CN118828005A - A CT detector data compression method based on register redundancy - Google Patents

Abstract

The invention discloses a register redundancy-based CT detector data compression method, which comprises the following steps: s1, data preprocessing: decomposing the original data into an initial value and a difference coefficient; s2, data compression: the method is characterized in that a single register is used for storing a plurality of difference coefficients by adopting a register disassembling mode; s3, data transmission: transmitting the compressed data through a slip ring; s4, data decompression: restoring the original data by using the transmitted initial value and the difference coefficient; s5, image reconstruction: and realizing image reconstruction by using the decompressed data. According to the invention, the original data is decomposed into the initial value and the difference coefficient in a data preprocessing mode, and the dynamic range of the difference coefficient is smaller, so that the large register redundancy is realized, the effective utilization of the register redundancy is realized in a register disassembling mode, the hardware resources are fully utilized, the compression efficiency is effectively improved, the pressure of the transmission bandwidth and the storage space is relieved, and the transmission reliability is improved.

Description

A CT detector data compression method based on register redundancy

Technical Field

The invention belongs to the technical field of medical X-ray imaging, and in particular is a CT detector data compression method based on register redundancy.

Background Art

Since its birth, CT examination has been one of the most important parts of medical examination. According to incomplete statistics, as early as 2016, the total number of CT examinations in my country reached an astonishing 290 million, accounting for 26.88% of the total number of radiological diagnosis and treatment, and maintained an annual growth rate of 11.7% in the past decade. However, the insufficient resolution of most CT scans due to reasons such as the number of scanning rows seriously affects the diagnosis of doctors, especially the diagnosis of early cancer, solitary pulmonary nodules and other diseases.

Existing high-resolution CT faces the problem of serious mismatch between data volume and transmission bandwidth. Taking 320-row CT as an example, the existing 320-row CT has a rotation speed of about 0.23s per revolution, the number of pixels in each row is about 900, the data of each pixel is 16-bit binary data, and the sampling rate of one revolution is about 4096. The required transmission bandwidth is at least 320*900*16*4096/0.23/(1024*1024*1024)＝76.42Gbps. The transmission bandwidth required for more advanced photon counting CT is more than 350Gbps. However, even with the most advanced multi-channel parallel optical fiber transmission, the maximum transmission bandwidth is only 40Gbps. In addition, the storage space required for a single scan data is hundreds of G. Large-scale data transmission will increase the risk of data loss and erroneous transmission.

In addition, CT data is mostly transmitted and stored in 16-bit binary data, which means that a single data needs a 16-bit register for storage and transmission. However, in actual use, the dynamic range of the CT raw data is often quite different from the maximum dynamic range that can be represented by the 16-bit register, resulting in register redundancy. Taking the CT data of 5000 as an example, its representation in the 16-bit register is 0001 00111000 1000. At this time, the value of the upper 3-bit register is 0, which has no practical meaning, resulting in the redundancy of the upper 3-bit register. When the data scale is large, a large number of registers will be redundant, wasting hardware resources.

With the continuous development of CT technology, the scanning speed, number of layers, and image resolution of CT equipment are also increasing, so the scale of CT data is also increasing, which also leads to the transmission bandwidth being far from meeting the needs of CT data transmission.

To address the above issues, US 7916830 B2 proposes an edge detection compression method, which first detects the edges in the projection data, then compares the detected edges with preset positive and negative thresholds to determine the boundaries, and finally compresses the projection samples or differential samples between the boundaries.

JP2003290216A patent proposes a method for classifying projection data by using a preset threshold value. The method first divides the projection data into an air information area and a subject information area, and then deletes the air information area.

Patent CN 104825183A proposes a packaging and transmission method for CT data. The method first splices and packages the detector data to obtain CT data of a single angle, then preprocesses the CT data of a single angle, and then compresses the CT data by utilizing the difference between the CT data of the current angle and the CT data of the previous angle.

Patent CN117201800A proposes a compression method based on spatial redundancy. The method first divides the projection data into several connected domains, then determines the importance of all connected domains, and then compresses them using run-length encoding.

It can be seen from the retrieved relevant literature that existing compression technologies often use regional compression, transform coding and other methods. Such methods do not change the essence that CT data cannot occupy all registers. Therefore, when the data volume is large, the existing compression technology has a large amount of register redundancy, resulting in a waste of hardware resources; the existing transmission bandwidth cannot meet the growing demand for CT data transmission. In addition, since the dynamic range of actual CT data is much smaller than the maximum dynamic range that can be represented by the register, when the CT data scale is large, there will be a large amount of register redundancy, resulting in a waste of hardware resources.

Summary of the invention

In order to solve the defects and shortcomings in the above-mentioned prior art, the present invention provides a CT detector data compression method based on register redundancy, which decomposes the original data into initial values and difference coefficients by data preprocessing. Since the dynamic range of the difference coefficient is small and has a large register redundancy, a register disassembly method is adopted to use a single register to store multiple difference coefficients, thereby realizing effective utilization of register redundancy, making full use of hardware resources, effectively improving compression efficiency, alleviating the pressure of transmission bandwidth and storage space, and improving transmission reliability.

The technical solution adopted by the present invention to solve the technical problem is: a CT detector data compression method based on register redundancy, comprising the following steps:

S1. Data preprocessing: First, calculate the minimum value of each dimension of the original data as the initial value; second, calculate the difference coefficient of each original data, the difference coefficient is the square root of the difference between the original data and the initial value, and the square root is rounded off and converted into integer data; finally, decompose each original data into the form of initial value and difference coefficient;

S2. Data compression: By disassembling the register, multiple difference coefficients are transmitted using one register, so as to make full use of the register redundant resources;

S3. Data transmission: The compressed data is transmitted through the slip ring;

S4. Data decompression: restore the original data using the transmitted initial value and difference coefficient;

S5. Image reconstruction: Reconstruct the image using the decompressed data.

Preferably, the meaning of taking the minimum value as the initial value in step S1 is to ensure that the sign bits of subsequent calculated values are unified, and are saved in an unsigned data format, thereby increasing the representation range of the data.

Preferably, the significance of taking the square root of the difference coefficient and then rounding it off in step S1 is to reduce the dynamic range of the data as much as possible without losing accuracy and increase register redundancy.

Preferably, the specific formula corresponding to the data preprocessing in step S1 is:

c _k =min(a _nmk )

Among them, c _k is the minimum value of each dimension of the original data, a _nmk is the original data collected by the detector, min means taking the minimum value of each dimension in the original data, b _nmk is the difference coefficient between the original data a _nmk and the initial value c _k , and round means rounding.

Preferably, in step S2, by disassembling the register, the 16-bit register that was originally only used to store and transmit a single data can be used to store and transmit multiple difference coefficients, thereby making full use of hardware resources.

Preferably, the number of bits of the register decomposition in step S2 can be comprehensively considered based on factors such as the actual data range and the acceptable error range. If a larger error can be accepted, the 16-bit register can be decomposed into four 4-bit registers, and the 16-bit register originally used to transmit a single data can transmit four data.

Preferably, the calculation formula for data decompression in step S4 is:

Among them, a′ _nmk is the decompressed data, c _k is the initial value, and b _nmk is the difference coefficient.

The present invention provides an efficient and reliable CT data compression method, which increases the redundancy of registers by data preprocessing, and then realizes the effective utilization of register resources by register disassembly, fully utilizes hardware resources, effectively improves compression efficiency, and can be used to alleviate the pressure of transmission bandwidth and storage space, fully utilize hardware resources, and improve transmission reliability.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic diagram of a data transmission process of the present invention;

FIG2 is a schematic diagram of data preprocessing of the present invention;

FIG3 is a schematic diagram of data compression and decompression according to the present invention.

DETAILED DESCRIPTION

The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

Example 1

A CT detector data compression method based on register redundancy, the basic process is shown in FIG1 , and the method mainly includes the following steps:

Data preprocessing: This step is shown in Figure 2. First, the minimum value of each dimension of the original data is calculated as the initial value. The significance of taking the minimum value as the initial value is to ensure that the sign bit of the subsequent calculated value is unified, and it can be saved in an unsigned data format to increase the data representation range. Secondly, the difference coefficient of each original data is calculated. The difference coefficient takes the square root of the difference between the original data and the initial value, and the square root is rounded and converted into integer data. The significance of taking the square root of the difference coefficient and then rounding it is to reduce the dynamic range of the data as much as possible without losing accuracy. A smaller data range means a larger register redundancy, and a larger register redundancy is more conducive to the subsequent register disassembly. Finally, each original data is decomposed into the form of an initial value and a difference coefficient.

If the original data is K-channel data of size n*m, the specific formula corresponding to the data preprocessing is:

c _k =min(a _nmk )

Among them, c _k is the minimum value of each dimension of the original data, a _nmk is the original data collected by the detector, min means taking the minimum value of each dimension in the original data, b _nmk is the difference coefficient between the original data a _nmk and the initial value c _k , and round means rounding off.

Data compression: This step is shown in Figure 3, and its main principles are:

The present invention adopts the form of decomposing the original data into an initial value and a difference coefficient, so that only the initial value and the difference coefficient need to be transmitted during the transmission process, further increasing the difference between the transmission value and the maximum value that the register can represent, and improving the redundancy of the register. Afterwards, by disassembling the register, the redundant resources of the register are fully utilized, so that the 16-bit register that can only be used to store and transmit a single data can be used to store and transmit multiple difference coefficients, making full use of hardware resources. It should be noted that the number of bits of the register disassembly can be comprehensively considered according to factors such as the actual data range and the acceptable error range. For example, the 16-bit register can be disassembled into high 8 bits and low 8 bits, and the high 8 bits and low 8 bits can be used to store and transmit a difference coefficient respectively, that is, the 16-bit register originally used to transmit a single data transmits two data. In addition, if a larger error can be accepted, the 16-bit register can also be disassembled into 4 4-bit registers, and the 16-bit register originally used to transmit a single data transmits 4 data.

Data transmission: The compressed data is transmitted through the slip ring.

Data decompression: This step is shown in Figure 3, and its main principles are:

The data transmitted through the slip ring is the initial value and the difference coefficient. The calculation formula for data decompression is:

Among them, a′ _nmk is the decompressed data, c _k is the initial value, and b _nmk is the difference coefficient.

Image reconstruction: Image reconstruction is achieved using decompressed data.

The data compression and decompression in the present invention specifically refers to: after preprocessing the original data, the initial value and difference coefficient corresponding to the original data are obtained, and then the register originally used to store and transmit the original data is split, and multiple difference coefficients are merged and stored in the original register. After the transmission is completed, the original data is restored using the initial value and difference coefficient.

Example 2

The following are some numerical examples of the present invention:

The collected original CT data is 4*4 4-channel data, using uint16 data type and 16-bit register storage. The specific values are as follows:

1 channel:

[58933 23352 12691 05711]

[27318 15776 60002 12471]

[49877 05303 46007 49865]

[34201 46955 24589 06670]

2 channels:

[65515 41558 16645 02770]

[53144 22331 38946 12626]

[33492 65142 58509 34918]

[13609 42656 64524 32210]

3 channels:

[42019 18651 49474 52413]

[54142 35131 03190 17814]

[16958 16618 40562 10448]

[26356 33360 16613 45530]

4 channels:

[19545 48309 28979 27304]

[13989 22749 05304 41803]

[15383 55080 42930 39026]

[16872 17848 43679 24329]

Data preprocessing:

Calculate the minimum value of each channel of the original data and record it as the initial value, which are: 5303, 2770, 3190, 5304

Calculate the difference between each original data and the initial value, record it as the difference coefficient, then the difference coefficient of the above original data is:

1 channel:

[232 134 086 020]

[148 102 234 085]

[211 000 202 211]

[170 204 139 037]

2 channels:

[250 197 118 000]

[224 140 190 099]

[175 250 236 179]

[104 200 249 172]

3 channels:

[197 124 215 222]

[226 179 000 121]

[117 116 193 085]

[152 174 116 206]

4 channels:

[119 207 154 148]

[093 132 000 191]

[100 223 194 184]

[108 112 196 138]

Data Compression:

The original 16-bit register is decomposed into high 8 bits and low 8 bits, which are used to store and transmit a difference coefficient respectively. For example, the first two difference coefficients in channel 1 are 232 and 134 respectively, and the storage forms of the original difference coefficients in the 16-bit register are: 0000 0000 1110 1000, 0000 0000 1000 0110. At this time, the high 8-bit register value of the 16-bit register is 0, which has no practical significance and is register redundancy. The compression method of the present invention is to store the effective bits of the difference coefficients 232 and 134 in the high 8 bits and low 8 bits of the 16-bit register respectively. Therefore, the storage form in the 16-bit register of the present invention is: 11101000 1000 0110. At this time, the highest bit register value in the 16-bit register is not 0, so each bit register has practical significance, and the 16-bit register that originally stores single data is used to store two difference coefficients. In addition, the decomposition of the original register and how to store and transmit the difference coefficient need to be changed according to the actual situation.

Data transmission: Transmit the initial value and difference coefficient.

Data decompression: Restore the original data based on the transmitted initial value and difference coefficient. For example, if the initial value of channel 1 is 5303, and the difference coefficients stored in the high 8 bits and low 8 bits of a 16-bit register are 232 and 134 respectively, then the original data are 5303+232*232 and 5303+134*134, that is, 59182 and 22857 respectively.

Image reconstruction: Reconstruct the image based on the original data restored by decompression.

In order to better illustrate that the present invention has the same compression rate for different CT raw data, 6670 with smaller raw data and 65515 with larger raw data are selected as examples. According to the data preprocessing method, the corresponding channel initial values are 5303 and 2270, and the difference coefficients are 37 and 250, respectively. Since the data transmitted in the channel of the present invention are the initial value and the difference coefficient, and the transmission bandwidth required for the initial value of each channel is 16 bits, and the transmission bandwidth required for each difference coefficient is 8 bits. Therefore, the present invention has the same compression rate for CT raw data of different values.

In addition, for CT raw data of different lengths, assuming that the size of the CT raw data is n*m*k, the required transmission bandwidth is n*m*k*16, and taking the decomposition of a 16-bit register into high 8 bits and low 8 bits as an example, the required transmission bandwidth of the present invention is k*16+n*m*k*8. Since the k value is generally much smaller than the n*m*k value, the present invention has approximately the same compression rate for CT raw data of different lengths.

To prove the effectiveness of the present invention, since CT raw data is closer to image data, the classic JPEG compression method in the field of image data compression is selected for comparison. Since the compression rate of the JPEG compression method is variable, while the compression rate of the present invention is fixed, in order to compare in the same dimension, the compression rate of the JPEG compression method is adjusted to the same as that of the present invention, and then the CT detector data compression method provided by the present invention is used for compression, and compared with the JPEG compression method, and the average value is taken after 10 tests.

Table (1) is a quantitative index comparison between the CT detector data compression method provided by the present invention and the JPEG compression method. As shown in Table 1, the experimental results show that the CT detector data compression method provided by the present invention is superior to the JPEG compression method in terms of RMSE, PSNR, and MSE evaluation indicators. Therefore, under the same compression rate, the error of the CT detector data compression method provided by the present invention is smaller than that of the JPEG compression method.

In addition, when the RMSE value of the JPEG compression method is adjusted to be the same as the RMSE value of the CT detector data compression method provided by the present invention, the compression rate of the JPEG compression method is 65.52%, while the compression rate of the present invention is 50.01%. Therefore, under the same error condition, the compression rate of the CT detector data compression method provided by the present invention is better than that of the JPEG compression method.

In summary, the effectiveness of the CT detector data compression method provided by the present invention is demonstrated.

Table (1)

Although specific embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the specific embodiments without departing from the principles and spirit thereof, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims (7)

1. A CT detector data compression method based on register redundancy, characterized in that it comprises the following steps: S1. Data preprocessing: First, calculate the minimum value of each dimension of the original data as the initial value; second, calculate the difference coefficient of each original data, the difference coefficient is the square root of the difference between the original data and the initial value, and the square root is rounded off and converted into integer data; finally, decompose each original data into the form of initial value and difference coefficient; S2. Data compression: By disassembling the register, multiple difference coefficients are transmitted using one register, so as to make full use of the register redundant resources; S3. Data transmission: The compressed data is transmitted through the slip ring; S4. Data decompression: restore the original data using the transmitted initial value and difference coefficient; S5. Image reconstruction: Reconstruct the image using the decompressed data. 2. A CT detector data compression method based on register redundancy according to claim 1, characterized in that: the meaning of taking the minimum value as the initial value in step S1 is to ensure that the sign bits of subsequent calculated values are unified, and are saved in an unsigned data format to increase the data representation range. 3. A CT detector data compression method based on register redundancy according to claim 1, characterized in that: the significance of taking the square root of the difference coefficient and then rounding it in step S1 is to reduce the dynamic range of the data as much as possible without losing accuracy and increase the register redundancy. 4. The CT detector data compression method based on register redundancy according to claim 1, characterized in that: the specific formula corresponding to the data preprocessing in step S1 is: c _k =min(a _nmk ) Among them, c _k is the minimum value of each dimension of the original data, a _nmk is the original data collected by the detector, min means taking the minimum value of each dimension in the original data, b _nmk is the difference coefficient between the original data a _nmk and the initial value c _k , and round means rounding off. 5. The CT detector data compression method based on register redundancy according to claim 1 is preferably characterized in that: in step S2, by means of register disassembly, a 16-bit register that was originally only used to store and transmit a single data can be used to store and transmit multiple difference coefficients, thereby making full use of hardware resources. 6. The CT detector data compression method based on register redundancy according to claim 1 is preferably characterized in that: the number of bits of the register decomposition in step S2 can be comprehensively considered based on factors such as the actual data range and the acceptable error range. If a larger error can be accepted, the 16-bit register can also be decomposed into four 4-bit registers, and the 16-bit register originally used to transmit a single data can transmit four data. 7. The method for compressing CT detector data based on register redundancy according to claim 1 is preferably characterized in that the calculation formula for data decompression in step S4 is: Among them, a′ _nmk is the decompressed data, c _k is the initial value, and b _nmk is the difference coefficient.