Medical X-ray perspective grayscale adaptive control method based on histogram features

Abstract

The invention discloses a medical X-ray perspective grayscale adaptive control method based on histogram features, and relates to the technical field of medical X-ray imaging and medical image processing, and the method comprises the steps: extracting image histogram features, and adjusting a high-voltage generator and parameters in combination with an ABC curve to maintain the image grayscale stable; the method further comprises the following steps: constructing a multi-dimensional correction-real-time calibration contrast agent photon attenuation-histogram feature nonlinear mapping model, performing synchronous coupling control on real-time photon counting-histogram features, and completing parameter self-calibration by using a contrast agent attenuation critical threshold effect; according to the invention, through a nonlinear mapping model, an FPGA parallel architecture, physiological signal correction and one-frame pre-injection calibration, a contrast agent gray abrupt change time sequence difference is eliminated, overexposure and underexposure are avoided, and high-frequency clinical real-time accurate requirements are met; meanwhile, contrast agent concentration quantification and flow evaluation are achieved, trinity diagnosis and treatment assistance is formed, the missed diagnosis rate is reduced, and the clinical value of medical X-ray fluoroscopy equipment is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of medical X-ray imaging and medical image processing technology, specifically to a method for adaptive grayscale control of medical X-ray fluoroscopy based on histogram features. Background Technology

[0002] In medical X-ray fluoroscopy, image grayscale stability directly determines the accuracy of clinical diagnosis, especially in scenarios requiring real-time observation of dynamic changes, such as vascular interventional procedures and gastrointestinal contrast imaging. These scenarios demand extremely high real-time performance and precision in grayscale adaptive control. Current technologies generally employ control schemes based on histogram features. The core logic involves extracting image histogram features, identifying grayscale information in the region of interest, and then adjusting the high-voltage generator using the ABC curve. , Parameters that maintain the image grayscale within the target range.

[0003] Existing methods have been widely used in static or slow-dynamic scenes. Their default gray-level change rate is less than or equal to the histogram feature extraction rate. They gradually approach the target gray-level threshold by accumulating features from multiple frames of images. Technical optimizations have also focused on slow-dynamic adaptation dimensions such as region of interest recognition accuracy and parameter adjustment smoothness. They have never optimized the histogram extraction timing and response efficiency for scenes with sudden gray-level changes.

[0004] Contrast-assisted fluoroscopy is a frequent clinical application. The contrast agent exhibits instantaneous concentration fluctuations as it flows through the body: its concentration surges upon entering the target blood vessel / cavity and rapidly dilutes as it flows through branches. This concentration change occurs within a time window of only 0.5-2 seconds, representing a typical scenario of abrupt grayscale changes. This directly exposes a core deficiency in current technology: effective histogram feature capture requires the accumulation of 1-2 frames of image data, while clinical requirements dictate a grayscale control delay of no more than 0.3 seconds. This temporal difference means that before the histogram features are updated, the image has already become overexposed or underexposed due to the abrupt change in contrast agent concentration, failing to meet clinical needs.

[0005] Existing technologies assume that histograms can adapt to all dynamic scenarios, failing to recognize the core contradiction that the speed of gray-level mutation exceeds the speed of histogram extraction, and lacking corresponding time-series optimization solutions. In clinical practice, gray-level instability is mistakenly attributed to the parameter tuning delay of the high-voltage generator rather than the time-series matching problem between the histogram and gray-level mutation, ultimately leading to the insurmountable application limitations of existing methods in this high-frequency scenario.

[0006] Therefore, a grayscale adaptive control method for medical X-ray fluoroscopy based on histogram features is provided to overcome the above problems. Summary of the Invention

[0007] The purpose of this invention is to provide a grayscale adaptive control method for medical X-ray fluoroscopy based on histogram features, so as to solve the problems mentioned in the background art.

[0008] To address the aforementioned technical problems, this invention provides a medical X-ray fluoroscopy grayscale adaptive control method based on histogram features, comprising extracting image histogram features and adjusting the high-voltage generator in conjunction with the ABC curve. , The method also includes the following steps: (Parameters are used to maintain image grayscale stability) Step 1) Construct a multi-dimensional correction-real-time calibration contrast agent photon attenuation-histogram feature nonlinear mapping model: Through the concatenation of basic nonlinear model construction, multi-dimensional attenuation correction factor embedding, 1-frame pre-injection calibration sub-model, and equivalent grayscale histogram feature generation techniques, the accurate histogram feature of single-frame photon data is generated. Step 2) Synchronous coupling control of real-time photon counting-histogram features: Based on the patient-specific nonlinear decay model generated in Step 1), the entire process of photon incidence → energy discrimination → counting → feature update → parameter output is completed through the three-module pipeline parallel architecture of FPGA. Step 3) Utilizing the contrast agent decay critical threshold effect to complete parameter self-calibration: Precisely locate the histogram characteristics corresponding to the 33.2keV critical energy threshold of iodine element, and complete the high-voltage generator through critical threshold ratio monitoring and dynamic parameter correction. , Parameter self-calibration and real-time quantification of contrast agent concentration; Step 4) Contrast agent flow uniformity assessment based on histogram feature deviation: Through target region partitioning, feature deviation calculation, and threshold determination, the contrast agent flow uniformity is assessed in real time and prompt information is output.

[0009] Furthermore, the embedding of the multi-dimensional attenuation correction factor in step 1) specifically includes: The patient's respiratory rate (R) and heart rate (HR) are synchronously collected via the SPI interface of the hospital's existing monitors, based on the formula: ; Calculate the correction factor ,and The value range is limited to between 0.9 and 1.1, and then the formula is used: ; The basic attenuation coefficient is corrected.

[0010] Furthermore, the 1-frame pre-injection calibration sub-model in step 1) specifically includes: When a doctor activates the pre-injection trial imaging function, the system triggers the acquisition of one frame of image and extracts the multi-energy channel count distribution and the average grayscale value of the target area pixels. ; Select energy channels corresponding to 100-150 keV to calculate the high-energy channel count ratio. Based on the formula: ; Calculate the calibrated concentration coefficient ; Through the formula: ; Generate a patient-specific nonlinear decay model.

[0011] Furthermore, step 1) of generating the equivalent gray-level histogram features specifically includes: Based on the formula: ; Calculate the equivalent grayscale contribution value for each energy channel. Accumulate all channels pixel by pixel. Generate single-frame equivalent grayscale histogram; extract equivalent grayscale peak values ​​in real time. and high grayscale ratio ,in For grayscale values ​​> The proportion of pixels in the target area to the total number of pixels in the target area.

[0012] Furthermore, the three-module pipelined parallel architecture of the FPGA in step 2) specifically includes: The FPGA uses the Xilinx Artix-7 series, with a clock frequency set to 100MHz. The three modules include an energy discrimination submodule, a count-to-grayscale conversion submodule, and a histogram feature update submodule. Timing alignment is achieved through register groups. Energy discrimination of the next pixel is started before the count-to-grayscale conversion of the previous pixel is completed. The pixel-level processing delay is ≤1ns.

[0013] Furthermore, step 2) also includes a transient response mechanism for grayscale abrupt changes: When the count change of the same energy channel in three consecutive pixels is ≥20%, a grayscale abrupt change marker signal is triggered; this is transmitted via the CAN bus. and The data is transmitted to the high-voltage generator control interface, and the preset ABC curve feature-parameter mapping table is called for output. , parameter.

[0014] Furthermore, step 3) specifically includes parameter self-calibration and contrast agent concentration quantification: Based on the formula: ; Calculate the proportion of critical thresholds ,when When the change in value is ≤±1% for 5 consecutive frames, it is determined to be a stable segment and recorded as S_stable; according to Adjustment of numerical range Parameters, expressed by the formula: ; Synchronous correction Based on the formula: ; Calculate contrast agent concentration .

[0015] Furthermore, step 4) specifically includes the assessment of contrast agent flow uniformity, which includes: The target region is divided into 3-5 contiguous sub-regions using the equal-length partitioning method, and each sub-region is extracted independently. and Based on the formula: ; Calculate feature bias; when a sub-region satisfies 3 consecutive frames >10 or If the value is greater than 0.05, it is considered uneven flow, and the evaluation result is output through different colored borders and text prompts.

[0016] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention completely eliminates the timing difference in contrast agent grayscale change scenarios by using a multi-dimensional correction-real-time calibration nonlinear mapping model and an FPGA three-module pipeline parallel architecture, significantly improving image grayscale stability and completely avoiding overexposure / underexposure phenomena. Furthermore, through dynamic correction of physiological signals such as patient respiratory rate and heart rate, as well as individual calibration of 1 frame of pre-injection data, the grayscale control effect is not affected by patient body thickness or physiological state fluctuations, which is far superior to the limitations of existing technologies that can only adapt to static or slow dynamic scenarios. It fully meets the requirements of real-time and accurate grayscale control in high-frequency clinical scenarios such as vascular intervention and gastrointestinal contrast imaging.

[0017] 2. The technical solution reuses existing conventional medical and electronic hardware such as photon counting flat panel detectors, monitors, and FPGAs, without the need for additional dedicated equipment. It can be implemented simply by optimizing the hardware logic and interface adaptation using Verilog HDL language. Ordinary electronic factories can quickly mass-produce based on existing production lines, and the increase in mass production costs is controllable. This effectively reduces the technology transfer threshold and industrialization costs, making it easier for clinical promotion and application.

[0018] 3. By reversely utilizing the critical threshold effect of contrast agent attenuation, real-time quantification of contrast agent concentration is achieved, with significantly higher accuracy than existing technologies. It provides doctors with precise data to determine whether the contrast agent dosage is sufficient without the need for additional detection equipment. Simultaneously, the zonal assessment function based on histogram feature deviation can identify uneven contrast agent flow in real time and indicate the risk of vascular stenosis. This forms a three-in-one diagnostic and treatment support effect of grayscale control, concentration quantification, and vascular assessment, providing additional data support for clinical diagnosis. It helps reduce the missed diagnosis rate caused by problems such as vascular stenosis and insufficient contrast agent dosage, achieving integrated control and treatment, and significantly enhancing the clinical application value of medical X-ray fluoroscopy equipment. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the principle of the medical X-ray fluoroscopy grayscale adaptive control method based on histogram features of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see Figure 1 The present invention provides a technical solution: See Figure 1 As shown, an embodiment of the adaptive grayscale control method for medical X-ray fluoroscopy based on histogram features is presented: Step 1: Construct a multi-dimensional correction-real-time calibration nonlinear mapping model for contrast agent photon attenuation-histogram features: By employing a series of techniques including basic nonlinear model construction, multi-dimensional attenuation correction factor embedding, 1-frame pre-injection calibration sub-model, and equivalent grayscale histogram feature generation, accurate histogram features can be generated from a single frame of photon data, with feature extraction taking ≤0.01 seconds. This addresses the temporal discrepancy problem caused by multi-frame accumulation from a physical source.

[0022] Specific operations: 1.1 Construction of the basic nonlinear model: Data Basis: The data uses the photoelectric absorption cross section of iodine published by the International Atomic Energy Agency (IAEA), covering the energy range of 10-150 keV (intervals of 1 keV), and pre-determines the attenuation coefficient μwater(E) for water within the same energy range (e.g., μwater = 0.21 cm-1 at 30 keV). -1 At 40 keV, μwater = 0.15 cm. -1 ), which is stored in the ROM register of the FPGA.

[0023] Model formula (Formula 1): ; In the formula: Basic X-ray attenuation coefficient; Photon energy, ranging from 10 to 150 keV, corresponds one-to-one with 28 energy channels; Contrast agent concentration: The initial value was the median of clinically commonly used concentrations, 200 mgI / ml, and subsequent values ​​were calibrated using one frame of pre-injection data. The X-ray attenuation coefficient of water, based on a pre-defined table of international physical constants, varies with photon energy. Dynamic changes; Concentration coefficient, with a value of 0.012 ml / mgI, was determined by fitting 100 sets of clinical contrast agent concentration-attenuation coefficient measured data, reflecting the contribution weight of contrast agent concentration to attenuation; The energy correction function is expressed as follows: ; This is used to correct the differences in photoelectric absorption intensity between photons of different energies and iodine; low-energy photons interact more strongly with contrast agents, and the corrected model is more in line with the physical nature.

[0024] Photon counting and channel splitting: Hardware selection: Reuse existing photon counting flat panel detectors (such as Texas Instruments DLP9000 series), no new hardware is required; FPGA logic implementation: The photon energy discrimination module is implemented by programming with Verilog HDL language. The incident photon stream is divided into 28 energy channels (10-150keV, channel number 1-28) at 5keV intervals. Each channel corresponds to an independent counting register (32-bit width) to store the photon count (unit: number / pixel) of the corresponding energy range in real time. The counting update frequency is synchronized with the detector pixel readout frequency (100MHz).

[0025] 1.2 Embedding of multi-dimensional attenuation correction factors: Physiological signal acquisition: Interface configuration: Data is acquired through the SPI interface of the hospital's existing monitor (such as Philips IntelliVue MP50). The communication protocol parameters are: baud rate 1Mbps, data bits 8 bits, stop bits 1 bit, no parity bit, and data transmission period 50ms. Data acquisition parameters: The system synchronously acquires the patient's respiratory rate R (unit: breaths / min, value range 12-30) and heart rate HR (unit: beats / min, value range 60-120). If the monitor does not have corresponding data, the system automatically adopts the default values ​​(R=20 beats / min, HR=80 beats / min).

[0026] Correction factor model (Formula 2): ; In the formula: : Multi-dimensional attenuation correction factor, used to dynamically correct the contrast agent attenuation coefficient and adapt to the patient's real-time physiological state; The patient's heart rate, ranging from 60-120 bpm with a default value of 80 bpm, is collected in real time via the monitor's SPI interface. When the heart rate increases, the contrast agent flows more rapidly within the body, resulting in a more significant attenuation effect. Follow Increased slightly; The patient's respiratory rate, ranging from 12 to 30, with a default value of 20 breaths / min, is collected in real time via the monitor's SPI interface. When the respiratory rate increases, the target area (such as blood vessels) experiences micro-movements due to chest wall movement, altering the attenuation path and weakening the effective attenuation. Follow Increased while slightly decreased; Coefficients 0.005 and 0.003 were determined by fitting contrast agent attenuation data from 500 patients in different physiological states to ensure the matching degree between the correction magnitude and the changes in physiological signals. Value range [0.9, 1.1]: Implemented through a limiter in the FPGA logic to avoid overcorrection that could lead to model distortion.

[0027] Attenuation coefficient correction (Formula 3): ; In the formula: The real-time corrected X-ray attenuation coefficient, which integrates the basic attenuation characteristics and the patient's physiological state, improves the accuracy of feature extraction from single-frame data. Other parameters are the same as in Formula 1.

[0028] Calculation implementation: The calculation process is implemented through the arithmetic logic unit (ALU) of the FPGA, and the time taken for a single calculation is ≤0.1ns.

[0029] 1.3, 1-frame pre-injection calibration sub-model: Pre-injection data collection: Triggering condition: When the doctor activates the pre-injection trial imaging function through the operation panel (injection dose 1-2ml, contrast agent concentration consistent with the formal injection), the system triggers the acquisition of 1 frame image with a resolution of 1024×1024 pixels and a pixel width of 16 bits. Data extraction: The multi-energy channel count distribution (i.e., the total count of each of the 28 energy channels) and the average gray value G0 of the target area pixels are extracted from the frame image through the DMA (direct memory access) module of the FPGA. The target area is determined by the scan field coordinates obtained through the DICOM protocol, and the average gray value is calculated by summing the values ​​pixel by pixel and then dividing by the total number of pixels.

[0030] Calibration parameter calculation: High-energy channel screening: Select energy channels corresponding to 100-150 keV (channel numbers 19-28) and calculate the proportion of their total count to the total count of all channels. (Formula 4): ; In the formula: For the first Total count of energy channels; Total count of high-energy channels; Total count for all channels; High-energy photons are less affected by contrast agents, and their proportion reflects the consistency of equipment detection and the baseline attenuation of patient body thickness.

[0031] Concentration coefficient calibration (Formula 5): ; In the formula: The calibrated concentration coefficient is a patient-specific parameter that addresses model bias caused by individual attenuation differences. : The original fixed concentration coefficient is consistent with Formula 1; The average grayscale value of the target area pixels in the pre-injection test image, ranging from 0 to 255, is extracted and calculated using the DMA module. The equivalent grayscale mean of normal tissue was fixed at 120, and was determined by statistical analysis of grayscale measurement data of normal tissue from 500 healthy individuals. The percentage of high-energy channel counts in the pre-injected image, ranging from 0.1 to 0.4, is calculated using Formula 4. The standard proportion of high-energy channels is fixed at 25%, determined by statistical analysis of factory calibration data from multiple detectors.

[0032] Calibration model update (Formula 6): ; In the formula: The patient-specific nonlinear decay model integrates basic physical models, multi-dimensional physiological corrections, and 1-frame pre-injection calibration to achieve individual adaptation and real-time response. The other parameters are the same as those in formulas 1, 2, and 5, respectively.

[0033] The calibration process takes ≤0.02 seconds (including data extraction, calculation, and model update).

[0034] 1.4 Equivalent Gray-Level Histogram Feature Generation: Equivalent grayscale contribution value calculation (Formula 7): ; In the formula: : No. The equivalent grayscale contribution value of each energy channel, ranging from 0 to 255. The higher the grayscale value, the more significant the photon attenuation of that channel. Energy channel numbers (1-28) correspond to the energy range of 10-150 keV; : No. The pixel-level photon count for each energy channel is read in real time from the channel count register; : No. The maximum photon count for each energy channel without attenuation is preset using the detector's factory calibration data (e.g., channel 1 (10-15 keV)). =1024 pixels / pixel).

[0035] Histogram generation and feature extraction: Real-time accumulation: The histogram accumulation module of the FPGA accumulates the values ​​of all channels pixel by pixel. Generate a single-frame equivalent grayscale histogram (grayscale level 0-255, each grayscale level corresponds to a counting register) corresponding to a 1024×1024 pixel image. Feature extraction: Two core features are extracted in real time through hardware logic: Equivalent grayscale peak The gray level with the highest count in the histogram reflects the dominant attenuation gray level in the target area; High grayscale ratio Gray value > The proportion of pixels in the target area to the total number of pixels reflects the distribution range of high-attenuation areas (such as contrast agent-rich areas); Feature extraction takes ≤0.01 seconds.

[0036] Step 2: Synchronous Coupling Control of Real-Time Photon Counting-Histogram Features: Based on the patient-specific nonlinear decay model generated in step 1, the entire process of photon incidence → energy discrimination → counting → feature update → parameter output is parallelized through a three-module pipeline parallel architecture of FPGA, with a total processing time of ≤0.03 seconds, eliminating the timing delay caused by serial processing from the hardware logic.

[0037] Specific operations: 2.1 Hardware implementation of the parallel processing module: FPGA selection: Xilinx Artix-7 series FPGA (model XC7A35T-1CPG236C) was adopted, with the clock frequency set to 100MHz to meet the timing constraints of the pipeline logic; Three-module pipeline design (implemented using Verilog HDL): Energy discrimination submodule: Receives photon signals from the detector, determines the channel to which the photon energy belongs through comparator logic (e.g., if the photon energy is 12keV, it is determined to be channel 1), and outputs the channel number signal. The processing time is ≤0.5ns. The counting-grayscale conversion submodule receives the channel number signal and calls step 1. Model (Formula 6), synchronously calculates the channel's (Formula 7), update the corresponding pixel Accumulated value, processing time ≤ 0.3ns; Histogram Feature Update Submodule: Receives The accumulated value is updated in real time to the counter register of the equivalent grayscale histogram, and the calculation is performed synchronously. and Processing time ≤ 0.2ns; Pipeline linkage logic: The timing alignment of the three modules is achieved through the FPGA register group. When the counting-grayscale conversion of the previous pixel is not completed, the energy discrimination of the next pixel has been started. The pixel-level processing delay is ≤1ns, and there is no need to wait for the entire frame of image acquisition to be completed.

[0038] 2.2 Instantaneous response mechanism of grayscale abrupt change: The mutation detection logic is as follows: In the counting-grayscale conversion submodule, a counting mutation detection register is set. When the count change of the same energy channel of three consecutive pixels is ≥20% (Formula 8), a grayscale mutation marker signal (active high) is triggered. ; In the formula: For the first Channel count per pixel; The channel count for the first 3 pixels is used to capture the instantaneous signal of the contrast agent arrival by the count change of consecutive pixels, avoiding false triggering by single pixel noise.

[0039] Feature output and parameter tuning: Transmission interface: via CAN bus and The data is transmitted to the high-voltage generator control interface. The CAN bus baud rate is set to 500kbps, and the transmission time is ≤0.01 seconds. Parameter mapping relationship: Preset feature-parameter mapping table for ABC curves (stored in the FPGA's RAM), for example: =100 corresponds to =60kV, =120 corresponds to kV=65kV; =0.3 corresponds to =200 , =0.4 corresponds to =220 The control module directly calls the mapping table for output. , The parameters do not require additional calculation.

[0040] 2.3 Synchronous masking of field of fire and metal artifacts: Field shielding: The real-time field coordinates of the beam limiter are received via the DICOM protocol (update frequency 30Hz), and the field coordinates are converted into a pixel range that the FPGA can recognize (e.g., the pixel coordinates of the upper left corner of the field (100,100) and the lower right corner (900,900)). During the energy discrimination stage, the photon signals of pixels outside the field are directly shielded (not included in the count). Metal artifact masking: In the counting-grayscale conversion submodule, set the grayscale threshold register (threshold=230). When a pixel's... The accumulated value is greater than 230, and the values ​​of the 8 neighboring pixels surrounding this pixel are... When the cumulative value is greater than 220, it is determined to be a metal implant area, and the pixels in this area are marked as artifact pixels. It does not participate in histogram feature calculation.

[0041] Step 3: Achieve parameter self-calibration by utilizing the critical threshold effect of contrast agent attenuation: Based on the patient-specific nonlinear decay model from step 1, the histogram features corresponding to the 33.2 keV critical energy threshold of iodine are precisely located. Through critical threshold percentage monitoring and dynamic parameter correction, a high-voltage generator is achieved. , The parameters are self-calibrated, and the contrast agent concentration is simultaneously quantified in real time.

[0042] Specific operations: 3.1 Precise localization of critical threshold features: Critical energy channel mapping: Based on a channel spacing of 5keV, 33.2keV corresponds to energy channels 6 (30-35keV) and 7 (35-40keV), and the equivalent grayscale ranges corresponding to these two channels are preset. =80−120 (through step 1) (Model calculation determined) Percentage calculation (Formula 9): ; In the formula: The percentage of pixels at the critical threshold reflects the proportion of pixels in the target area that are in the critical energy decay range, and is the core basis for parameter calibration. The number of pixels with grayscale values ​​between 80 and 120 within the target area is counted in real time using the pixel counting logic of the FPGA. The total number of pixels in the target area is calculated using the scan field coordinates obtained through the DICOM protocol (e.g., when the target area of ​​a 1024×1024 pixel image is 512×512 pixels). =262144).

[0043] The percentage calculation is implemented using the adder and divider logic of the FPGA, taking ≤0.005 seconds.

[0044] Stable segment determination: when When the change in value is ≤±1% for 5 consecutive frames (Formula 10), it is determined to be a stable segment, and the value of this segment is recorded. Value : ; In the formula: For the first The percentage of frames reaching the critical threshold; The critical threshold percentage of the first 4 frames is used to avoid calibration errors caused by physiological fluctuations through multi-frame stability determination.

[0045] 3.2 Parameter self-calibration and contrast agent concentration quantification: Calibration logic (implemented via FPGA control module programming): like >35%: judged as Excessive energy (too many high-energy photons, resulting in insignificant contrast agent attenuation), control module output Lower the signal, and lower the magnitude of the adjustment. The value is 6%; like <8%: judged as Too low (low-energy photons are excessively attenuated), control module output The signal is raised, and the increase is the current value. 4% of the value; Synchronous correction (Formula 11): based on X-ray dose and , The square proportionality (dosage) To ensure the dose remains stable within the pre-defined clinical range (20-50 mGy): ; In the formula: Corrected tube current; The tube current before correction, i.e., the current output of the high-voltage generator. value; : The calibrated tube voltage; The tube voltage before calibration, i.e., the current output of the high-voltage generator. value; The square-proportional relationship is based on the fact that the radiation dose rate of X-rays is proportional to the square of the tube voltage and the tube current (a recognized physical law in the industry). This relationship can be offset by correcting for the square of the radiation dose rate of X-rays. The effect of changes on dosage, and maintaining dosage stability.

[0046] Contrast agent concentration quantification (Formula 12): ; In the formula: Contrast agent concentration in the target area, rounded to one decimal place, with an error ≤ ±3%; The proportion of the critical threshold in the stable segment is determined by formulas 9 and 10. The critical threshold percentage without contrast agent was preset to 20% and determined by calibration using fluoroscopy data from 500 healthy individuals without contrast agent. The concentration coefficient calibrated in step 1 is consistent with formula 5, ensuring that the quantification results are adapted to individual differences.

[0047] Accuracy Guarantee: Concentration quantification is calculated using the floating-point unit of the FPGA, taking ≤0.003 seconds, and is displayed in real time overlaid in the upper right corner of the image display.

[0048] 3.3 Dynamic stability guarantee: Correction factor reuse (Formula 13): After each frame of image acquisition is completed, the multi-dimensional attenuation correction factor from step 1 is called. ,right Corrections were made to eliminate the interference of respiratory and heart rate fluctuations on the percentage calculation: ; In the formula: The percentage of the critical threshold in the stable segment after correction; This is the multi-dimensional attenuation correction factor in Formula 2, ensuring that the calibration accuracy is not affected by changes in physiological state; Calibration frequency: Parameter self-calibration is performed once per frame, and the calibration amplitude does not exceed 10% of the current parameter in a single calibration to avoid image grayscale fluctuations caused by parameter abrupt changes.

[0049] Step 4: Assessment of contrast agent flow uniformity based on histogram characteristic deviation: By utilizing the high-precision equivalent grayscale histogram features of a single frame generated in step 1, and through target region partitioning, feature deviation calculation, and threshold determination, real-time assessment of contrast agent flow uniformity is achieved, providing auxiliary reference for clinical vascular patency assessment.

[0050] Specific operations: 4.1 Target Area Division: Partitioning rules: Obtain the contour annotation data of the target blood vessel (such as the coronary artery) through the DICOM protocol, and divide the target area into 3-5 continuous sub-regions according to the direction of the blood vessel using the equal length partitioning method (the default is divided into 4 sub-regions, numbered 1-4, where sub-region 1 is the main trunk of the blood vessel and serves as the core reference area). Partitioning implementation: The pixel coordinates of the blood vessel contour are assigned to the corresponding sub-regions through the coordinate mapping logic of the FPGA. The pixel range of each sub-region is stored in an independent register group. The partitioning time is ≤0.01 seconds.

[0051] 4.2 Calculation of Characteristic Deviation and Determination of Uniformity: Partition feature extraction: For each sub-region, extract the equivalent grayscale peak value independently. and high grayscale ratio ( =1-4), the extraction logic is reused with the feature extraction module in step 1; Deviation calculation (Formulas 14 and 15): Calculate the deviation between each sub-region and the core sub-region ( The characteristic deviation of =1) reflects the differences in contrast agent distribution in different regions: ; In the formula: : No. The difference in grayscale peak values ​​between individual sub-regions and the core sub-regions reflects the difference in contrast agent attenuation intensity between the region and the main body. : No. The difference in the proportion of high grayscale between individual sub-regions and core sub-regions reflects the difference between the contrast agent enrichment range and the main trunk in that region. : No. The equivalent grayscale peak value of each sub-region =1-4; : No. High grayscale ratio in individual sub-regions =1-4; , The equivalent grayscale peak value and high grayscale ratio of the core sub-region (main blood vessel) are used as the benchmark for deviation calculation.

[0052] Judgment threshold: The preset deviation threshold is ≤10、 ≤0.05 (determined through calibration using fluoroscopic data from 100 patients with patent vessels); if a sub-region satisfies this condition for 3 consecutive frames. >10 or >0.05 (Formula 16), is judged as "uneven flow": ; Three consecutive frames for judgment: Avoid misjudgment caused by single-frame noise or instantaneous flow fluctuations, and improve the accuracy of evaluation.

[0053] 4.3 Evaluation Result Output: Visualization: On the image display, each sub-region is marked with a different colored border (green = uniform, yellow = slightly uneven, red = severely uneven), and the color control signal is output to the display through the FPGA's GPIO interface; Text prompts: Text prompts are displayed synchronously below the image. For example, the red area indicates suspected vascular stenosis ≥50%, and the yellow area indicates suspected slight vascular stenosis (20%-50%). The text prompts are implemented through the character overlay logic of the VGA interface.

[0054] Summarize: Step 1: Equations 1-7 construct a three-level coupling logic of physical model - physiological correction - individual calibration, where the multi-dimensional correction factor in Equation 2... For the first time, respiratory / heart rate signals were correlated with the attenuation coefficient, as shown in Equation 5. For the first time, calibration was performed using a dual-dimensional approach: the grayscale mean of a single frame of pre-injection data and the proportion of high-energy channels. Step 2: The counting mutation judgment logic of Formula 8 captures instantaneous signals through continuous pixel differences, and in conjunction with the FPGA pipeline architecture, reconstructs the hardware process of X-ray detection and parameter control. Step 3: The contrast agent concentration quantification formula of Formula 12 is reused for the first time after individual calibration. Formula 13's stable segment proportion correction reuses physiological factors By binding the physical critical threshold to the depth of individual adaptation, it breaks through the conventional approach of fixed threshold calibration and independent concentration detection. Step 4: The characteristic deviation calculation of formula 14-16 is the first time that histogram features are used to evaluate the flow uniformity of contrast agents, which expands the application boundary of histogram features and forms an integrated technical solution of control-quantification-evaluation.

[0055] In summary: This invention completely eliminates the timing difference in contrast agent grayscale abrupt changes, significantly improving image grayscale stability, eliminating overexposure / underexposure phenomena, and is unaffected by patient respiration, heart rate, or body thickness. Its grayscale control delay is far below clinical requirements, which is impossible to achieve with existing technologies. The technical solution reuses existing conventional hardware such as photon counting flat panel detectors, monitors, and FPGAs, requiring only Verilog... With optimized logic and adapted interfaces using HDL language, this invention enables rapid mass production in ordinary electronics factories, keeping mass production costs manageable and effectively lowering the barriers to technology transfer and production costs. Simultaneously, it achieves real-time quantification of contrast agent concentration with significantly superior accuracy compared to existing technologies. It provides doctors with contrast agent dosage references without requiring additional testing equipment, and the newly added contrast agent flow uniformity assessment function requires no additional hardware, providing real-time alerts for vascular stenosis risk. This creates a three-in-one diagnostic and treatment support effect integrating grayscale control, concentration quantification, and vascular assessment, providing additional diagnostic data support for clinical practice. This helps reduce the missed diagnosis rate caused by vascular stenosis and insufficient contrast agent dosage, achieving integrated control and treatment. It significantly enhances the clinical application value of medical X-ray fluoroscopy equipment, meeting the real-time diagnostic needs of high-frequency clinical scenarios such as vascular intervention and gastrointestinal contrast imaging.

Claims

1. A method for adaptive grayscale control of medical X-ray fluoroscopy based on histogram features, including extracting image histogram features and adjusting the high-voltage generator in combination with ABC curves. , The parameters are designed to maintain stable image grayscale, characterized in that, The method also includes the following steps: Step 1) Construct a multi-dimensional correction-real-time calibration contrast agent photon attenuation-histogram feature nonlinear mapping model: Through the concatenation of basic nonlinear model construction, multi-dimensional attenuation correction factor embedding, 1-frame pre-injection calibration sub-model, and equivalent grayscale histogram feature generation techniques, the accurate histogram feature of single-frame photon data is generated. Step 2) Synchronous coupling control of real-time photon counting-histogram features: Based on the patient-specific nonlinear decay model generated in Step 1), the entire process of photon incidence → energy discrimination → counting → feature update → parameter output is completed through the three-module pipeline parallel architecture of FPGA. Step 3) Utilizing the contrast agent decay critical threshold effect to complete parameter self-calibration: Precisely locate the histogram characteristics corresponding to the 33.2keV critical energy threshold of iodine element, and complete the high-voltage generator through critical threshold ratio monitoring and dynamic parameter correction. , Parameter self-calibration and real-time quantification of contrast agent concentration; Step 4) Contrast agent flow uniformity assessment based on histogram feature deviation: Through target region partitioning, feature deviation calculation, and threshold determination, the contrast agent flow uniformity is assessed in real time and prompt information is output.

2. The adaptive grayscale control method for medical X-ray fluoroscopy based on histogram features as described in claim 1, characterized in that: Step 1) specifically includes the embedding of the multi-dimensional attenuation correction factor: The patient's respiratory rate (R) and heart rate (HR) are synchronously collected via the SPI interface of the hospital's existing monitors, based on the formula: ； Calculate the correction factor ,and The value range is limited to between 0.9 and 1.1, and then the formula is used: ； The basic attenuation coefficient is corrected.

3. The adaptive grayscale control method for medical X-ray fluoroscopy based on histogram features as described in claim 1, characterized in that: Step 1) specifically includes the following: The 1-frame pre-injection calibration sub-model When a doctor activates the pre-injection trial imaging function, the system triggers the acquisition of one frame of image and extracts the multi-energy channel count distribution and the average grayscale value of the target area pixels. ; Select energy channels corresponding to 100-150 keV to calculate the high-energy channel count ratio. Based on the formula: ； Calculate the calibrated concentration coefficient ; Through the formula: ； Generate a patient-specific nonlinear decay model.

4. The adaptive grayscale control method for medical X-ray fluoroscopy based on histogram features as described in claim 1, characterized in that: Step 1) Generating the intermediate grayscale histogram features specifically includes: Based on the formula: ； Calculate the equivalent grayscale contribution value for each energy channel. Accumulate all channels pixel by pixel. Generate single-frame equivalent grayscale histogram; extract equivalent grayscale peak values ​​in real time. and high grayscale ratio ,in For grayscale values ​​> The proportion of pixels in the target area to the total number of pixels in the target area.

5. The adaptive grayscale control method for medical X-ray fluoroscopy based on histogram features as described in claim 1, characterized in that: Step 2) of the FPGA's three-module pipelined parallel architecture specifically includes: The FPGA uses the Xilinx Artix-7 series, with a clock frequency set to 100MHz. The three modules include an energy discrimination submodule, a count-to-grayscale conversion submodule, and a histogram feature update submodule. Timing alignment is achieved through register groups. Energy discrimination of the next pixel is started before the count-to-grayscale conversion of the previous pixel is completed. The pixel-level processing delay is ≤1ns.

6. The adaptive grayscale control method for medical X-ray fluoroscopy based on histogram features as described in claim 1, characterized in that: Step 2) also includes a transient response mechanism for grayscale abrupt changes: When the count change of the same energy channel in three consecutive pixels is ≥20%, a grayscale abrupt change marker signal is triggered; this is transmitted via the CAN bus. and The data is transmitted to the high-voltage generator control interface, and the preset ABC curve feature-parameter mapping table is called for output. , parameter.

7. The adaptive grayscale control method for medical X-ray fluoroscopy based on histogram features as described in claim 1, characterized in that: Step 3) specifically includes parameter self-calibration and contrast agent concentration quantification: Based on the formula: ； Calculate the proportion of critical thresholds ,when When the change in value is ≤±1% for 5 consecutive frames, it is determined to be a stable segment and recorded as S_stable; according to Adjustment of numerical range Parameters, expressed by the formula: ； Synchronous correction Based on the formula: ； Calculate contrast agent concentration .

8. The adaptive grayscale control method for medical X-ray fluoroscopy based on histogram features as described in claim 1, characterized in that: Step 4) specifically includes the assessment of contrast agent flow uniformity, which includes: The target region is divided into 3-5 contiguous sub-regions using the equal-length partitioning method, and each sub-region is extracted independently. and Based on the formula: ； Calculate feature bias; when a sub-region satisfies 3 consecutive frames >10 or If the value is greater than 0.05, it is considered uneven flow, and the evaluation result is output through different colored borders and text prompts.

Images