Photon counting CT data compression method

Abstract

The invention discloses a photon counting CT (Computed Tomography) data compression method, which is used for storing data bits with optimal data matching in different energy levels and different areas. The method comprises the following specific steps: S1, energy spectrum partitioning: acquiring projection data, extracting the projection data through a data processing unit, and then dividing the energy level of the projection data into a plurality of regions; s2, dynamic threshold decision making: dividing the energy spectrum data according to a scanning area and calculating an area division threshold of the energy spectrum data; s3, performing projection data area division and bit width automatic allocation: calculating the maximum storage bit number corresponding to area data; and S4, packaging the original data: storing the data according to the maximum storage bit number, and recording the region position information and the storage bit number information. The method is used for carrying out lossless compression on the original data volume collected by the photon counting CT, and the data volume needing to be transmitted is effectively reduced under the condition that the data precision is not lost; computing resources consumed by the algorithm are small, and good real-time performance is achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of lossless data compression, and particularly relates to a photon counting CT data compression method. Background Art

[0002] As a new generation of CT, photon counting CT provides better performance than conventional CT. Its core principle is that through a photon counting detector, X-ray photons attenuated by the human body are counted, and the energy level range of a certain photon is distinguished, usually 4 to 8 energy level segments. In addition, the pixel size of the photon counting detector is usually one order of magnitude smaller than that of the conventional CT detector. These characteristics can bring important clinical values, such as higher spatial resolution, better low-contrast resolution, removal of metal artifacts, and the ability of material decomposition, that is, clinical differential diagnosis. However, the smaller physical size of the detector pixel will generate more raw data. Especially in order to obtain better energy resolution, data acquisition of multiple energy levels is often carried out, which leads to a transmission bandwidth of hundreds of GB / s for transmitting the projection data collected by the photon counting CT compared with the conventional CT, bringing huge challenges to both data transmission and storage of the CT system, especially when information loss caused by data compression or processing is not desired in clinical applications.

[0003] However, the X-ray will be greatly attenuated after passing through the object to be scanned; the spectral attenuation degrees of different energy levels by the object are different, and the lower the energy level, the greater the attenuation degree; the distribution of the incident spectrum at different energy levels is different, that is, the number of incident photons at different energy levels is different.

[0004] Currently, the main literature on the compression processing of CT raw data includes the following several articles:

[0005] US20210057084A1 MEDICAL DATA PROCESSING APPARATUS,MEDICAL DATA PROCESSING METHOD,AND MEDICAL IMAGE DIAGNOSTIC proposes a compression algorithm for multi-channel data, and its data organization form is similar to that of photon counting CT. The core idea of this patent is to first convert the raw data into frequency domain data through data conversion, then compress and convert it by the method of AI, and finally perform quantization coding. This method requires a high-performance processor to execute complex machine learning algorithms and data processing.

[0006] PichaShunhavanich2016 Lossless Compression of Projection Data from Photon Counting Detectors proposed a prediction algorithm. It predicts the current value through the weighted average of adjacent samples (adjacent views and channels) and the predicted values of energy levels, and encodes and records the residual error. At the same time, a context model based on local gradients is used to mark boundary information. The data compression effect is achieved through the compression encoding of the predicted value and the residual. The advantage of this method is that it can perform lossy and lossless compression. The disadvantages are that the effectiveness of the prediction algorithm needs to be further confirmed, the implementation process is relatively complex, and it has limitations in high-speed transmission.

[0007] Sjo ·· Sjo lin2017 Compression of CT sinogram data by decimation in the view direction proposed a data compression method of decimating one in ten in the sampling view direction, which can improve the high-contrast resolution (compared with the low view sampling rate) without increasing the data acquisition amount. Its core idea is to first increase the sampling rate and then perform filtering and downsampling. The advantage of this method is that it combines the actual application scenario of improving high-contrast resolution. However, increasing sampling and then performing downsampling, the data compression effect will depend on the specific application, and the image effect obtained by downsampling is lossy, and its effectiveness also depends on the specific clinical application.

[0008] Kimura2021 Data traffic compression in spectral photon-counting CT imaging based on human visual characteristics proposed a data compression method for photon-counting CT, aiming to compress the data volume while maintaining the spatial resolution, and the final image remains visually consistent. Its core idea is to collect data at a high resolution for a certain energy level, and then collect data at a low resolution (interval collection or combined collection) for the data used to maintain the energy spectrum information, and then merge the images reconstructed from the two groups of data to achieve an image that takes into account both energy spectrum information and resolution information. The advantage of this method is that it comprehensively considers retaining energy spectrum information and spatial resolution information. The disadvantage is that there are differences between the spatial resolution information of different energy levels, so there will be differences between the final synthesized image effect and the original image, and its effectiveness depends on the specific clinical application.

[0009] Yirong Yang 2023 "Empirical optimization of energy bin weights for compressing measurements with realistic photon counting x-ray detectors" proposed an energy-weighted photon counting CT data compression algorithm. Its core idea is to perform weighted merging of data at different energy levels while maintaining the effects of virtual monochromatic images and virtual non-contrast images, retaining data at 2 or 3 (depending on the type of material decomposition) energy levels, thereby achieving the effect of reducing the data volume. The advantages of this method are that it fully combines the image category information of photon counting CT, while the disadvantages are high complexity and lossy nature. Summary of the Invention

[0010] The present invention aims to solve the defects and deficiencies existing in the above-mentioned prior art, and provides a photon counting CT data compression method for lossless compression of the original data volume collected by photon counting CT, effectively reducing the data volume to be transmitted without loss of data accuracy; the algorithm is simple and consumes little computing resources.

[0011] The technical solution of the present invention is as follows: A photon counting CT data compression method that matches the best number of data bits for data in different energy levels and different regions for storage, and the steps are as follows:

[0012] S1. Energy spectrum partitioning: Collect projection data, extract it through a data processing unit, and then divide the energy levels of the projection data into multiple regions;

[0013] S2. Dynamic threshold decision: Divide the energy spectrum data according to the scanning region and calculate its region division threshold;

[0014] S3. Projection data region partitioning and automatic bit width allocation: Calculate the maximum number of storage bits corresponding to the region data;

[0015] S4. Original data encapsulation: Store the data according to the maximum number of storage bits, and record the region position information and the storage bit number information.

[0016] Preferably, step S1 specifically refers to dividing the projection data according to the view angle and row, and the processing unit uses the single-row data as the minimum unit, and the currently processed view angle and row are marked with vid and sid.

[0017] Preferably, step S2 specifically includes:

[0018] 1) Set the ratios ratio1 and ratio2 according to the intensity for the two regions of the object to be scanned and air;

[0019] 2) Traverse to obtain the maximum projection data value of the current perspective and row:

[0020] val mx = max(data(1:N c , 1:N b , sid, vid))

[0021] where data is the energy spectrum data, N c is the number of channels, and N b is the number of energy levels;

[0022] 3) Calculate the division thresholds for the two regions:

[0023] trd1 = val mx × ratio1

[0024] trd2 = val mx × ratio2;

[0025] 4) Use trd1 and trd2 to divide the projection data into 5 regions in the channel direction:

[0026] rgn(1) = pos(1):pos(2), data(pos(1):pos(2), 1:N b , sid, vid) < trd1

[0027] rgn(2) = pos(2)+1:pos(3)-1

[0028] rgn(3) = pos(3):pos(4), data(pos(3):pos(4), 1:N b , sid, vid) < trd2

[0029] rgn(4) = pos(4)+1:pos(5)-1

[0030] rgn(5) = pos(5):pos(6), data(pos(5):pos(6), 1:N b , sid, vid) < trd1

[0031] where pos(1) = 1 and pos(6) = N c .

[0032] Preferably, the specific operation of dividing the projection data region in step S3 is to statistically calculate the maximum value of each energy level in each region:

[0033] val mx (rid, bid) = max(data(rgn(rid), bid, sid, vid)

[0034] Among them, rid is the current area and bid is the current energy level.

[0035] Preferably, the specific operation of bit width automatic allocation in step S3 is to calculate val mx The maximum number of storage bits corresponding to (rid, bid):

[0036] bits(rid,bid) = nbit,2 nbit <val mx (rid,bid) ≤ 2 nbit+1 -1.

[0037] Preferably, the specific operation of step S4 is to store the data according to nbit, and additionally record the pos and nbit information for data recovery.

[0038] The beneficial effects of the present invention are as follows: The present invention is used to losslessly compress the original data volume collected by photon counting CT, effectively reducing the data volume to be transmitted without loss of data accuracy; the algorithm is simple and consumes little computing resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application, and those of ordinary skill in the art can obtain other drawings without creative efforts based on these drawings.

[0040] Figure 1 It is a schematic diagram of the data compression process of the present invention;

[0041] Figure 2 It is a schematic diagram of the data layout of a single view in the present invention;

[0042] Figure 3 It is a schematic diagram of the data phantom for compression in the present invention;

[0043] Figure 4 For Figure 3 It is a schematic diagram of the 7 - energy - level acquisition data distribution of the data phantom in the attached drawings, and each line is the data of one energy level. DETAILED DESCRIPTION OF THE INVENTION

[0044] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0045] Embodiment 1

[0046] The present invention is based on the following three facts: 1. X-rays will be significantly attenuated after passing through the object to be scanned; 2. The attenuation degrees of spectra with different energy levels by the object are different, and the lower the energy level, the greater the attenuation degree; 3. The distributions of the incident spectrum at different energy levels are different, that is, the numbers of incident photons at different energy levels are different.

[0047] A data compression method based on the energy level distribution of the scanned object is provided, which stores the data matching the best data bits for different regions at different energy levels. The basic process is as Figure 1 shown, and the specific steps are as follows:

[0048] S1. Projection data acquisition and processing unit extraction:

[0049] The projection data is divided according to the view angle and row (see the appendix for the data layout Figure 2 ), and the processing unit takes the single-row data as the minimum unit. The currently processed view angle and row are marked with vid and sid;

[0050] S2. Divide the energy spectrum data according to the scanned area:

[0051] 1) Set the intensity ratios ratio1 and ratio2 for the two regions of the object to be scanned and the air;

[0052] 2) Traverse to obtain the maximum projection data value of the current view angle and row:

[0053] val mx = max(data(1:N c , 1:N b , sid, vid))

[0054] where data is the energy spectrum data, N c is the number of channels, and N b is the number of energy levels;

[0055] 3) Calculate the division thresholds of the two regions:

[0056] trd1 = val mx × ratio1

[0057] trd2 = val mx × ratio2;

[0058] 4) Divide the projection data into 5 regions in the channel direction using trd1 and trd2:

[0059] rgn(1) = pos(1):pos(2), data(pos(1):pos(2), 1:N b , sid, vid) < trd1

[0060] rgn(2) = pos(2) + 1:pos(3) - 1

[0061] rgn(3) = pos(3):pos(4), data(pos(3):pos(4), 1:N b , sid, vid) < trd2

[0062] rgn(4) = pos(4) + 1:pos(5) - 1

[0063] rgn(5) = pos(5):pos(6), data(pos(5):pos(6), 1:N b , sid, vid) < trd1

[0064] where pos(1) = 1 and pos(6) = N c ;

[0065] S3 - S4. Perform data compression:

[0066] S3. Statistically calculate the maximum value of each energy level in each region:

[0067] val mx (rid, bid) = max(data(rgn(rid), bid, sid, vid)

[0068] where rid is the current region and bid is the current energy level;

[0069] Calculate val mx (rid, bid) corresponding maximum storage bits:

[0070] bits(rid, bid) = nbit, 2 nbit < val mx (rid, bid) ≤ 2 nbit+1 -1;

[0071] S4. Store the data according to nbit and additionally record the pos and nbit information for data recovery.

[0072] Example 2

[0073] An embodiment of the present invention is as follows:

[0074] 1. Adopt 7 - level data acquisition, and the storage bit for each data before compression is 16 bits;

[0075] 2. Adopt a 120 kVp scan, and the energy levels are set as: [25 34 50 60 70 90 100];

[0076] 3. Perform data acquisition on the digital phantom shown in Attachment Figure 3 The data distribution of a single view and a single row obtained is as shown in Attachment Figure 4 shown;

[0077] 4. Set: ratio1 = 0.3, ratio2 = 0.12, and the calculated data regions are: pos(2) = 357, pos(3) = 847, pos(4) = 1554, pos(5) = 2044

[0078] rgn(1) = 1:357

[0079] rgn(2) = 358:846

[0080] rgn(3) = 847:1554

[0081] rgn(4) = 1555:2043

[0082] rgn(5) = 2044:2400;

[0083] 5. The maximum values of each energy level in each region are as follows (energy level x region):

[0084]

[0085] 6. The maximum storage bits corresponding to the maximum values of each energy level in each region are:

[0086]

[0087] 7. The compression ratios (after compression / before compression) corresponding to each energy level in each region are:

[0088]

[0089] 8. The overall compression ratio of the current data is: 0.8392;

[0090] 9. The compression ratios of each energy level: 0.5535 0.8521 0.9076 0.9076 0.9262 0.8637 0.8637;

[0092] 10. The compression ratios of each region are: 0.7857 0.9554 0.7321 0.9554 0.7857;

[0094] 11. The additional information amount is:

[0095] 1) 4 pos information, each piece of information is recorded using 12 bits, which is 48 bits;

[0096] 2) The bits information of each data segment of each energy level is recorded using 4 bits, which is: 5 x 7 x 4 = 140 bits;

[0097] In summary: Each row of each perspective requires 188 bits of data for compressed information recording. Taking the current data as an example, the amount of data saved by compression is: 2400 x 7 x 16 x 0.8392 = 43223 bits ≈ 5.28 KB, and the additional recorded information accounts for 0.43% of the saved amount.

[0098] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present invention. Any reference signs in the claims should not be regarded as limiting the claimed rights.

[0099] In addition, it should be understood that although this specification is described according to embodiments, not every embodiment only contains an independent technical solution. This narrative way of the specification is only for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A photon counting CT data compression method stores data with the best number of bits matched for different energy levels and different regions. It is characterized in that: The steps are as follows: S1. Energy spectrum partitioning: Collect projection data, extract it through a data processing unit, and then divide the energy levels of the projection data into multiple regions; S2. Dynamic threshold decision-making: Divide the energy spectrum data according to the scanning region and calculate the region division threshold; S3. Projection data region partitioning and bit width automatic allocation: Calculate the maximum storage bit number corresponding to the region data; S4. Original data encapsulation: Store the data according to the maximum storage bit number, and record the region position information and storage bit number information.

2. The photon counting CT data compression method according to claim 1, wherein: Specifically, in step S1, the projection data is divided according to the view angle and row. The processing unit takes the single-row data as the minimum unit, and the currently processed view angle and row are marked with vid and sid.

3. A photon counting CT data compression method according to claim 1, characterized in that: Step S2 specifically includes: 1) Set the intensity ratios ratio1 and ratio2 for the two regions of the object to be scanned and air; 2) Traverse to obtain the maximum projection data value of the current view angle and row; val mx = max(data(1:N c , 1:N b , sid, vid)) Among them, data is the energy spectrum data, N c is the number of channels, N b is the number of energy levels; 3) Calculate the division thresholds of the two regions; trd1 = val mx × ratio1 trd2 = val mx × ratio2; 4) Use trd1 and trd2 to divide the projection data into 5 regions in the channel direction: rgn(1) = pos(1):pos(2), data(pos(1):pos(2), 1:N b , sid, vid) < trd1 rgn(2) = pos(2) + 1:pos(3) - 1 rgn(3) = pos(3):pos(4), data(pos(3):pos(4), 1:N b , sid, vid) < trd2 rgn(4) = pos(4) + 1:pos(5) - 1 rgn(5) = pos(5):pos(6), data(pos(5):pos(6), 1:N b , sid, vid) < trd1 Among them, pos(1) = 1 and pos(6) = N c .

4. A photon counting CT data compression method according to claim 1, characterized in that: Specifically, in step S3, the projection data region partitioning is as follows: Statistical maximum value of each energy level in each region: val mx (rid, bid) = max(data(rgn(rid), bid, sid, vid) where rid is the current region and bid is the current energy level.

5. A photon counting CT data compression method according to claim 1, characterized in that: Specifically, in step S3, the bit width automatic allocation is as follows: Calculate val mx (rid, bid) corresponding maximum storage bits: bits(rid,bid) = nbit,2 nbit <val mx (rid,bid) ≤ 2 nbit+1 -1.

6. A photon counting CT data compression method according to claim 1, characterized in that: Specifically, step S4 is as follows: Store the data according to nbit, and additionally record the pos and nbit information for data recovery.

Images