track deviation distance parameter, so that the error influence caused by abnormal data or error items is reduced, the needle track deviation distance parameter K is more accurately generated, and the accuracy and the reliability of needle track deviation distance calculation on DICOM image mapping data are improved. Meanwhile, the adjustment value in the formula can be adjusted according to actual conditions and is applied to DICOM image mapping data, so that the flexibility and applicability of the algorithm are improved.

Preferably, step S5 comprises the steps of:

step S51: establishing a mapping relation of image needle track model learning reinforcement by using a deep reinforcement learning model, and generating an initial image needle track optimization mathematical model;

step S52: acquiring a historical image needle track alignment model;

Step S53: performing rewarding function design of the initial image needle track optimization mathematical model according to the historical image needle track alignment model, and generating a rewarding function of the initial image needle track optimization mathematical model;

step S54: performing model training on the initial image needle track optimization mathematical model by using the historical image needle track alignment model to generate an image needle track optimization mathematical model;

step S55: and transmitting the image needle track model to an image needle track optimizing mathematical model for model test, and optimizing the image needle track model according to the rewarding function, so as to generate an optimized image needle track model.

The invention establishes a mapping relation of image needle track model learning reinforcement by utilizing a deep reinforcement learning model, wherein the deep reinforcement learning model can select an optimal strategy according to the current state (namely the initial state of the image needle track model) and the action (namely the optimization step), so that the optimization of the image needle track model is realized, an initial image needle track optimization mathematical model is generated by learning the mapping relation, and a foundation is provided for subsequent model training and optimization. Historical image track alignment models are obtained, wherein the historical models are optimization models which are trained and verified before, experience and knowledge of image track optimization are provided, and the historical models are used for design and model training of subsequent reward functions. According to the historical image track alignment model, a reward function of the initial image track optimization mathematical model is designed, the reward function is used for evaluating the effect of the image track model in the optimization process, rewards or punishments are given according to the optimization result, and the deep reinforcement learning model can be guided to learn and optimize the image track model better through the reasonable design of the reward function. The historical image needle track alignment model is utilized to carry out model training on the initial image needle track optimization mathematical model, the image needle track model can be continuously adjusted and optimized according to the rewarding function through training the deep reinforcement learning model, so that the model can better adapt to the image needle track optimization requirements under different conditions, and the trained image needle track optimization mathematical model has stronger optimization capability. The image needle track model is transmitted to a trained image needle track optimizing mathematical model for model test, the image needle track model can be optimized according to the current state and action by the model according to the rewarding function designed before, the optimized image needle track model is finally generated through repeated optimization processes, and the optimization process can enable the image needle track model to be more attached to real patient anatomy structure and image data, and accuracy and reliability of the model are improved.

The method has the advantages that the DICOM image data in the medical database and TPS report image data are correspondingly integrated, data cleaning, noise reduction, contrast enhancement and other treatments are carried out, the accuracy and consistency of the data are ensured, the noise and interference of the data are reduced, and high-quality standard image data are provided for subsequent steps. Extracting the skin layer image in the DICOM image data determines the boundary of the skin layer and other tissue structures, providing more accurate edge information for subsequent needle model construction, which helps to ensure accurate positioning of the needle and avoid penetration of the needle through the skin. By performing three-dimensional space mapping and optimization of needle track coordinates on TPS report image data, an initial needle track model is generated, accurate correspondence of the needle track model and a patient anatomy structure is ensured, and the needle track model reflects the position and the shape in the DICOM image more truly, so that the visualization effect and the practical application precision of the needle track model are improved. The image needle track model can be learned and optimized according to the current state and action, model training is carried out according to the historical image needle track alignment model, and optimization is carried out according to a reward function, and the optimization process can continuously adjust and improve the image needle track model, so that the image needle track model can be better adapted to anatomical structures and image data of different patients, and accuracy and reliability of the needle track model are improved.

Drawings

FIG. 1 is a flowchart illustrating a method for generating a three-dimensional needle track model in DICOM images based on TPS report according to the present invention;

FIG. 2 is a flowchart illustrating the detailed implementation of step S3 in FIG. 1;

FIG. 3 is a flowchart illustrating the detailed implementation of step S34 in FIG. 2;

FIG. 4 is a flowchart illustrating the detailed implementation of step S4 in FIG. 1;

FIG. 5 is a flowchart illustrating the detailed implementation of step S5 in FIG. 1;

the achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.

Detailed Description

The following is a clear and complete description of the technical method of the present patent in conjunction with the accompanying drawings, and it is evident that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, are intended to fall within the scope of the present invention.

Furthermore, the drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus a repetitive description thereof will be omitted. Some of the block diagrams shown in the figures are functional entities and do not necessarily correspond to physically or logically separate entities. The functional entities may be implemented in software or in one or more hardware modules or integrated circuits or in different networks and/or processor methods and/or microcontroller methods.

It will be understood that, although the terms "first," "second," etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another element. For example, a first element could be termed a second element, and, similarly, a second element could be termed a first element, without departing from the scope of example embodiments. The term "and/or" as used herein includes any and all combinations of one or more of the associated listed items.

In order to achieve the above objective, referring to fig. 1 to 5, the present invention provides a method for generating a three-dimensional needle track model in DICOM images based on TPS reports, comprising the following steps:

In the embodiment of the present invention, as described with reference to fig. 1, a flow chart of steps of a method for generating a three-dimensional needle track model in DICOM images based on TPS reports according to the present invention is shown, and in the embodiment, the method for generating the three-dimensional needle track model in DICOM images based on TPS reports includes the following steps:

in an embodiment of the invention, chest-related data of a patient is obtained from a medical database. DICOM image data of the patient is included, including CT scan and MRI images, and chest treatment reports generated by a Treatment Planning System (TPS). The integration of DICOM image data with TPS report image data corresponding to the image data may be achieved by matching patient identity information with examination time, etc. And then, carrying out data preprocessing on the integrated image data, removing noise and artifacts in the image, carrying out image registration and standardization to generate high-quality standard image data, and preparing for the subsequent steps.

in the embodiment of the invention, the extraction processing of the skin layer image is performed on the standard image data, and the image processing technology such as image threshold segmentation and morphological operation is utilized to extract the skin layer region from the standard image data, so as to obtain the skin layer DICOM image data, wherein the skin layer DICOM image data only contains the information of skin tissues.

in the embodiment of the invention, the needle track coordinate information of the TPS report image is extracted from the standard image data, which can be realized through image segmentation and a target detection algorithm, after the needle track coordinate data is obtained, the needle track coordinate is mapped and optimized in the standard image data in a three-dimensional space by utilizing a computer graphics and image registration technology, so as to obtain a needle track model, and the needle track model can accurately represent the actual needle track position and shape of a patient.

In the embodiment of the invention, the DICOM image data of the skin layer is aligned and optimized by using the needle track model generated before in a space mapping way. The method can be realized through a registration algorithm and a space transformation technology, the optimized DICOM image mapping data is obtained, corresponding information of a skin layer and a needle track model is contained, the optimized DICOM image mapping data is utilized to reconstruct a three-dimensional needle track model of an image of the needle track model, and the needle track model and the DICOM image data are fused together to generate an image needle track model.

In the embodiment of the invention, the generated image needle track model is input into the deep reinforcement learning model, the deep reinforcement learning model learns and optimizes the image needle track model, in the learning process, the model is adjusted and optimized according to the corresponding relation between the position and the shape of the needle track and DICOM image data, and the optimized image needle track model is obtained through the training of reinforcement learning, thereby being more in accordance with the characteristics of the DICOM image data and having higher accuracy and adaptability.

Preferably, step S1 comprises the steps of:

In an embodiment of the invention, chest-related data of a patient is obtained from a medical database. DICOM image data of the patient is included, including CT scan and MRI images, and chest treatment reports generated by a Treatment Planning System (TPS). The DICOM image data and the TPS report image data are integrated correspondingly, the DICOM image data and the TPS report image data are associated by matching the identity information and the checking time of the patient, the DICOM image data and the TPS report image data are ensured to be from the same patient and are data of the same treatment plan, the integrated data are generated into image data to be processed, and the image data to be processed are further processed and optimized in the follow-up steps. The method aims at carrying out data cleaning treatment on the image data to be processed, and aims at removing noise and artifacts in the image, smoothing the image and removing abnormal points, and the method is realized through an abnormal value detection method of an image processing technology, so that the cleaned image data is cleaner and more reliable, and a better foundation is provided for subsequent image processing and analysis. The noise reduction processing is carried out on the cleaned image data by utilizing the image data noise reduction formula, the noise intensity, the filtering intensity, the gradient information and other factors in the image data are considered in the formula, the noise in the image data is reduced by reasonably adjusting the parameters in the formula, the details and the edge information of the image are kept, the image data after noise reduction are clearer and more accurate, and the subsequent image analysis and needle track model construction are facilitated. The image contrast enhancement processing is carried out on the cleaned image data, the image contrast enhancement is realized through image processing technologies such as histogram equalization and the like, different tissue structures and needle track information in the image can be more prominent and clear, subsequent image analysis and model construction are facilitated, and the enhanced image data is easier to observe and understand. The standard image size adjustment is carried out on the noise reduction image data, the standard image size adjustment is realized through an interpolation algorithm and a space transformation technology, the image data can be unified to the same space size and pixel spacing, so that direct comparison and registration can be carried out between different images, and the generation of the standard image data provides a consistent reference for subsequent image processing and needle track model construction.

Preferably, the image data denoising formula in step S14 is as follows:

The invention utilizes an image data to reduce noiseA formula which fully considers the abscissa x of the cleaning image data, the ordinate y of the cleaning image data, the pixel value I (x, y) with the coordinates of (x, y) in the cleaning image data, the weight information mu of noise intensity, the filtering intensity k, the noise data quantity F (x, y) with the coordinates of (x, y) in the cleaning image data and the smoothness degree for controlling gradient informationGradient +.f. of pixel value with coordinates (x, y) in the cleaning image data>And interactions between functions to form a functional relationship:

That is to say,the weight information of the noise intensity is expressed as a weight according to the noise ratio of the video, and when the noise intensity is relatively large, the weight information of the noise intensity is also larger; the filtering strength is used for balancing the trade-off between noise reduction and detail preservation, and the filtering treatment of noise is emphasized more by the larger filtering strength, but detail loss can be caused, and the proper filtering strength can reduce noise and keep image detail at the same time; the noise data amount with the coordinates (x, y) in the cleaning image data is a function related to the image data and is used for describing the distribution condition of noise in an image, and the noise can be estimated and noise reduced more accurately by analyzing the noise data amount with the coordinates (x, y) in the cleaning image data; the gradient of the pixel value with the coordinates of (x, y) in the image data is cleaned to measure the edge characteristics of the image, and the definition of the edge information is maintained in the noise reduction process by introducing the gradient information; the smoothness degree used for controlling the gradient information is used for adjusting gradient noise and ringing effect of the image. The noise level in the image data is reduced by the functional relation, so that the quality and definition of the image are improved, the image data after noise reduction is more suitable for subsequent image processing and analysis, and the gradient information of the image data is considered, so that the details and the edge information of the image are kept in the noise reduction process The method is particularly important for reconstructing the three-dimensional needle track model, and the constructed three-dimensional needle track model can be more accurate. And the abnormal adjustment value tau of the functional relation is utilized to adjust and correct the functional relation, so that the error influence caused by abnormal data or error items is reduced, noise data D (x, y) with coordinates of (x, y) in the image data after noise reduction is more accurately generated, and the accuracy and reliability of image data noise reduction processing of the cleaning image data are improved. Meanwhile, the weight information and the adjustment value in the formula can be adjusted according to actual conditions and are applied to different cleaning image data, so that the flexibility and applicability of the algorithm are improved.

Preferably, step S2 comprises the steps of:

In the embodiment of the invention, the standard image data is subjected to DICOM image data extraction processing, and the DICOM image data subjected to data preprocessing is extracted to obtain the optimized DICOM image data. The optimized DICOM image data comprises information of skin layers and other tissue structures, in order to further extract the skin layer information, an image segmentation algorithm is utilized to extract the skin layer image of the DICOM image from the optimized DICOM image data, the skin region in the image can be separated according to characteristics of pixel values or characteristics of textures, shapes and the like of regions by the algorithm, the skin layer DICOM image data is obtained, only the skin tissue information is contained, and the method can be used for subsequent needle track planning and treatment planning.

Preferably, step S3 comprises the steps of:

As an example of the present invention, referring to fig. 2, a detailed implementation step flow diagram of step S3 in fig. 1 is shown, where step S3 includes:

in the embodiment of the invention, the standard image data is extracted into TPS report image data, the TPS report image data subjected to data preprocessing is extracted, and information related to needle track treatment, such as the type, the number and the position coordinates of the needle track, is extracted from the TPS report, so that optimized TPS report image data is obtained, wherein the optimized TPS report image data only comprises the information related to the needle track treatment.

in the embodiment of the invention, needle track coordinate extraction of TPS report images is performed on optimized TPS report image data, and in order to further extract needle track coordinate data, an image processing and target detection algorithm is needed. The algorithms can automatically identify and extract the position coordinates of the needle tracks according to the characteristics or structures in the images, so that the needle track coordinate data are obtained, the position information of all the needle tracks is contained, and a foundation is provided for the subsequent needle track model construction.

in the embodiment of the invention, the three-dimensional space mapping of the needle track coordinates is performed according to the needle track coordinate data to generate an initial needle track model, and the needle track coordinate data of a patient is assumed to be obtained, wherein the needle track coordinate data contains the position information of all needle tracks. By such processing, an initial needle track model is obtained which accurately represents the position and orientation of the needle track in the patient's body.

In the embodiment of the invention, the optimization verification of the needle track relative position of the initial needle track model is realized through an edge detection technology and an image registration algorithm, the edge contour related to the needle track is extracted from the optimized TPS report image data, and is compared with the initial needle track model, and the position and the shape of the needle track model can be adjusted through the optimization verification so as to be more in line with the actual condition in DICOM image data, thereby obtaining the final needle track model which can more accurately represent the actual needle track position and the shape of a patient.

Preferably, step S34 includes the steps of:

As an example of the present invention, referring to fig. 3, a detailed implementation step flow diagram of step S34 in fig. 2 is shown, where step S34 includes:

in the embodiment of the invention, the optimized TPS report image data is processed by utilizing an edge detection technology to extract the peripheral outline of the image, the edge detection is a common image processing technology, the edge of an object in the image can be found by identifying the jump or gradient change of the pixel value, the peripheral outline in the image is found to better define the needle track area, and the peripheral outline data of the image is obtained and is used for calculating the needle track optimization parameters in the subsequent steps.

in the embodiment of the invention, the relative positions of the needle track coordinate and the peripheral contour data are calculated according to the needle track coordinate data and the peripheral contour data of the image to generate the needle track optimization parameters, and the needle track coordinate data and the peripheral contour data of the image are determined, so that the information of the relative positions is obtained by calculating the distance and the angle between the needle track coordinate and the peripheral contour data, and can be used for the subsequent needle track model optimization, and the position and the shape of the needle track model can be better adjusted by calculating the needle track optimization parameters to more accurately match the actual condition of a patient.

In the embodiment of the invention, the position and the shape of the needle track model are adjusted according to the needle track optimization parameters, so that the needle track model better accords with the actual condition of a patient, the needle track model is continuously adjusted and optimized through optimization verification until the optimal matching effect is achieved, and finally the needle track model is obtained, and the position and the direction of the needle track in the body of the patient are accurately represented.

Preferably, step S4 comprises the steps of:

As an example of the present invention, referring to fig. 4, a detailed implementation step flow diagram of step S4 in fig. 1 is shown, where step S4 includes:

in the embodiment of the invention, the DICOM image data of the skin layer is spatially mapped and aligned according to the generated needle track model, the DICOM image data of the skin layer is required to be aligned with the needle track model so as to ensure that the position and the direction of the needle track model are consistent with the actual condition of a patient, and the DICOM image data of the skin layer is mapped into the coordinate system of the needle track model by using a spatial transformation technology to obtain DICOM image mapping data which represents the spatial position of the DICOM image data of the skin layer in the coordinate system of the needle track model.

in the embodiment of the invention, the needle track deviation distance calculation formula is utilized to calculate the needle track deviation distance of the DICOM image mapping data so as to generate the needle track deviation distance parameter, and the needle track deviation distance parameter is used for measuring the deviation degree between the DICOM image mapping data and the needle track model. The lane departure distance parameter is calculated using a calculation formula, which may involve pixel values of the image, gradient information, and geometric features of the lane model, by which the accuracy of DICOM image map data and the differences from the lane model are evaluated.

in the embodiment of the invention, the needle track deviation distance parameter tells us which positions in the DICOM image mapping data deviate from the positions of the needle track model, and the needle track positions in the images can be mapped into the needle track model coordinate system more accurately by adjusting the DICOM image mapping data according to the needle track deviation distance parameter, so that the accuracy and precision of the needle track model are improved, and the optimized DICOM image mapping data are generated.

In the embodiment of the invention, the DICOM image mapping data is combined with the needle track model, the needle track shape in the image is reconstructed in the three-dimensional space, and the final image needle track model is obtained, wherein the image needle track model comprises the appearance and the needle track position of a treatment area of a user, accurately represents the position and the shape of the needle track in the body of the user, and generates the image needle track model.

Preferably, step S41 comprises the steps of:

In the embodiment of the invention, the skin layer DICOM image data is marked according to the needle track coordinate data of the TPS report image, the position of the skin layer DICOM image data corresponding to the needle track coordinate data is marked, the region corresponding to the needle track position in the skin layer DICOM image data is marked according to the needle track coordinate data, and the marking is realized by drawing a special mark in the DICOM image or giving a specific numerical value in the pixel value, and the marked region indicates the position of the puncture needle track in the skin layer. The skin layer DICOM image is mapped into the information in the needle track model to align the needle track model with the DICOM image, a marking area corresponding to the needle track position marked before is found in the DICOM image, and by aligning the needle track model mapping data with the marking area, DICOM image mapping data can be generated, which is expressed as a preliminary position mapped in the needle track model.

That is to say,the initial adjustment value of the needle track position is used for determining an initial adjustment point of the needle track position, and the accurate position of the needle track position can be found by adjusting the initial adjustment value of the needle track position; the relative distance between the boundary of the DICOM image mapping data and the needle track coordinate position is used for considering the boundary condition of the DICOM image mapping data and factors influencing the needle track position; the angle of the puncture needle track is used for helping to measure the direction and the inclination degree of the needle track position in the DICOM image; the difference value of the boundary distance between the boundary of the DICOM image mapping data and the boundary of the needle track model is used for further controlling the position of the needle track model in the DICOM image; the distance difference between the marking position of the DICOM image mapping data and the needle track coordinate position is used for considering the deviation between the marking position of the DICOM image and the real needle track position; d (D)The deviation angle of the mark position and the needle track coordinate position of the ICOM image mapping data is used for helping to measure the deviation degree of the mark position. The needle track position can be optimized according to the initial adjustment value of the needle track position through the functional relation, so that the boundary of the needle track model and DICOM image data is more consistent, accurate positioning of the needle track model is facilitated, and the accuracy of the needle track position is improved; determining the boundary condition of the DICOM image mapping data by considering the relative distance between the boundary of the DICOM image mapping data and the needle track position and the boundary distance difference between the boundary of the DICOM image mapping data and the needle track model, thereby better controlling the position of the needle track model in the DICOM image; the distance difference and the deviation angle between the marking position of the DICOM image mapping data and the needle track coordinate position are used for considering the position difference between the marking position in the DICOM image and the needle track coordinate, so that the accuracy of the needle track model is further optimized, and the optimization and the accurate positioning of the needle track position are realized. And the functional relation is adjusted and corrected by utilizing the abnormal adjustment value delta of the needle track deviation distance parameter, so that the error influence caused by abnormal data or error items is reduced, the needle track deviation distance parameter K is more accurately generated, and the accuracy and the reliability of needle track deviation distance calculation on DICOM image mapping data are improved. Meanwhile, the adjustment value in the formula can be adjusted according to actual conditions and is applied to DICOM image mapping data, so that the flexibility and applicability of the algorithm are improved.

Preferably, step S5 comprises the steps of:

step S52: acquiring a historical image needle track alignment model;

As an example of the present invention, referring to fig. 5, a detailed implementation step flow diagram of step S5 in fig. 1 is shown, where step S5 includes:

in the embodiment of the invention, the depth reinforcement learning model is utilized to establish the image needle track model to learn the reinforced mapping relation, and the mapping relation can be used for carrying out needle track optimization on the transmitted image needle track model so as to ensure that the needle track position of the image needle track model is more accurate.

Step S52: acquiring a historical image needle track alignment model;

in the embodiment of the invention, the related information is obtained from a history record or an existing image track alignment model. These historical models may be the result of previous image needle model optimizations or expert validated and approved models to evaluate the performance of our initial image needle optimization mathematical model.

In the embodiment of the invention, the rewarding function of the initial image track optimization mathematical model is designed according to the historical image track alignment model. The reward function is an important part in the deep reinforcement learning model, and is used for evaluating the action of the model and giving corresponding rewards or punishments, and hopes that the model can be more approximate to and meet the result of the historical video track alignment model when optimizing the video track model by designing the proper reward function, for example, the reward function is used for judging whether the track position of the video track model is correct or incorrect, and if the track position is incorrectly deducted, the track position is correctly added.

in the embodiment of the invention, the initial image needle track optimization mathematical model is subjected to model training by using the historical image needle track alignment model, and the initial model is gradually optimized through training, so that the result of the historical model is more similar to and met when the image needle track model is optimized, and after the training is finished, the image needle track optimization mathematical model verified by the historical model is obtained.

In the embodiment of the invention, the initial image needle track optimization mathematical model is transmitted to the image needle track optimization mathematical model for model test, in the model test process, the performance of the model is evaluated by using the previously designed reward function, the image needle track model is optimized according to the feedback of the reward function, and the optimized image needle track model is finally obtained through continuous iteration and optimization, so that the model can better meet the result of the historical model and has higher accuracy and stability, thereby generating the optimized image needle track model.

The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.

The foregoing is only a specific embodiment of the invention to enable those skilled in the art to understand or practice the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for generating a three-dimensional needle track model in DICOM images based on TPS reports, comprising the steps of:

2. The method of generating a three-dimensional needle track model in DICOM images based on TPS reports according to claim 1, wherein step S1 comprises the steps of:

3. The method of generating a three-dimensional needle track model in DICOM images based on TPS reports of claim 2, wherein the image data denoising formula in step S14 is as follows:

wherein D (x, y) is noise data with coordinates (x, y) in the image data after noise reduction,x is represented as an abscissa of the cleaning image data, y is represented as an ordinate of the cleaning image data, I (x, y) is represented as a pixel value of the cleaning image data having coordinates (x, y), μ is represented as weight information of noise intensity, k is represented as filter intensity, F (x, y) is represented as an amount of noise data of the cleaning image data having coordinates (x, y),expressed as smoothness for controlling gradient information, < >>The gradient is expressed as a gradient of pixel values of coordinates (x, y) in the cleaning image data, and τ is expressed as an abnormal adjustment value of a functional relation.

4. The method of generating a three-dimensional needle tract model in DICOM images based on TPS reports of claim 3 wherein step S2 comprises the steps of:

5. The method of generating a three-dimensional needle track model in DICOM images based on TPS reports of claim 4, wherein step S3 comprises the steps of:

6. The method of generating a three-dimensional needle track model in DICOM images based on TPS reports of claim 5, wherein step S34 comprises the steps of:

7. The method of generating a three-dimensional needle tract model in DICOM images based on TPS reports of claim 6, wherein step S4 comprises the steps of:

8. The method of generating a three-dimensional needle track model in DICOM images based on TPS reports of claim 7, wherein step S41 comprises the steps of:

9. The method of generating a three-dimensional needle track model in DICOM images based on TPS reports of claim 8, wherein the needle track deviation distance calculation formula in step S42 is as follows:

10. The method of generating a three-dimensional needle track model in DICOM images based on TPS reports of claim 9, wherein step S5 comprises the steps of:

step S52: acquiring a historical image needle track alignment model;