Lossless compression method for medical image data

Abstract

The invention relates to the technical field of medical data compression, and discloses a lossless compression method for medical image data, which comprises the following steps of: 1, acquiring three-dimensional medical image data and ray beam path data in a radiotherapy plan, and segmenting the three-dimensional medical image data into a plurality of voxel blocks; and performing clinical importance grade labeling on the voxel blocks according to ray beam path data. According to the technical scheme, physiological motion signals of a patient are obtained to generate a displacement time table, motion compensation is executed based on the displacement time table to generate a prediction image, and finally reversible differential storage is conducted on a real image and the prediction image, so that the purposes of fully utilizing time sequence redundancy of three-dimensional medical image data, remarkably improving the compression ratio and improving the image quality are achieved. And compared with the technical scheme that compression is adopted piece by piece and time relevance is not utilized in the prior art, the problems that the compression efficiency is low and excessive storage resources are occupied are solved.

Description

Technical Field

[0001] This invention relates to the field of medical data compression technology, specifically a lossless compression method for medical image data. Background Technology

[0002] In modern precision radiotherapy, image-guided techniques, especially four-dimensional computed tomography (CT) and cone-beam computed tomography (CBCT), have become key tools for accurately monitoring and locating tumor and surrounding organ movement. These techniques generate large-volume three-dimensional imaging data with time-series data at each treatment fraction, providing a basis for the development, validation, and adaptive adjustment of radiotherapy plans.

[0003] The massive amount of image data presents the following serious challenges:

[0004] When processing a large number of multiple three-dimensional image sequences generated during radiotherapy, the image-by-image compression method is often used, which fails to effectively utilize the massive data redundancy between three-dimensional space and time series, resulting in low compression efficiency and excessive storage resources.

[0005] The pixel values ​​in medical images are related to the dose calculation results. However, if the existing compression methods cannot guarantee the complete reversibility of the data, they will destroy the accuracy of the pixel values, causing subsequent dose tracing and clinical reassessment to lose a reliable data foundation.

[0006] The workflow of online adaptive radiotherapy is highly sensitive to time delays. Existing compression schemes often require complete decompression of the entire image roll when retrieving data. This time-consuming global decompression operation hinders the rapid retrieval and recalculation of data in the planning system, making it difficult to meet the efficiency requirements of same-day clinical planning and execution.

[0007] Existing data encapsulation methods lack effective voxel-level consistency verification mechanisms and block data traceability functions when exchanging data across different medical information systems, making it difficult to ensure the compliance and reliability of dosage assessment during data transfer and reproduction.

[0008] To address the aforementioned issues, this invention proposes a lossless compression method for medical imaging data. This method acquires and analyzes patient physiological motion signals, performs motion compensation to generate predicted images, calculates reversible differential data between the actual and predicted images, and performs adaptive lossless compression on the reversible differential data based on the clinical importance level of the X-ray beam path data. Simultaneously, it constructs a beam path priority directory for rapid location of high-importance voxel blocks and performs sampling verification before storage. This significantly improves compression efficiency while ensuring complete lossless and reversible data compression, and provides rapid location and retrieval of imaging data and consistency verification functions for clinical workflows. Summary of the Invention

[0009] To address the shortcomings of existing technologies, this invention provides a lossless compression method for medical image data, thereby resolving the problems mentioned in the background section.

[0010] To achieve the above objectives, the present invention provides the following technical solution: a lossless compression method for medical image data, comprising:

[0011] Step 1: Acquire 3D medical image data and X-ray beam path data from the radiotherapy plan; divide the 3D medical image data into many voxel blocks; and label the voxel blocks with clinical importance levels based on the X-ray beam path data.

[0012] Step 2: Acquire patient physiological motion signals during 3D medical image data acquisition; analyze physiological motion signals to generate a displacement timetable recording the expected movement of voxel blocks;

[0013] Step 3: Obtain the 3D medical image from the previous moment as a reference image; perform motion compensation processing on the reference image according to the displacement time table to generate the predicted image for the current moment;

[0014] Step 4: Calculate the reversible difference data between the current 3D medical image data and the predicted image; based on the clinical importance level, use an adaptive strategy to perform lossless compression and storage of the reversible difference data;

[0015] Step 5: Based on the clinical importance level labeling, select high-importance voxel blocks; construct a bundle priority directory to record the mapping relationship between the positions of high-importance voxel blocks and the offsets within the compressed data;

[0016] Step 6: Before the compression and storage are completed, extract voxel blocks along the X-ray beam data path for sampling and verification; when the sampling and verification results show an anomaly, automatically increase the storage priority of the corresponding voxel block and record the anomaly information.

[0017] Preferably, the clinical importance level labeling includes: labeling voxel blocks located inside the X-ray beam path data as high importance; and labeling voxel blocks far from the X-ray beam path data as low importance.

[0018] Preferably, the motion compensation process includes performing translation or deformation pull-back operations on the voxel blocks in the reference image according to the movement parameters recorded in the displacement time table.

[0019] Preferably, the beam path priority directory is used to directly extract the high importance level voxel blocks according to the directory when performing dose recalculation without decompressing the entire roll of image data, thereby achieving rapid location and retrieval;

[0020] The automatic elevation of the storage priority of the corresponding voxel block includes: when the sampling verification result shows an anomaly, stopping the storage of the reversible differential data for the voxel block and directly storing the original uncompressed data of the voxel block instead.

[0021] Preferably, step one further includes:

[0022] Sub-step Acquire the three-dimensional medical image data containing time series data to be compressed. Simultaneously acquire X-ray beam path data in the radiotherapy plan. The three-dimensional medical image data With the aforementioned beam path data All are sets of voxels in three-dimensional space;

[0023] in, For 3D medical imaging data, The three-dimensional spatial region represented by the ray beam path data;

[0024] Sub-step The three-dimensional medical image data According to fixed dimensions Cut into many fixed-size three-dimensional voxel blocks ;

[0025] in, , , This is the index of the voxel block in three-dimensional space. For index Three-dimensional voxel blocks, , , The fixed number of voxels in the voxel block;

[0026] Sub-step Traverse all the aforementioned three-dimensional voxel blocks ; Calculate each three-dimensional voxel block With the aforementioned beam path data Number of spatially intersecting voxels ;

[0027] ,

[0028] in, For three-dimensional voxel blocks With X-ray beam path data The number of voxels that intersect in space. The number of elements in the set. The spatial intersection operation for sets;

[0029] Sub-step For the number of spatially intersecting voxels For a voxel block with a value of zero, calculate the three-dimensional voxel block. To the ray beam path data Shortest spatial distance ;

[0030] ,

[0031] in, For three-dimensional voxel blocks Data to X-ray beam path The shortest spatial distance, For three-dimensional voxel blocks Any voxel point within For X-ray beam path data Any voxel point within For calculation points With point The Euclidean distance between them This is for calculating the minimum value;

[0032] Sub-step Based on the number of spatially intersecting voxels With the shortest spatial distance By setting a preset distance threshold For each three-dimensional voxel block Assign clinical importance level labels ;Judgment conditions:

[0033] like Clinical importance level label Assign it a high importance level;

[0034] like and ≥ Clinical importance level label Assign a value of low importance.

[0035] like and Clinical importance level label Assign it a value of medium importance.

[0036] in, To be assigned to a three-dimensional voxel block Clinical importance level label, This is a preset spatial distance threshold used to distinguish between medium and low importance levels.

[0037] Preferably, step two further includes:

[0038] Sub-step Acquire patient physiological motion signals during the acquisition of the three-dimensional medical image data. ; At the point of time The amplitude of the acquired raw physiological motion signals;

[0039] Sub-step For at a certain point in time The amplitude of the raw physiological motion signal collected Application of bandpass filters The signal is processed to eliminate noise and baseline drift, resulting in a filtered physiological motion signal. ;

[0040] ,

[0041] Then, for the filtered physiological motion signal Amplitude normalization was performed to obtain normalized physiological motion signals. ;

[0042] in, This is the filtered physiological motion signal. Let be the transfer function of the bandpass filter. For convolution operations, This is the physiological motion signal after amplitude normalization;

[0043] Sub-step By analyzing the normalized physiological motion signals Applying Hilbert transform Calculate the normalized physiological motion signal The analyzed signal; the instantaneous phase angle of the analyzed signal is extracted. The instantaneous phase angle of the analyzed signal This refers to the respiratory phase;

[0044] ,

[0045] in, To analyze the signal, For Hilbert transform operations, The imaginary unit, To analyze the instantaneous amplitude of the signal, The instantaneous phase angle of the analytical signal;

[0046] Sub-step , for the segmented voxel blocks Predefined maximum displacement vector The maximum displacement vector For the voxel block The maximum expected three-dimensional displacement amplitude and direction during a complete respiratory cycle; among which... voxel blocks The preset three-dimensional maximum displacement vector, For index Three-dimensional voxel blocks;

[0047] Sub-step According to the respiratory phase and the maximum displacement vector Calculate the current time The voxel block Expected movement The expected movement amount By the breathing phase Scalar function mapped to motion amplitude Perform calculations;

[0048] ,

[0049] The scalar function use Calculations are performed to determine the respiratory phase. Mapped to The range of motion;

[0050] in, voxel blocks At the point of time The expected three-dimensional movement vector, Let be the scalar mapping function from respiratory phase to motion amplitude. It is a cosine function;

[0051] Sub-step All the voxel blocks All corresponding image acquisition timestamps The expected amount of movement Perform aggregation; generate a displacement timetable that records the expected movement of all voxel blocks. The displacement timetable Used for motion compensation processing;

[0052] in, A displacement timetable for recording the expected movement of voxel blocks.

[0053] Preferably, step three further includes:

[0054] Sub-step Get the previous time. 3D medical images As a reference image; simultaneously acquiring the voxel block At the present moment Expected movement The displacement timetable ;

[0055] in, For reference image, For displacement timetable, For index Three-dimensional voxel blocks, voxel blocks At the point of time The expected three-dimensional movement vector;

[0056] Sub-step Initialize with the reference image An empty three-dimensional matrix of the same spatial dimension is used as the current moment to be generated. Predicted images ; Traverse the predicted images Each target voxel The target voxel The three-dimensional spatial coordinates are ;

[0057] in, For the predicted image of the current moment to be generated, For predicting images The target voxel in For target voxels Three-dimensional spatial coordinates;

[0058] Sub-step According to the target voxel Three-dimensional spatial coordinates Determine the target voxel Voxel block ; from the displacement time table Retrieving the voxel block in the middle The corresponding expected movement amount ;

[0059] Sub-step Using the target voxel Three-dimensional spatial coordinates With the expected amount of movement Calculate the target voxel In the reference image Inverse mapping source coordinates ;

[0060] ,

[0061] in, For target voxels In reference image The inverse mapping source coordinates, Reverse mapping source coordinates Three-dimensional spatial coordinates, , , The value is not an integer;

[0062] Sub-step Determine the coordinates of the reverse mapping source. Is it located in the reference image? Effective space boundary Internal; Judgment conditions for:

[0063] ( ≥ ) ( ≤ ) ( ≥ ) ( ≤ ) ( ≥ ) ( ≤ ),

[0064] in, The boundary condition determination result, For reference image Effective spatial boundary, , , For reference image The maximum coordinate in the dimension, For logical AND operation;

[0065] Sub-step If the boundary condition determination result If true, in the reference image Locating the inverse mapping source coordinates Nearby voxels; Obtain the voxel values ​​of the nearby voxels. With the reverse mapping source coordinates Distance weights to the neighboring voxels ; using trilinear interpolation algorithm Calculate the target voxel Predicted voxel values ;

[0066] ,

[0067] If the boundary condition determination result If false, the predicted voxel value will be... Assign a preset background value ;

[0068] in, For the target voxel The predicted voxel values, It is a trilinear interpolation function. The inverse mapping source coordinates The surrounding voxel values ​​of neighboring voxels The inverse mapping source coordinates The corresponding number Trilinear interpolation weights The preset background value;

[0069] Sub-step The predicted voxel value Assign to the predicted image The three-dimensional spatial coordinates described in Place; ,

[0070] traversing all the target voxels Then, the predicted image for the current moment is completed. The generation of the predicted image Used for invertible difference calculations. For the predicted image In coordinates The voxel value at that location.

[0071] Preferably, step four further includes:

[0072] Sub-step Get the current time 3D medical imaging data Get the current time of generation Predicted images ; Traverse the current time 3D medical imaging data Each element in The voxel Spatial coordinates are ; Obtain the voxel Current voxel value at [location] With predicted voxel values ; Calculate the voxels Reversible difference voxel value at the location ;

[0073] ,

[0074] in, For the current moment 3D medical imaging data, The predicted image for the current moment. The current time 3D medical imaging data In coordinates The current voxel value at that location, The current time Predicted images In coordinates The predicted voxel value at the location, For coordinates The invertible difference voxel value at the location;

[0075] Sub-step , all coordinates The reversible difference voxel value Combining to construct invertible difference data The reversible differential data With the current time 3D medical imaging data They have the same spatial dimensions;

[0076] in, The data consists of invertible difference voxel values.

[0077] Sub-step According to the segmented voxel blocks Spatial Index From the reversible differential volume data Extract the corresponding reversible difference voxel blocks ;

[0078] in, For index Three-dimensional voxel blocks, For the invertible differential volume data The index is Reversible difference volumetric blocks;

[0079] Sub-step Obtain the voxel block Corresponding clinical importance level label Define high-priority compression parameters. Compression parameters with low importance level The high-importance compression parameters Includes small compressed block size With high strength verification level The low-importance level compression parameters Includes large compression block size With low strength verification level ; Judgment condition for: ;

[0080] in, As a clinical importance level label, High-priority compression parameter set, Low importance level compression parameter set, For small compression block size, For large compression block size, For high-strength verification level, Low strength verification level;

[0081] Sub-step The hierarchical lossless compression process traverses all the aforementioned reversible differential voxels. Based on the aforementioned clinical importance level labels For each reversible difference volumetric block Select compression parameters If the aforementioned clinical importance level label For high importance level, compression parameters The value is assigned to the high-importance compression parameter. If the aforementioned clinical importance level label For low importance level, compression parameters The value is assigned to the low importance level compression parameter. Apply lossless compression algorithm With the compression parameters For the reversible differential volumetric block Compress the data to generate compressed data blocks. ;

[0082] ,

[0083] in, For the selected compression parameters, It is a lossless compression algorithm. This refers to the compressed data blocks.

[0084] Sub-step The compressed data stream encapsulates and stores all the generated compressed data blocks. The data is collected in a predetermined order to form a complete compressed data stream. The compressed data stream The compressed data stream is stored in a data container. Used to construct a bundle priority catalog; where... For a complete compressed data stream.

[0085] Preferably, step five further includes:

[0086] Sub-step Obtain all the voxel blocks. Corresponding clinical importance level label ; Retrieve all compressed data blocks The compressed data stream Initialize an empty bundle priority directory. Initialize the current data offset. It is zero;

[0087] in, For the voxel block The corresponding clinical importance level label, To fully compress the data stream, For the bundle priority directory to be built, To the compressed data stream The current data offset used to track the current position;

[0088] Sub-step According to the voxel block Spatial Index In sequence, traverse the compressed data stream Each compressed data block contained ; Obtain the compressed data block Data length ;

[0089] ,

[0090] in, This is the spatial index of the voxel block. For index compressed data blocks, For the compressed data block Data length, This is a function for calculating the length of a data block.

[0091] Sub-step Obtain the current spatial index. The corresponding clinical importance level label Determine the clinical importance level label. Is it of high importance? ; Judgment condition for:

[0092] ,

[0093] in, These are preset constant values ​​used to represent high importance levels. The Boolean result of the importance level determination criteria. Equals comparison operator;

[0094] Sub-step The directory mapping relationship is constructed if the determination condition is met. If true, construct a mapping entry. The mapping entries Includes the spatial index As the key, and the current data offset As a value;

[0095] ,

[0096] The mapping entry Add to the bundle priority directory middle;

[0097] ,

[0098] in, For mapping entries containing key-value pairs, This represents the mapping relationship between keys and values. To add the entry to the table of contents;

[0099] Sub-step Offset updates and directory generation use the compressed data blocks. The data length Update the current data offset ;

[0100] ,

[0101] After traversing all the compressed data blocks After that, complete the bundle priority directory. The construction.

[0102] Preferably, step six further includes:

[0103] Sub-step Obtain the defined ray beam path data. To obtain reference images for historical comparison. Define the set of sampling lines. The sampling line set Data from multiple ray beam paths line segments constitute;

[0104] in, For the ray beam path data, Reference images for historical comparison. For the set of sampling lines, For the set of sampling lines One of the sampling lines;

[0105] Sub-step Get the current time 3D medical imaging data ; Traverse the set of sampling lines The sampling lines mentioned in ; Calculate the current time 3D medical imaging data Along the sampling line average voxel intensity value ;

[0106] ,

[0107] in, This refers to the current 3D medical image data.

[0108] The three-dimensional medical image data at the current moment. Along the sampling line The average intensity value of voxels, For the sampling line voxel points on Quantity, For the sampling line On the voxel point, The three-dimensional medical image data at the current moment. At voxel points voxel intensity values;

[0109] Sub-step traverse the set of sampling lines The sampling lines mentioned in Using the reference image ; Calculate the reference image Along the sampling line average voxel intensity value ;

[0110] ,

[0111] in, The reference image Along the sampling line The average intensity value of voxels, The reference image At voxel points voxel intensity values;

[0112] Sub-step Calculate the average intensity value of the current image voxels. With the historical voxel average intensity value absolute difference between ;

[0113] ,in, For the sampling line The statistical absolute difference on the above, This is for absolute value operations;

[0114] Sub-step Define a preset anomaly detection threshold. The statistical absolute difference With the aforementioned anomaly detection threshold Compare; generate abnormal status flags ; Judgment condition for:

[0115] like The abnormal status flag The value is assigned to be true;

[0116] like ≤ The abnormal status flag The value assigned is false;

[0117] in, The preset anomaly detection threshold, For the sampling line Abnormal status indicators, This is the sampling verification result;

[0118] Sub-step When the abnormal state flag When true, locate the sampling line. All voxel blocks traversed Automatically lift the positioned voxel block Storage priority To "store as is"; the voxel block Spatial Index With the abnormal status flag As abnormal information Record to the log file;

[0119] in, For index Three-dimensional voxel blocks, For the voxel block Storage priority, This is the recorded abnormal information.

[0120] This invention provides a lossless compression method for medical image data. It has the following beneficial effects:

[0121] 1. This invention employs a technical solution that acquires the patient's physiological motion signals to generate a displacement timetable, performs motion compensation based on the displacement timetable to generate a predicted image, and finally performs reversible differential storage of the real image and the predicted image. This achieves the technical effect of fully utilizing the temporal redundancy of three-dimensional medical image data, significantly improving the compression ratio, and reducing storage occupation. Compared with the existing technology that uses slice-by-slice compression and fails to utilize the temporal correlation, this invention solves the problems of low compression efficiency and excessive storage resource occupation.

[0122] 2. This invention employs a technical solution that calculates reversible differential data between the current real image and the motion-compensated predicted image, and stores the differential data using a lossless compression algorithm. This achieves the technical effect of ensuring that all voxel values ​​can be accurately restored through the differential data and the predicted image, thus guaranteeing the accuracy of pixel values. Compared with existing compression methods that cannot guarantee complete data reversibility, this invention solves the problem of losing a reliable data foundation for subsequent dose tracing and clinical reassessment.

[0123] 3. The present invention adopts a technical solution that marks the clinical importance level of voxel blocks based on X-ray beam path data and constructs a beam path priority directory for recording the mapping relationship between the position of high-importance voxel blocks and the data offset. This achieves the technical effect of quickly locating key voxel blocks based on the directory without decompressing the entire roll of data when calling the data. Compared with the existing technical solution that requires global decompression of the entire roll of image when calling the data, this solves the problem of high time delay sensitivity in online adaptive radiotherapy workflow.

[0124] 4. The present invention adopts a technical solution that extracts voxel blocks along the X-ray beam data path for sampling and verification before the compression and storage are completed, and automatically raises the storage priority and records the abnormal information for voxel blocks with abnormal verification results. This achieves the technical effect of establishing voxel consistency verification and block data traceability functions at the data encapsulation level. Compared with the data encapsulation methods in the prior art that lack effective verification mechanisms and traceability functions, this invention solves the problem that it is not easy to guarantee the compliance and reliability of dose assessment in data flow. Attached Figure Description

[0125] Figure 1 This is a flowchart of the present invention;

[0126] Figure 2 This is a schematic diagram illustrating the generation of motion-compensated predicted images and reversible difference data. Detailed Implementation

[0127] To enable those skilled in the art to understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort should fall within the scope of protection of the present invention.

[0128] The present invention will now be described in detail with reference to the accompanying drawings:

[0129] Example:

[0130] Please see the appendix Figure 1 and attached Figure 2 This invention provides a lossless compression method for medical image data, comprising:

[0131] Step 1: Acquire the three-dimensional medical image data containing time series data to be compressed, and simultaneously acquire the radiation beam path data in the radiotherapy plan. Divide the three-dimensional medical image data into many voxel blocks of fixed size; label the voxel blocks with clinical importance levels based on the radiation beam path data.

[0132] Step 2: Acquire patient physiological motion signals during 3D medical image data acquisition; analyze physiological motion signals to generate a displacement timetable recording the expected movement of voxel blocks;

[0133] Step 3: Obtain the 3D medical image from the previous moment as a reference image; perform motion compensation processing on the reference image according to the displacement time table to generate the predicted image for the current moment;

[0134] Step 4: Calculate the reversible difference data between the current 3D medical image data and the predicted image; based on the clinical importance level, use an adaptive strategy to perform lossless compression and storage of the reversible difference data;

[0135] Step 5: Based on the clinical importance level labeling, select high-importance voxel blocks; construct a bundle priority directory to record the mapping relationship between the positions of high-importance voxel blocks and the offsets within the compressed data;

[0136] Step 6: Before the compression and storage are completed, extract voxel blocks along the X-ray beam data path for sampling and verification; when the sampling and verification results show an anomaly, automatically increase the storage priority of the corresponding voxel block and record the anomaly information.

[0137] Step 1: After the method of this invention is started, the patient's... Using timestamps, retrieve the time-series 3D medical image data to be compressed, and load the corresponding radiotherapy plan file to parse the X-ray beam path data. ;

[0138] Acquired 3D medical image data The acquired 3D medical image data is loaded into memory. Logically divided into many voxel blocks in three-dimensional space according to a preset fixed size. And record the spatial index of each block. ;

[0139] Reading X-ray beam path data The geometric definition is used. A 3D geometric collision detection algorithm is employed to determine the collision characteristics of each voxel block. Does the spatial bounding box correspond to the ray beam data? The geometric paths intersect.

[0140] If voxel block Spatial range and beam path data Intersecting internal paths constitute a voxel block in memory. Corresponding clinical importance level label Assign a value indicating high importance.

[0141] If they do not intersect, it is considered a voxel block. Corresponding clinical importance level label It is assigned a low importance level.

[0142] Step 2: Acquire patient physiological motion signals synchronously recorded during image acquisition. And for The process involves first applying a digital bandpass filter to remove high-frequency noise and baseline drift, resulting in the filtered physiological motion signal. Then apply Hilbert transform Extracting filtered physiological motion signals Instantaneous respiratory phase .

[0143] Based on a pre-established motion model or learned through 4D-CT initial phase, combined with the current moment respiratory phase Through mapping function Calculate the values ​​of each voxel block exist Expected movement at time ;

[0144] All movements are stored in a key-value pair data structure, namely a displacement timetable. ;

[0145] Step 3: Obtain the 3D medical image from the previous time step and the generated displacement time table. Initialize an all-zero matrix in GPU memory; launch the GPU kernel function, which predicts the image for the current time step. Each target voxel is assigned a thread.

[0146] Each thread first locates its corresponding voxel block. Then, according to the displacement timetable The corresponding expected movement is retrieved from the database; the thread performs a "reverse mapping" operation to calculate the source coordinates; simultaneously, the thread checks for bounds. If not, trilinear interpolation is performed to calculate the predicted voxel values. After the GPU completes execution, the predicted image at the current time step is... Read back from video memory to main memory.

[0147] Step four further includes:

[0148] Sub-step Get the current time 3D medical imaging data Get the current time of generation Predicted images ; Traverse the current time 3D medical imaging data Each element in voxels Spatial coordinates are ; Obtaining voxels Current voxel value at [location] With predicted voxel values ; Calculate voxels Reversible difference voxel value at the location ;

[0149] ,

[0150] in, For the current moment 3D medical imaging data, The predicted image for the current moment. For the current moment 3D medical imaging data In coordinates The current voxel value at that location, For the current moment Predicted images In coordinates The predicted voxel value at the location, For coordinates The invertible difference voxel value at the location;

[0151] Sub-step , all coordinates Reversible difference voxel value Combining to construct invertible difference data Invertible difference data With the current moment 3D medical imaging data They have the same spatial dimensions;

[0152] in, The data consists of invertible difference voxel values.

[0153] Sub-step Based on the segmented voxel blocks Spatial Index From invertible difference data Extract the corresponding reversible difference voxel blocks ;

[0154] in, For index Three-dimensional voxel blocks, For invertible difference data The index is Reversible difference volumetric blocks;

[0155] Sub-step Obtain voxel blocks Corresponding clinical importance level label Define high-priority compression parameters. Compression parameters with low importance level High-priority compression parameters Includes small compressed block size With high strength verification level Low importance level compression parameters Includes large compression block size With low strength verification level ; Judgment condition for: ;

[0156] in, As a clinical importance level label, High-priority compression parameter set, Low importance level compression parameter set, For small compression block size, For large compression block size, For high-strength verification level, Low strength verification level;

[0157] Sub-step Hierarchical lossless compression is performed by traversing all reversible difference prime blocks. Based on clinical importance level labels For each reversible difference volumetric block Select compression parameters If clinical importance level label For high importance level, compression parameters Assigned as a high-priority compression parameter If clinical importance level label For low importance level, compression parameters Assigning values ​​to low importance level compression parameters Apply lossless compression algorithm With compression parameters For reversible difference volumetric blocks Compress the data to generate compressed data blocks. ;

[0158] ,

[0159] in, For the selected compression parameters, It is a lossless compression algorithm. This refers to the compressed data blocks.

[0160] Sub-step The compressed data stream encapsulates and stores all the generated compressed data blocks. The data is collected in a predetermined order to form a complete compressed data stream. Compress the data stream Store in a data container and compress the data stream. Used to construct a bundle priority catalog; where... For a complete compressed data stream.

[0161] Step five further includes:

[0162] Sub-step Obtain all voxel blocks Corresponding clinical importance level label ; Retrieve all compressed data blocks compressed data stream Initialize an empty bundle priority directory. Initialize the current data offset. It is zero;

[0163] in, voxel blocks The corresponding clinical importance level label, To fully compress the data stream, For the bundle priority directory to be built, To compress data streams The current data offset used to track the current position;

[0164] Sub-step According to voxel blocks Spatial Index Sequentially traverse the compressed data stream Each compressed data block contained ; Obtain compressed data blocks Data length ;

[0165] ,

[0166] in, For the spatial index of the voxel block. For index compressed data blocks, To compress data blocks Data length, This is a function for calculating the length of a data block.

[0167] Sub-step Get the current spatial index Corresponding clinical importance level label Determine the clinical importance level label Is it of high importance? ; Judgment condition for:

[0168] ,

[0169] in, These are preset constant values ​​used to represent high importance levels. The Boolean result of the importance level determination criteria. Equals comparison operator;

[0170] Sub-step Directory mapping relationship construction if the following conditions apply If true, construct a mapping entry. ; Mapping entries Includes spatial index As the key, and the current data offset As a value;

[0171] ,

[0172] Map entries Add to bundle priority directory middle;

[0173] ,

[0174] in, For mapping entries containing key-value pairs, This represents the mapping relationship between keys and values. To add the entry to the table of contents;

[0175] Sub-step Offset updates and directory generation use compressed data blocks Data length Update current data offset ;

[0176] ,

[0177] Traversing all compressed data blocks After that, complete the bundle priority catalog. The construction.

[0178] Step six further includes:

[0179] Sub-step Obtain the defined ray beam path data To obtain reference images for historical comparison. Define the set of sampling lines. Sampling line set Data from multiple ray beam paths line segments constitute;

[0180] in, For X-ray beam path data, Reference images for historical comparison. For the set of sampling lines, For the set of sampling lines One of the sampling lines;

[0181] Sub-step Get the current time 3D medical imaging data ; Traverse the set of sampling lines Each sampling line in ; Calculate the current time 3D medical imaging data Along the sampling line average voxel intensity value ;

[0182] ,

[0183] in, This refers to the current 3D medical image data.

[0184] 3D medical image data at the current moment Along the sampling line The average intensity value of voxels, For sampling lines voxel points on Quantity, For sampling lines On the voxel point, 3D medical image data at the current moment At voxel points voxel intensity values;

[0185] Sub-step Traverse the set of sampling lines Each sampling line in Use reference images ; Calculate reference image Along the sampling line average voxel intensity value ;

[0186] ,

[0187] in, For reference image Along the sampling line The average intensity value of voxels, For reference image At voxel points voxel intensity values;

[0188] Sub-step Calculate the average intensity value of the current image voxels. Compared with historical voxel average intensity value absolute difference between ;

[0189] ,in, For sampling lines The statistical absolute difference on the above, This is for absolute value operations;

[0190] Sub-step Define a preset anomaly detection threshold. ; Statistical absolute difference With anomaly detection threshold Compare; generate abnormal status flags ; Judgment condition for:

[0191] like Abnormal status flag The value is assigned to be true;

[0192] like ≤ Abnormal status flag The value assigned is false;

[0193] in, The preset anomaly detection threshold, For sampling lines Abnormal status indicators, This refers to the sampling and verification results;

[0194] Sub-step When the abnormal status flag When true, the location sampling line is... All voxel blocks traversed Automatically lifted and positioned voxel blocks Storage priority To "store as is"; voxel blocks Spatial Index Abnormal status flags As abnormal information Record to the log file;

[0195] in, For index Three-dimensional voxel blocks, voxel blocks Storage priority, This is the recorded abnormal information.

[0196] Data preprocessing was completed, especially the clinical importance level labeling of voxel blocks based on X-ray beam path data, which provides a priority basis for adaptive compression and rapid localization, ensuring the targeted nature of subsequent processing.

[0197] By analyzing the patient's physiological motion signals, the expected movement of the voxel block is generated, laying the foundation for motion compensation prediction and serving as a prerequisite for utilizing time series redundancy.

[0198] Based on the displacement timetable, motion compensation is performed on the reference image to generate a predicted image that is highly similar to the image at the current time. Motion between images has been removed, ensuring that the residual between the current image and the predicted image is minimal, thus significantly improving compression efficiency.

[0199] Reversible differential data ensures that the data is absolutely lossless and reversible, meeting the requirements of dosimetry; based on the clinical importance level, an adaptive strategy is used for lossless compression, achieving a balance between the storage reliability of high-importance voxel blocks and the efficient compression of low-importance voxel blocks, thus optimizing the overall storage performance.

[0200] High-importance voxel blocks crucial for dose calculation were selected based on clinical importance, and a bundle-priority catalog was constructed. This catalog records the offset of key data in the compressed stream, enabling the direct extraction of required data blocks without decompressing the entire image roll during dose recalculation, significantly improving clinical workflow efficiency.

[0201] Sampling and verification are performed on the front-end ray beam path of storage to establish a built-in voxel-level consistency verification mechanism. When the verification result shows an anomaly, the storage priority of the corresponding voxel block is automatically increased to ensure that critical data is not distorted due to compression, and the anomaly information is recorded to provide reliability and compliance traceability functions in the data flow and reproduction process.

[0202] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A lossless compression method for medical image data, characterized in that, include: Step 1: Acquire 3D medical image data and X-ray beam path data from the radiotherapy plan, and divide the 3D medical image data into many voxel blocks; Voxel blocks were labeled with clinical importance levels based on X-ray beam path data; Step 2: Acquire patient physiological motion signals during 3D medical image data acquisition; analyze physiological motion signals to generate a displacement timetable recording the expected movement of voxel blocks; Step 3: Obtain the 3D medical image from the previous moment as a reference image; perform motion compensation processing on the reference image according to the displacement time table to generate the predicted image for the current moment; Step 4: Calculate the reversible difference data between the current 3D medical image data and the predicted image; based on the clinical importance level, use an adaptive strategy to perform lossless compression and storage of the reversible difference data; Step 5: Based on the clinical importance level labeling, select high-importance voxel blocks; construct a bundle priority directory to record the mapping relationship between the positions of high-importance voxel blocks and the offsets within the compressed data; Step 6: Before the compression and storage are completed, extract voxel blocks along the X-ray beam data path for sampling and verification; when the sampling and verification results show an anomaly, automatically increase the storage priority of the corresponding voxel block and record the anomaly information.

2. The lossless compression method for medical image data according to claim 1, characterized in that, The clinical importance level labeling includes: labeling voxel blocks located inside the X-ray beam path data as high importance; and labeling voxel blocks far from the X-ray beam path data as low importance.

3. The lossless compression method for medical image data according to claim 1, characterized in that, The motion compensation process includes performing translation or deformation pull-back operations on the voxel blocks in the reference image according to the movement parameters recorded in the displacement time table.

4. The lossless compression method for medical image data according to claim 1, characterized in that, The beam path priority directory is used to extract the high importance voxel blocks directly from the directory when performing dose recalculation without decompressing the entire roll of image data, thus enabling rapid location and retrieval. The automatic priority increase of the storage of the corresponding voxel block includes: when the sampling verification result shows an anomaly, stopping the storage of the reversible differential data for the voxel block and directly storing the original uncompressed data of the voxel block.

5. The lossless compression method for medical image data according to claim 1, characterized in that, Step one further includes: Sub-step Acquire the three-dimensional medical image data containing time series data to be compressed. Simultaneously acquire X-ray beam path data in the radiotherapy plan. The three-dimensional medical image data With the aforementioned beam path data All are sets of voxels in three-dimensional space; in, For 3D medical imaging data, The three-dimensional spatial region represented by the ray beam path data; Sub-step The three-dimensional medical image data According to fixed dimensions Cut into many fixed-size three-dimensional voxel blocks ; in, , , This is the index of the voxel block in three-dimensional space. For index Three-dimensional voxel blocks, , , The fixed number of voxels in the voxel block; Sub-step Traverse all the aforementioned three-dimensional voxel blocks ; Calculate each three-dimensional voxel block With the aforementioned beam path data Number of spatially intersecting voxels ; ， in, For three-dimensional voxel blocks With X-ray beam path data The number of voxels that intersect in space. The number of elements in the set. The spatial intersection operation for sets; Sub-step For the number of spatially intersecting voxels For a voxel block with a value of zero, calculate the three-dimensional voxel block. To the ray beam path data Shortest spatial distance ; ， in, For three-dimensional voxel blocks Data to X-ray beam path The shortest spatial distance, For three-dimensional voxel blocks Any voxel point within For X-ray beam path data Any voxel point within For calculation points With point The Euclidean distance between them This is an operation to find the minimum value; Sub-step Based on the number of spatially intersecting voxels With the shortest spatial distance By setting a preset distance threshold For each three-dimensional voxel block Assign clinical importance level labels ;Judgment conditions: like Clinical importance level label Assign it a high importance level; like and ≥ Clinical importance level label Assign a value of low importance. like and Clinical importance level label Assign it a value of medium importance. in, To be assigned to a three-dimensional voxel block Clinical importance level label, This is a preset spatial distance threshold used to distinguish between medium and low importance levels.

6. The lossless compression method for medical image data according to claim 1, characterized in that, Step two further includes: Sub-step Acquire patient physiological motion signals during the acquisition of the three-dimensional medical image data. ; At the point of time The amplitude of the acquired raw physiological motion signals; Sub-step For at a certain point in time The amplitude of the raw physiological motion signal collected Application of bandpass filters The signal is processed to eliminate noise and baseline drift, resulting in a filtered physiological motion signal. ; ， Then, for the filtered physiological motion signal Amplitude normalization was performed to obtain normalized physiological motion signals. ; in, This is the filtered physiological motion signal. Let be the transfer function of the bandpass filter. For convolution operations, This is the physiological motion signal after amplitude normalization; Sub-step By analyzing the normalized physiological motion signals Applying Hilbert transform Calculate the normalized physiological motion signal The analyzed signal; the instantaneous phase angle of the analyzed signal is extracted. The instantaneous phase angle of the analyzed signal This refers to the respiratory phase; ， in, To analyze the signal, For Hilbert transform operations, The imaginary unit, To analyze the instantaneous amplitude of the signal, The instantaneous phase angle of the analytical signal; Sub-step , for the segmented voxel blocks Predefined maximum displacement vector The maximum displacement vector For the voxel block The maximum expected three-dimensional displacement amplitude and direction during a complete respiratory cycle; among which... voxel blocks The preset three-dimensional maximum displacement vector, For index Three-dimensional voxel blocks; Sub-step According to the respiratory phase and the maximum displacement vector Calculate the current time The voxel block Expected movement The expected movement amount By the breathing phase Scalar function mapped to motion amplitude Perform calculations; ， The scalar function use Calculations are performed to determine the respiratory phase. Mapped to The range of motion; in, voxel blocks At the point of time The expected three-dimensional movement vector, Let be the scalar mapping function from respiratory phase to motion amplitude. It is a cosine function; Sub-step All the voxel blocks All corresponding image acquisition timestamps The expected amount of movement Perform aggregation; generate a displacement timetable that records the expected movement of all voxel blocks. The displacement timetable Used for motion compensation processing; in, A displacement timetable for recording the expected movement of voxel blocks.

7. The lossless compression method for medical image data according to claim 1, characterized in that, Step three further includes: Sub-step Get the previous time. 3D medical images As a reference image; simultaneously acquiring the voxel block At the present moment Expected movement The displacement timetable ; in, For reference image, For displacement timetable, For index Three-dimensional voxel blocks, voxel blocks At the point of time The expected three-dimensional movement vector; Sub-step Initialize with the reference image An empty three-dimensional matrix of the same spatial dimension is used as the current moment to be generated. Predicted images ; Traverse the predicted images Each target voxel The target voxel The three-dimensional spatial coordinates are ; in, For the predicted image of the current moment to be generated, For predicting images The target voxel in For target voxels Three-dimensional spatial coordinates; Sub-step According to the target voxel Three-dimensional spatial coordinates Determine the target voxel Voxel block ; from the displacement time table Retrieving the voxel block in the middle The corresponding expected movement amount ; Sub-step Using the target voxel Three-dimensional spatial coordinates With the expected amount of movement Calculate the target voxel In the reference image Inverse mapping source coordinates ; ， in, For target voxels In reference image The inverse mapping source coordinates, Reverse mapping source coordinates Three-dimensional spatial coordinates, , , The value is not an integer; Sub-step Determine the coordinates of the reverse mapping source. Is it located in the reference image? Effective space boundary Internal; Judgment conditions for: ( ≥ ) ( ≤ ) ( ≥ ) ( ≤ ) ( ≥ ) ( ≤ )， in, The boundary condition determination result, For reference image Effective spatial boundary, , , For reference image The maximum coordinate in the dimension, For logical AND operation; Sub-step If the boundary condition determination result If true, in the reference image Locating the inverse mapping source coordinates Nearby voxels; Obtain the voxel values ​​of the nearby voxels. With the reverse mapping source coordinates Distance weights to the neighboring voxels ; using trilinear interpolation algorithm Calculate the target voxel Predicted voxel values ; ， If the boundary condition determination result If false, the predicted voxel value will be... Assign a preset background value ; in, For the target voxel The predicted voxel values, It is a trilinear interpolation function. The inverse mapping source coordinates The surrounding voxel values ​​of neighboring voxels The inverse mapping source coordinates The corresponding number Trilinear interpolation weights The preset background value; Sub-step The predicted voxel value Assign to the predicted image The three-dimensional spatial coordinates described in Place; , traversing all the target voxels Then, the predicted image for the current moment is completed. The generation of the predicted image Used for invertible difference calculations. For the predicted image In coordinates The voxel value at that location.

8. The lossless compression method for medical image data according to claim 1, characterized in that, Step four further includes: Sub-step Get the current time 3D medical imaging data Get the current time of generation. Predicted images ; Traverse the current time 3D medical imaging data The various elements in The voxel Spatial coordinates are ; Obtain the voxel Current voxel value at [location] With predicted voxel values ; Calculate the voxels Reversible difference voxel value at the location ; ， in, For the current moment 3D medical imaging data, The predicted image for the current moment. The current time 3D medical imaging data In coordinates The current voxel value at that location, The current time Predicted images In coordinates The predicted voxel value at the location, For coordinates The reversible difference voxel value at the location; Sub-step , all coordinates The reversible difference voxel value Combining to construct invertible difference data The reversible differential data With the current time 3D medical imaging data They have the same spatial dimensions; in, The data consists of invertible difference voxel values. Sub-step According to the segmented voxel blocks Spatial Index From the reversible differential volume data Extract the corresponding reversible difference voxel blocks ; in, For index Three-dimensional voxel blocks, For the invertible difference volume data The index is Reversible difference volumetric blocks; Sub-step Obtain the voxel block Corresponding clinical importance level label Define high-priority compression parameters. Compression parameters with low importance level The high-importance compression parameters Includes small compressed block size With high strength verification level The low-importance level compression parameters Includes large compression block size With low strength verification level ; Judgment condition for: ; in, As a clinical importance level label, High-priority compression parameter set, Low importance level compression parameter set, For small compression block size, For large compression block size, For high-strength verification level, Low strength verification level; Sub-step The hierarchical lossless compression process traverses all the aforementioned reversible differential voxels. Based on the aforementioned clinical importance level labels For each reversible difference volumetric block Select compression parameters If the aforementioned clinical importance level label For high importance level, compression parameters The value is assigned to the high-importance compression parameter. If the aforementioned clinical importance level label For low importance level, compression parameters The value is assigned to the low importance level compression parameter. Apply lossless compression algorithm With the compression parameters For the reversible differential volumetric block Compress the data to generate compressed data blocks. ; ， in, For the selected compression parameters, It is a lossless compression algorithm. This refers to the compressed data blocks. Sub-step The compressed data stream encapsulates and stores all the generated compressed data blocks. The data is collected in a predetermined order to form a complete compressed data stream. ; the compressed data stream The compressed data stream is stored in a data container. Used to construct a bundle priority catalog; where... For a complete compressed data stream.

9. The lossless compression method for medical image data according to claim 1, characterized in that, Step five further includes: Sub-step Obtain all the voxel blocks. Corresponding clinical importance level label ; Retrieve all compressed data blocks The compressed data stream Initialize an empty bundle priority directory. Initialize the current data offset. It is zero; in, For the voxel block The corresponding clinical importance level label, To fully compress the data stream, For the bundle priority directory to be built, To the compressed data stream The current data offset used to track the current position; Sub-step According to the voxel block Spatial Index In sequence, traverse the compressed data stream Each compressed data block contained ; Obtain the compressed data block Data length ; ， in, The spatial index of the voxel block. For index compressed data blocks, For the compressed data block Data length, A function for calculating the length of a data block; Sub-step Obtain the current spatial index. The corresponding clinical importance level label Determine the clinical importance level label. Is it of high importance? ; Judgment condition for: ， in, These are preset constant values ​​used to represent high importance levels. The Boolean result of the importance level determination criteria. Equals comparison operator; Sub-step The directory mapping relationship is constructed if the determination condition is met. If true, construct a mapping entry. The mapping entries Includes the spatial index As the key, and the current data offset As a value; ， The mapping entry Add to the bundle priority directory middle; ， in, For mapping entries containing key-value pairs, This represents the mapping relationship between keys and values. To add the entry to the table of contents; Sub-step Offset updates and directory generation use the compressed data blocks. The data length Update the current data offset ; ， After traversing all the compressed data blocks After that, complete the bundle priority directory. The construction of.

10. A lossless compression method for medical image data according to claim 1, characterized in that, Step six further includes: Sub-step Obtain the defined ray beam path data. To obtain reference images for historical comparison. Define the set of sampling lines. The sampling line set Data from multiple ray beam paths line segments constitute; in, For the ray beam path data, Reference images for historical comparison. For the set of sampling lines, For the set of sampling lines One of the sampling lines; Sub-step Get the current time 3D medical imaging data ; Traverse the set of sampling lines The sampling lines mentioned in ; Calculate the current time 3D medical imaging data Along the sampling line average voxel intensity value ; ， in, This refers to the current 3D medical image data. The three-dimensional medical image data at the current moment. Along the sampling line The average intensity value of voxels, For the sampling line voxel points on Quantity, For the sampling line On the voxel point, The three-dimensional medical image data at the current moment. At voxel points voxel intensity values; Sub-step traverse the set of sampling lines The sampling lines mentioned in Using the reference image ; Calculate the reference image Along the sampling line average voxel intensity value ; ， in, The reference image Along the sampling line The average intensity value of voxels, The reference image At voxel points voxel intensity values; Sub-step Calculate the average intensity value of the current image voxels. With the historical voxel average intensity value absolute difference between ; ,in, For the sampling line The statistical absolute difference on the above, This is for absolute value operations; Sub-step Define a preset anomaly detection threshold. The statistical absolute difference With the aforementioned anomaly detection threshold Compare; generate abnormal status flags ; Judgment condition for: like The abnormal status flag The value is assigned to be true; like ≤ The abnormal status flag The value assigned is false; in, The preset anomaly detection threshold, For the sampling line Abnormal status indicators, This is the sampling verification result; Sub-step When the abnormal status flag When true, locate the sampling line. All voxel blocks traversed Automatically lift the positioned voxel block Storage priority To "store as is"; the voxel block Spatial Index With the abnormal status flag As abnormal information Record to the log file; in, For index Three-dimensional voxel blocks, For the voxel block Storage priority, This is the recorded abnormal information.

Images