Multi-gain flat panel detector and reading data processing method

Abstract

The present invention discloses a multi-gain flat-panel detector and a reading data processing method, belonging to the field of image processing. The multi-gain flat-panel detector includes a scintillator layer and a reading circuit layer. Before reading, the reading circuit layer detects and corrects the signal and then outputs the image. Each pixel of the read image corresponds to its own independent thin-film transistor array, control circuit, and analog-to-digital converter, thereby achieving independent pixel reading. In addition, the overall capacitance of the charge amplifier can be adjusted by software control of the switching of the circuit containing the individual capacitors in the charge amplifier, thereby achieving control of the reading gain, solving the problems of data saturation and low data signal-to-noise ratio that are prone to occur in flat-panel detectors.

Description

Technical Field

[0001] The present invention relates to the field of image processing, and in particular to a multi-gain flat panel detector and a reading data processing method. Background Art

[0002] Cone-beam computed tomography (CBCT) offers many potential advantages for point-of-care imaging of head trauma, including a small footprint, mechanical simplicity, volumetric acquisition in a single rotation, and isotropic spatial resolution, as exemplified by mobile intraoperative C-arms, dedicated musculoskeletal imaging, breast imaging, and maxillofacial imaging.

[0003] CBCT uses a dynamic flat-panel detector. Under the premise of unchanged dose, the flat-panel detector (FPD) can use signal amplification to reduce the relative contribution of electronic noise through high-gain detector readings, thereby improving image resolution.

[0004] However, there's a trade-off between signal amplification (increasing the signal relative to electronic noise) and detector dynamic range. Due to the detector's limited readout range, signal amplification means the detector reading is more likely to reach saturation, preventing effective transmission. This is especially true in a bare beam near the target object, where absorption along the beam path is minimal and the number of photons reaching the detector surface is high. The detector reading is much higher than that of a beam transmitted through the object, making saturation extremely likely. This trade-off is particularly pronounced when detecting low-contrast features (such as ICH lesions) within a highly attenuating object (e.g., the skull). Summary of the Invention

[0005] The present disclosure provides a multi-gain flat panel detector and a reading data processing method, which can solve the problem in related technologies that the flat panel detector has a limited dynamic range and is prone to high-dose data saturation and low low-dose data signal-to-noise ratio. The technical solution is as follows:

[0006] In one aspect, a multi-gain flat panel detector is provided, the detector comprising: a scintillator layer and a readout circuit layer;

[0007] The reading circuit layer includes a thin film transistor array, a control circuit and an analog-to-digital converter, wherein the thin film transistor array includes a photodiode;

[0008] The scintillator layer is connected to the reading circuit layer, and the scintillator layer is used to convert X-rays into optical signals. The reading circuit layer captures the optical signals through photodiodes in the thin film transistor array and converts the optical signals into electrical signals for storage before reading.

[0009] The thin film transistor array is used to detect the light signal captured by the photodiode before reading; the control circuit is used to sample, correct and multiplex the electrical signal stored before reading; the analog-to-digital converter is used to convert the voltage signal after sampling and multiplexing into a reading output signal.

[0010] In another aspect, a method for processing reading data is provided, the method comprising:

[0011] Acquire an N*N data set according to a reading output signal, wherein the reading output signal is collected by the multi-gain flat panel detector according to any one of claims 1 to 6, wherein N is a positive integer greater than or equal to 2;

[0012] Inputting the data set into a data processing model to obtain N*N images to be processed by the data processing model;

[0013] performing denoising on the N*N images using the data processing model, wherein the denoising is performed according to a current denoising strategy of the model, wherein each image has a different gain;

[0014] A target projection image is outputted through the data processing model, wherein the target projection image is obtained according to the average weight of the N*N images after denoising.

[0015] The beneficial effects of the present invention include at least:

[0016] An embodiment of the present application provides a multi-gain flat-panel detector, which includes a scintillator layer and a reading circuit layer; before reading, the signal is detected and corrected by the reading circuit layer, and then an image is output. Each pixel of the read-out image corresponds to its own independent thin-film transistor array, control circuit and analog-to-digital converter, thereby realizing independent pixel reading; in addition, the overall capacitance of the charge amplifier can be adjusted by software controlling the switches of the circuits where the capacitors in the charge amplifier are located, thereby realizing control of the reading gain and solving the problem of data saturation and low data signal-to-noise ratio that are prone to occur in flat-panel detectors. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A schematic structural diagram of a multi-gain flat panel detector provided by an exemplary embodiment of the present application is shown;

[0018] Figure 2 A flowchart of a reading data processing method provided by an exemplary embodiment of the present application is shown;

[0019] Figure 3 The denoising training process of the data processing model is shown. DETAILED DESCRIPTION

[0020] To make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0021] In this document, "plurality" refers to two or more. "And / or" describes a relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can mean: A exists alone, A and B exist simultaneously, or B exists alone. The character " / " generally indicates an "or" relationship between the associated objects.

[0022] Please refer to Figure 1 , which shows a schematic structural diagram of a multi-gain flat panel detector provided by an exemplary embodiment of the present application.

[0023] The detector includes a scintillator layer 110 and a reading circuit layer 120 .

[0024] The reading circuit layer 120 includes a thin film transistor array 121 , a control circuit 122 and an analog-to-digital converter 123 , wherein the thin film transistor array 121 includes a photodiode.

[0025] The scintillator layer 110 is connected to the reading circuit layer 120. The scintillator layer 110 is used to convert X-rays into optical signals. The reading circuit layer 120 captures the optical signals through the photodiodes in the thin film transistor array 121 and converts the optical signals into electrical signals for storage before reading.

[0026] Furthermore, the thin film transistor array 121 is used to detect the light signal captured by the photodiode before reading; the control circuit 122 is used to sample, correct and multiplex the electrical signal stored before reading; and the analog-to-digital converter 123 is used to convert the voltage signal after sampling and multiplexing into a reading output signal.

[0027] On the basis of the above content, the following content is further disclosed.

[0028] According to the processing flow, in the control circuit, the control circuit includes a charge amplifier, a sample-and-hold circuit, a CDS correlated double sampling circuit and a multiplexing circuit.

[0029] Among them, the charge amplifier is used to convert the signal charge into a voltage signal through an integrating capacitor; the sample and hold circuit is used to sample and process the voltage signal; the CDS circuit is used to correct the voltage signal after sampling; and the multiplexing circuit is used to multiplex the voltage signal after correction.

[0030] During the readout phase, the accumulated signal charge of each pixel is sequentially transferred to the integrating capacitor of the charge amplifier channel, thereby converting the charge signal into a voltage signal. The voltage signal is processed by the sample-and-hold circuit, subjected to CDS correction, multiplexing, and then converted into a digital word stream.

[0031] In one possible embodiment, the charge amplifier is implemented by connecting capacitors of different capacitances in parallel, and a control switch is connected to the circuit where each capacitor is located, and the control switch is used to adjust the capacitance of the corresponding capacitor; capacitors of different capacitances are used to achieve different gain readings of the charge amplifier.

[0032] In a multi-gain flat panel detector, the IC contains the above-mentioned adjustable charge amplifier to meet the reading requirements of different gains. Different charge amplifiers can be realized by connecting multiple groups of capacitors with different capacitances in parallel. The circuit where each capacitor is located should include a switch to control whether the capacitor is turned on or off. Theoretically, N C Capacitors with different capacitance can be realized in the above case Readings at different gains.

[0033] In the multi-gain flat-panel detector provided by this solution, the gain used by each pixel during each reading can be controlled by software, that is, the control switch is manipulated by software. Specifically, the software can control the switches in the circuit where the capacitors in the charge amplifier are located to adjust the overall capacitance of the charge amplifier, thereby achieving control of the reading gain.

[0034] In the above embodiment, a description of the structure and function of a multi-gain flat-panel detector is provided. Before reading, the signal is detected and corrected by the reading circuit layer, and then the image is output. Each pixel of the read-out image should correspond to its own independent thin-film transistor array, control circuit and analog-to-digital converter, so as to achieve independent pixel reading; in addition, the overall capacitance of the charge amplifier can be adjusted by software controlling the switches of the circuits where the capacitors in the charge amplifier are located, so as to achieve control of the reading gain.

[0035] Please refer to Figure 2 , which shows a flow chart of a reading data processing method provided by an exemplary embodiment of the present application. The method includes:

[0036] Step 201 : Acquire an N*N data set according to a reading output signal, wherein the reading output signal is collected by the multi-gain flat panel detector of the above embodiment, where N is a positive integer greater than or equal to 2.

[0037] In one example, the pixel size of the flat panel detector is 139 microns. When used for CBCT imaging, an N*N pixel merging method is adopted. In the embodiment of the present application, N is 3 for schematic illustration, that is, 9 adjacent pixels are merged into one.

[0038] For example, for a 1024*1024 X-ray projection matrix (readout output signal), each pixel is synthesized using this 3*3 pixel binning. Using the multi-gain flat-panel detector described in the above embodiment for data acquisition, the actual dataset obtained is 1024*1024*9, i.e., a 3*3 projection matrix (i.e., a 3*3 dataset). Pixels 1-9 have different gain modes. Therefore, for the same object, the transmittance resolution and noise model of the nine pixels will be inconsistent under different gain modes.

[0039] Step 202: Input the data set into the data processing model to obtain N*N images to be processed by the data processing model.

[0040] Step 203 , denoising is performed on the N*N images through the data processing model, where the denoising is performed according to the current denoising strategy of the model, wherein each image has a different gain.

[0041] In a possible implementation, step 303 includes the following content.

[0042] Content 1: Use the data processing model to extract image features and learn noise patterns for N*N images to obtain the current denoising strategy.

[0043] Furthermore, content one includes the following: performing image feature extraction on N*N images to obtain a preset denoising strategy; performing denoising processing on the N*N images according to the preset denoising strategy to obtain a preset projection image, and the preset projection image is obtained based on the average weight of the N*N images after denoising; in response to the difference value between the preset projection image and the demonstration image being less than the target difference value, determining the preset denoising strategy as the current denoising strategy; in response to the difference value between the preset projection image and the demonstration image being not less than the target difference value, adjusting the preset denoising strategy through the data processing model.

[0044] Content 2: De-noise N*N images using the data processing model according to the current denoising strategy.

[0045] Step 204 : Outputting a target projection image through the data processing model, wherein the target projection image is obtained according to the average weight of the N*N images after denoising.

[0046] In a possible implementation, step 304 includes the following content.

[0047] Content 1. Obtain N*N images after denoising.

[0048] Content 2: Use the data processing model to superimpose N*N images and perform weighted averaging to obtain the target projection image and output it.

[0049] like Figure 3 As shown, it shows the denoising training process of the data processing model.

[0050] In one example, reference Figure 3 , 1024*1024*9 (i.e. 3*3 dataset) is used as the input of the data processing model, and 1024*1024 with high dynamic range and high signal-to-noise ratio (i.e. target projection image) is used as the output of the data processing model for model training.

[0051] The data processing model decomposes the 3x3 dataset into nine 1024x1024 images with varying signal-to-noise ratios. Each image has a nearly identical structure. The data processing model first extracts image features and learns noise patterns. Learning from a single image is relatively weak. Using the nine decomposed images as auxiliary learning enhances the ability to accurately learn noise. This allows the model to remove noise while accurately restoring image details and avoiding artifacts, resulting in the current denoising strategy.

[0052] Next, the images processed by the real-time denoising strategy are superimposed and weighted averaged to obtain a preset projection image. The preset projection image is compared with the high-quality noise-free image using a certain calculation method (such as MSE) to determine the effect of the image after the initial denoising of the data processing model, and the comparison result is fed back. In response to the difference between the preset projection image and the demonstration image being less than the target difference (for example, 0.0001), the preset denoising strategy is determined as the current denoising strategy, and the current denoising model is output and saved; in response to the difference between the preset projection image and the demonstration image being not less than the target difference (for example, 0.0001), the preset denoising strategy is adjusted through the data processing model, the model denoising training continues, and the denoised image is compared with the high-quality noise-free image. This cycle is repeated until the difference between the two is small enough, and the model is corrected and learned.

[0053] To ensure the robustness of the model, we trained it on a dataset of 100,000 images, encompassing a wide range of subjects and captured under a variety of shooting conditions. After model training, we tested it on a dataset that had not been used in the training to ensure that the model accurately outputs images with a high dynamic range and a high signal-to-noise ratio.

[0054] In summary, the present invention uses a multi-gain data acquisition strategy to obtain readings under different gain modes, and then processes the data through a model data processing method to obtain a projection image with a high dynamic range and a high signal-to-noise ratio.

[0055] Those skilled in the art will appreciate that in one or more of the above examples, the functions described herein can be implemented using hardware, software, firmware, or any combination thereof. When implemented using software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any medium that facilitates the transmission of computer programs from one place to another. The storage medium can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0056] The above description is merely an optional embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A multi-gain flat panel detector, characterized in that: The detector comprises: a scintillator layer and a reading circuit layer; The reading circuit layer includes a thin film transistor array, a control circuit and an analog-to-digital converter, wherein the thin film transistor array includes a photodiode; The scintillator layer is connected to the reading circuit layer, and the scintillator layer is used to convert X-rays into optical signals. The reading circuit layer captures the optical signals through the photodiodes in the thin film transistor array and converts the optical signals into electrical signals for storage before reading. The thin film transistor array is used to detect the light signal captured by the photodiode before reading; the control circuit is used to sample, correct and multiplex the electrical signal stored before reading; the analog-to-digital converter is used to convert the voltage signal after sampling and multiplexing into a reading output signal; Wherein, the control circuit includes a charge amplifier, a sample-and-hold circuit, a CDS correlated double sampling circuit and a multiplexing circuit; The charge amplifier is used to convert the signal charge into a voltage signal through an integrating capacitor. The charge amplifier is implemented by connecting capacitors of different capacitances in parallel. A control switch is connected to the circuit where each capacitor is located. The control switch is used to adjust the capacitance of the corresponding capacitor. The capacitors of different capacitances are used to achieve different gain readings of the charge amplifier. The sample and hold circuit is used to sample and process the voltage signal; The CDS circuit is used to perform correction processing on the voltage signal after sampling; The multiplexing circuit is used to multiplex the voltage signals after the correction process.

2. The multi-gain flat panel detector according to claim 1, characterized in that: The photodiode stores the signal charge via a photodiode capacitor.

3. The multi-gain flat panel detector according to claim 1, wherein: The control switch is controlled by software.

4. A reading data processing method, characterized in that: The method comprises: Acquire an N*N data set according to a reading output signal, wherein the reading output signal is collected by the multi-gain flat panel detector according to any one of claims 1 to 3, wherein N is a positive integer greater than or equal to 2; Inputting the data set into a data processing model to obtain N*N images to be processed by the data processing model; performing denoising on the N*N images using the data processing model, wherein the denoising is performed according to a current denoising strategy of the model, wherein each image has a different gain; A target projection image is outputted through the data processing model, wherein the target projection image is obtained according to the average weight of the N*N images after denoising.

5. The reading data processing method according to claim 4, characterized in that: The denoising process is performed on the N*N images by the data processing model, wherein the denoising process is performed according to the current denoising strategy of the model, including: Performing image feature extraction and noise law learning on the N*N images respectively through the data processing model to obtain a current denoising strategy; De-noising is performed on the N*N images using the data processing model according to the current de-noising strategy.

6. The reading data processing method according to claim 5, characterized in that: Outputting a target projection image through the data processing model, wherein the target projection image is obtained according to an average weight of the N*N images after denoising, includes: Get N*N images after denoising; The data processing model is used to superimpose the N*N images and then weight average them to obtain the target projection image and output it.

7. The reading data processing method according to claim 5, characterized in that: The data processing model is used to extract image features and learn noise patterns on the N*N images to obtain a current denoising strategy, including: Performing image feature extraction on the N*N images to obtain a preset denoising strategy; De-noising the N*N images according to the preset de-noising strategy to obtain a preset projection image, wherein the preset projection image is obtained according to the average weight of the N*N images after de-noising; In response to a difference value between the preset projection image and the demonstration image being smaller than a target difference value, determining the preset denoising strategy as a current denoising strategy; In response to a difference value between the preset projection image and the demonstration image being not less than the target difference value, the preset denoising strategy is adjusted through the data processing model.

Images