Method and apparatus for processing scanning data, and computer device and storage medium

Abstract

Disclosed is a method for processing scanning data. The method comprises: acquiring scanning data (S210); and performing joint processing on the scanning data by using a preset deep learning network, so as to obtain a processing result of the scanning data (S220), wherein the scanning data comprises scanning data of a plurality of different energy spectrums, and the deep learning network is configured to perform feature extraction on the scanning data of each energy spectrum among the scanning data of the plurality of different energy spectrums, and perform feature extraction on the scanning data of all energy spectrums among the scanning data of the plurality of different energy spectrums. Further disclosed are an apparatus for processing scanning data, and a computer device and a storage medium.

Description

Method, apparatus, computer device and storage medium for processing scan data

[0001] Related applications

[0002] This disclosure claims the benefit of priority to Chinese Patent Application No. 202311269305.8 filed on September 27, 2023 with the State Intellectual Property Office of China, the entire disclosure of which is incorporated herein by reference for all purposes. Technical Field

[0003] The present disclosure relates to the field of medical imaging technology, and in particular to a method, device, and storage medium for processing scan data. Background Art

[0004] For the scan data obtained by computed tomography (CT), as well as the pre-processed data obtained after pre-processing the scan data, a preset deep learning network can be used to perform noise reduction, region segmentation, and scan result analysis to obtain corresponding image processing results. For multi-spectral scan data, or other different types of scan data with consistent image structures, conventional technology usually uses a deep learning network to process each type of scan data separately to obtain a processing result for each type of scan data. In this way, the noise reduction degree, segmentation accuracy, and efficiency of the processing results are relatively low.

[0005] Currently, no effective solution has been proposed to the problem that conventional technologies use deep learning networks to process multiple types of multi-energy spectrum scanning data with consistent image structures separately, resulting in low accuracy and efficiency of the processing results.

[0006] Summary of the Invention

[0007] In this embodiment, a method, apparatus, and storage medium for processing scan data are provided to solve the problem in the related art that the accuracy and efficiency of processing results of processing multiple types of scan data separately are low.

[0008] A first aspect of the present disclosure provides a method for processing scan data, comprising: acquiring scan data; and jointly processing the scan data using a preset deep learning network to obtain a processing result of the scan data. The scan data includes scan data of multiple different energy spectra. The deep learning network is configured to perform feature extraction on the scan data of each energy spectrum in the scan data of the multiple different energy spectra, and to perform feature extraction on the scan data of all energy spectra in the scan data of the multiple different energy spectra.

[0009] In the above-mentioned first aspect, the deep learning network includes both a single extraction layer for performing feature extraction on the scanning data of each energy spectrum, and a common extraction layer for performing feature extraction on the scanning data of the various energy spectra.

[0010] In the above first aspect, the features obtained by extracting the features from the scanning data of each energy spectrum are only used by the scanning data of the energy spectrum; the features obtained by extracting the features from the scanning data of various energy spectra are shared by the scanning data of the multiple energy spectra.

[0011] In the first aspect above, the scanning data is obtained through a medical scanning device; the medical scanning device includes at least one of a photon counting computed tomography device, an energy spectrum computed tomography device, a direct digital photography device, a single photon emission computed tomography device, and a positron emission tomography device.

[0012] In the above-mentioned first aspect, the acquiring of scanning data includes: acquiring initial scanning data obtained by computed tomography based on photon counting; dividing the initial scanning data into a plurality of different energy spectrum segments to obtain scanning data of a plurality of different energy spectra corresponding to the plurality of energy spectrum segments.

[0013] In the first aspect above, the acquiring of scanning data includes: acquiring scanning data of a plurality of different energy spectra respectively corresponding to the plurality of energy spectra based on spectral computed tomography.

[0014] In the above-mentioned first aspect, the acquisition of scanning data of multiple different energy spectra corresponding to multiple energy spectra based on energy spectrum computed tomography includes: performing computed tomography scanning using at least two different energies to obtain scanning data corresponding to different energy spectra; or performing computed tomography scanning based on two different light sources to obtain scanning data corresponding to different energy spectra.

[0015] In the above-mentioned first aspect, before using a preset deep learning network to process the scan data to obtain the processing result of the scan data, the method further includes: using the scan data as input of the deep learning network respectively.

[0016] In the first aspect above, the scanning data is processed using a preset deep learning network to obtain the processing results of the scanning data, including: using a single extraction layer of the deep learning network to extract the data features of the scanning data of each energy spectrum respectively, and obtaining single data features corresponding to the scanning data of different energy spectra; using a common extraction layer of the deep learning network to jointly extract the data features of the scanning data of different energy spectra to obtain common data features; and obtaining the processing results of the scanning data of different energy spectra based on the single data features and the common data features.

[0017] In the first aspect above, the scanning data are jointly processed using a preset deep learning network to obtain the processing results of the scanning data, which also includes: based on the single data features, the scanning data of each energy spectrum are separately optimized using the deep learning network; based on the common data features, the scanning data of different energy spectra are jointly optimized using the deep learning network.

[0018] In the above-mentioned first aspect, the method also includes: if the scanning data is multi-energy spectrum data that has not been materially decomposed, performing material decomposition on the output result of the deep learning network, and constraining the material decomposition process based on a preset loss function to obtain a material graph corresponding to the multi-energy spectrum data; or if the scanning data is material decomposition data, converting the output result of the deep learning network into target energy spectrum data, and using the loss function structure in the deep learning network to calculate the loss function of the target energy spectrum data.

[0019] In the above-mentioned first aspect, the processing of the scanning data performed by the deep learning network includes at least one of the following processing: noise reduction processing of the scanning data of the multiple different energy spectra; extraction of regions of interest in the scanning data of the multiple different energy spectra; and analysis of scanning results of the scanning data of the multiple different energy spectra.

[0020] A second aspect of the present disclosure provides a device for processing scan data, the device comprising an acquisition module and an output module. The acquisition module is configured to acquire scan data. The output module is configured to process the scan data using a preset deep learning network to obtain a processing result of the scan data. The scan data includes scan data of multiple different energy spectra. The deep learning network is configured to perform feature extraction on the scan data of each energy spectrum in the scan data of the multiple different energy spectra, and to perform feature extraction on the scan data of various energy spectra in the scan data of the multiple different energy spectra.

[0021] In the above second aspect, the deep learning network includes both a single extraction layer for performing feature extraction on the scanning data of each energy spectrum, and a common extraction layer for performing feature extraction on the scanning data of the various energy spectra.

[0022] In the above first aspect, the features obtained by extracting the features from the scanning data of each energy spectrum are only used by the scanning data of the energy spectrum; the features obtained by extracting the features from the scanning data of various energy spectra are shared by the scanning data of the multiple energy spectra.

[0023] In the above-mentioned second aspect, the acquisition module is further configured to: acquire initial scanning data obtained by computed tomography based on photon counting; divide the initial scanning data into multiple different energy spectrum segments to obtain scanning data of multiple different energy spectra corresponding to the multiple energy spectrum segments.

[0024] In the above second aspect, the acquisition module is further configured to: acquire scanning data of multiple different energy spectra corresponding to the multiple energy spectra based on energy spectrum computed tomography.

[0025] In the above-mentioned second aspect, the acquisition module is further configured to: perform computed tomography scanning using at least two different energies to obtain scanning data corresponding to different energy spectra; or perform computed tomography scanning based on two different light sources to obtain scanning data corresponding to different energy spectra.

[0026] In the above second aspect, the acquisition module is further configured to: use the scanning data of the different energy spectra as inputs of the deep learning network respectively.

[0027] In the above-mentioned second aspect, the output module is further configured to: use the single extraction layer of the deep learning network to extract the data features of the scanning data of each energy spectrum respectively, and obtain the single data features corresponding to the scanning data of different energy spectra; use the common extraction layer of the deep learning network to jointly extract the data features of different types of scanning data to obtain common data features; and obtain the processing results of the scanning data of different energy spectra based on the single data features and the common data features.

[0028] In the above-mentioned second aspect, the output module is further configured to: based on the single data feature, use the deep learning network to perform separate optimization processing on the scanning data of each energy spectrum; based on the common data feature, use the deep learning network to perform joint optimization processing on the scanning data of different energy spectra.

[0029] In the above-mentioned second aspect, the output module is further configured as follows: if the scanning data is multi-energy spectrum data that has not been materially decomposed, the output result of the deep learning network is materially decomposed, and the material decomposition process is constrained based on a preset loss function to obtain a material map corresponding to the multi-energy spectrum data; or if the scanning data is material decomposition data, the output result of the deep learning network is converted into target energy spectrum data, and the loss function structure in the deep learning network is used to calculate the loss function of the target energy spectrum data.

[0030] In the above-mentioned second aspect, the output module is further configured to perform at least one of the following processes: noise reduction processing on the scanning data of the multiple different energy spectra; extraction of regions of interest in the scanning data of the multiple different energy spectra; and analysis of scanning results of the scanning data of the multiple different energy spectra.

[0031] A third aspect of the present disclosure provides a computer device comprising a processor and a memory. The memory stores a computer program executable by the processor. When the processor executes the computer program, the method for processing scan data according to the first aspect is implemented.

[0032] A third aspect of the present disclosure provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the processor is prompted to implement the method for processing scan data described in the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The above and other aspects, features and advantages of the present disclosure will become apparent and readily understood from the description of the embodiments with reference to the following drawings, in which:

[0034] FIG1 is a structural block diagram of a terminal that executes a method for processing scan data according to an embodiment of the present disclosure;

[0035] FIG2 is a flowchart of a method for processing scan data according to an embodiment of the present disclosure;

[0036] FIG3 is a flowchart of a method for processing scan data according to another embodiment of the present disclosure;

[0037] FIG4 is a flow chart of a method for denoising multi-energy spectrum scanning data according to an embodiment of the present disclosure;

[0038] FIG5 is a flow chart of a method for performing image segmentation on multi-energy spectrum scanning data according to an embodiment of the present disclosure;

[0039] FIG6 is a structural block diagram of an apparatus for processing scan data according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0040] To make the objectives, features, and advantages of the present disclosure more apparent and understandable, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of them. All other embodiments obtained by persons of ordinary skill in the art based on the embodiments of the present disclosure without creative effort shall fall within the scope of protection of the present disclosure.

[0041] In this document, reference to the terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this document, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0042] Although the terms "first" and "second" may be used herein to describe various features or elements, these features or elements should not be limited by these terms unless otherwise specifically indicated. These terms can be used to distinguish one feature or element from another feature or element. Therefore, the first feature or element described below can be referred to as the second feature or element, and similarly, the second feature or element described below can be referred to as the first feature or element, without departing from the scope of this disclosure. In the description of the present disclosure, "plurality" refers to at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0043] The terms used in this article are only used to describe the purpose of specific embodiments and are not intended to limit the present disclosure. For example, as used in this article, the singular form "a" is also intended to include the plural form, unless the context clearly indicates otherwise. It will be further understood that when used in this article, the term "including" specifies the presence of the features, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, steps, operations, elements, components and / or groups thereof. As used in this article, the term "and / or" includes any and all combinations of one or more associated listed items and can be abbreviated as " / ".

[0044] It should be understood that when a feature or element is referred to as being "connected," "attached," or "coupled" to another feature or element, it can be directly connected, attached, or coupled to the other feature or element, or there can be intervening features or elements. In contrast, when a feature or element is referred to as being "directly connected," "directly attached," or "directly coupled" to another feature or element, there are no intervening features or elements. Although described or illustrated with respect to one embodiment, the features and elements so described or illustrated may apply to other embodiments.

[0045] One aspect of the present disclosure provides a method for processing scanned data. This method can be executed on a computer device such as a terminal or a server. Terminals include fixed terminals such as desktop computers and mobile terminals such as laptops, tablets, and mobile phones. The following describes the details of an embodiment of the present disclosure, using the execution of the method for processing scanned data provided by the present disclosure on a terminal as an example.

[0046] FIG1 is a block diagram of a terminal for executing a method for processing scanned data according to an embodiment of the present disclosure. As shown in FIG1 , the terminal may include one or more (only one is shown in FIG1 ) processors 102 and a memory 104 for storing data. The processor 102 may include, but is not limited to, a central processing unit (CPU), a microcontroller unit (MCU), a digital signal processor (DSP), an embedded processor, and other processing devices. The processor may be implemented by a dedicated integrated circuit, a programmable logic device, a system on a chip, and the like. The terminal may also include a transmission device 106 and an input / output device 108 for communication functions. It will be understood by those skilled in the art that the structure shown in FIG1 is merely illustrative and does not limit the structure of the above-mentioned terminal. For example, the terminal may also include more or fewer components than those shown in FIG1 , or have a different configuration than that shown in FIG1 .

[0047] The memory 104 can be used to store computer programs, such as software programs and modules of application software, such as the computer program corresponding to the method for processing scanned data in this embodiment. The processor 102 executes various functional applications and data processing by running the computer programs stored in the memory 104. The memory 104 may include high-speed random access memory and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories may be connected to the terminal via a wired or wireless network. Examples of wireless networks include, but are not limited to, the Internet, the Internet of Things, mobile communication networks, and combinations thereof.

[0048] The transmission device 106 is used to receive data from or send data to the outside via network communication. Network communication includes wireless network communication provided by the terminal's communication provider. In one embodiment, the transmission device 106 includes a network interface controller (NIC), which can be connected to other network devices via a base station to transmit information via the Internet. In another embodiment, the transmission device 106 can be a radio frequency (RF) module, which is used to access the Internet wirelessly.

[0049] In one embodiment of the present disclosure, a method for processing scan data is provided. As shown in FIG. 2 , the method includes step S210 and step S220 .

[0050] In step S210, scan data is acquired. The scan data is obtained by a medical scanning device. The medical scanning device includes at least one of a photon counting computed tomography (PCD-CT) device, a spectral computed tomography (spectral CT) device, a direct digital radiography (DR) device, a single photon emission computed tomography (SPECT) device, and a positron emission tomography (PET-CT) device.

[0051] In some embodiments, the scan data includes multiple types of scan data that have consistent image structures but differ in other specific aspects, i.e., different types of scan data with consistent image structures. These different types of scan data may include different types of scan data obtained by scanning the same scan object (e.g., a body part in the same physical location) using different scanning sequences. These scan data can represent image structures corresponding to the structure of the scan object. Because the scans are performed on the same scan object, the image structures represented by these scan data are consistent. These different types of scan data may also include pre-processed data obtained by pre-processing the acquired scan data, such as material decomposition data obtained by performing material decomposition on the scan data. From the perspective of the images themselves, scan data with consistent image structures can be at least two types of images with consistent corresponding structural composition. From the perspective of scanning and reconstruction, scan data with consistent image structures include at least two types of scan data with consistent scanning ranges, reconstruction ranges, and reconstruction parameters. For example, for different types of scan data, if the pixel positions corresponding to the heart on the image are consistent and the pixel positions corresponding to artifacts are consistent, then these scan data are considered to have consistent image structures.

[0052] The different types of scan data described above may be scan data of multiple energy spectra obtained by scanning the scanned object during a CT scan using X-rays distributed using at least two different energy spectra (e.g., a high-energy spectrum and a low-energy spectrum), hereinafter referred to as multi-energy spectrum scan data. For example, the scan data may be multi-energy spectrum scan data comprising multiple energy spectral segments obtained through photon counting CT scanning, multi-energy spectrum scan data obtained through separate CT scans using different light sources, or multi-energy spectrum scan data obtained through separate CT scans using different energies. The different types of scan data described above may also be scan data of different modalities obtained by performing different scan modalities on the same scanned object. Scans using different modalities include CT scans, positron emission tomography (PET) scans, and the like. For example, a PET-CT device may be used to scan the scanned object, such that the different types of scan data obtained include PET scan data and CT scan data obtained by performing PET and CT scans on the same scanned object. The different types of scan data described above may also be images of different base materials obtained by performing different material decomposition processes on the same CT scan data. In addition, the above-mentioned different types of scan data can also be pre-processed data obtained after the scan data undergoes other pre-processing. The scan data can be either forward projection raw data after scanning or filtered back-projection images.

[0053] In some embodiments, the scanning data includes scanning data of multiple different energy spectra, wherein each energy spectrum scanning data corresponds to an energy spectrum segment, such as 30-60 kV, 60-140 kV, etc.

[0054] In step S220, the scan data is processed using a preset deep learning network to obtain a processing result of the scan data. The deep learning network is configured to perform feature extraction on the scan data of each energy spectrum among the multiple different energy spectrum scan data, that is, to perform feature extraction on each single type of scan data among the different types of scan data (hereinafter referred to as "individual feature extraction"), and to perform feature extraction on the scan data of various energy spectra among the multiple different energy spectrum scan data, that is, to perform feature extraction on all types of scan data among the different types of scan data (hereinafter referred to as "common feature extraction"). The categories of the features extracted by "individual feature extraction" and the features extracted by "common feature extraction" can be the same or different, and the categories of these features include local texture, edge and color information, etc., as well as global shape, structure and semantic information, etc. "Individual feature extraction" is to perform feature extraction on a single type of scan data, and the extracted features are only used by the scan data of that type. For example, the processing result corresponding to the energy spectrum scan data is obtained using the individually extracted features. "Common feature extraction" involves extracting features shared by multiple categories of scan data. For example, the common features are used to generate processing results corresponding to scan data from multiple energy spectra. In some embodiments, the processing results corresponding to the multiple energy spectra scan data include processing results corresponding to each of the multiple energy spectra scan data.

[0055] In some embodiments, the deep learning network includes both a single extraction layer for extracting individual features from a single type of scan data, and a common extraction layer for extracting common features from multiple types of scan data. For example, for two types of energy spectrum scan data, such as first energy spectrum scan data and second energy spectrum scan data, the deep learning network may have two extraction layers for extracting features from the two types of energy spectrum scan data, wherein the first extraction layer extracts features that can be shared by the first energy spectrum scan data and the second energy spectrum scan data from the first energy spectrum scan data and the second energy spectrum scan data, and the second extraction layer extracts features that are only used by the first energy spectrum scan data from the first energy spectrum scan data, or extracts features that are only used by the second energy spectrum scan data from the second energy spectrum scan data. In this case, the first extraction layer can be referred to as a common extraction layer, and the second extraction layer can be referred to as a single extraction layer.

[0056] The above-mentioned deep learning network can be any neural network used for medical image processing, such as the medical image segmentation model Unet++, the dense convolutional network (Dense Convolutional Network, referred to as DenseNet) and the residual network ResNet, etc., which are not specifically limited in this embodiment. A single type of scan data refers to one type of scan data among the above-mentioned different types of scan data, for example, one type of scan data corresponding to one energy spectrum among multiple types of scan data corresponding to multiple energy spectra. Multiple types of scan data refer to some or all types of scan data in this group of different types of scan data with consistent image structure.

[0057] The deep learning network includes a single extraction layer that extracts features from each type of scan data separately, and a common extraction layer that extracts features from all different types of scan data. When extracting features from each type of scan data separately, the deep learning network first distinguishes the scan data by type and then extracts features from each type of scan data separately. When extracting features from all different types of scan data together, the network extracts features from the mixed scan data without distinguishing the different types of scan data.

[0058] Additionally, the joint processing of different types of scan data using the preset deep learning network can include noise reduction processing on the input different types of scan data to optimize the image quality of the scan data; for example, segmenting or identifying regions of interest on the input different types of scan data to output corresponding segmented images or recognition results; or performing scan result analysis on the input scan data, etc. Scan result analysis includes structural analysis of the scanned object image to identify lesions, energy spectrum analysis of the scanned object image to identify material components, etc. Based on this, the structure of the deep learning network will vary depending on the application scenario, and the structure and specific implementation of the individual extraction layer and the common extraction layer will also vary. In multi-spectral scan data, by jointly performing noise reduction, image segmentation, and scan result analysis on the multi-spectral scan data, the accuracy of the processing results can be improved. For example, joint noise reduction of multi-spectral scan data not only achieves simultaneous noise reduction of scan data of different energy spectra, improving noise reduction efficiency, but also improves noise reduction performance and optimizes noise reduction results by extracting common features between the data features of the scan data of different energy spectra and the individual features of the scan data of each energy spectra. Therefore, for multi-energy spectrum scanning data, the above method can improve the accuracy of processing results, enhance processing performance and improve processing efficiency.

[0059] In the above steps S210 to S220, scan data is acquired, and the scan data is jointly processed using a preset deep learning network to obtain processing results of different types of scan data. The deep learning network used also includes a single extraction layer for performing individual feature extraction on a single type of scan data, and a common extraction layer for performing common feature extraction on multiple types of scan data. Thus, by jointly inputting different types of scan data that are consistent in image structure, such as multi-energy spectrum scan data, different types of scan data that are consistent in image structure, such as multi-energy spectrum scan data, can be processed simultaneously based on the deep learning network, thereby improving the accuracy of the output results of the deep learning network and also improving the efficiency of scan data analysis.

[0060] In one embodiment, the step S210 of acquiring scan data may include: acquiring initial scan data obtained by photon counting computed tomography; dividing the initial scan data into multiple energy spectrum segments (e.g., at least two energy spectrum segments) to obtain a plurality of scan data corresponding to the multiple energy spectrum segments. The energy spectrum segment division method can be selected based on the image quality of each energy spectrum segment after image reconstruction and / or the number of photons obtained in each energy spectrum segment. Specifically, after a photon counting CT scan, the obtained energy spectrum data is divided into multiple energy spectrum segments, e.g., at least two energy spectrum segments. For example, if a total energy spectrum of 140 kV is divided into two energy spectrum segments, the first energy spectrum segment can be 30-60 kV, and the second energy spectrum segment can be 60-140 kV, or the first energy spectrum segment can be 20-65 kV, and the second energy spectrum segment can be 65-140 kV. Each energy spectrum segment corresponds to scan data of a different energy spectrum, thereby obtaining a plurality of scan data corresponding to the multiple energy spectrum segments. Therefore, in this embodiment, a variety of scanning data corresponding to multiple energy spectrum bands can be obtained based on a single photon counting CT.

[0061] In another embodiment, the step S210 of acquiring scan data may include acquiring multiple scan data corresponding to multiple energy spectra (e.g., at least two energy spectra) based on spectral computed tomography. For example, CT scans may be performed using at least two different energies to obtain different types of scan data corresponding to different energy spectra; or CT scans may be performed using two different light sources to obtain different types of scan data corresponding to different energy spectra.

[0062] In one embodiment, before using a preset deep learning network to jointly process different types of scan data to obtain processing results of different types of scan data, the above-mentioned method for processing scan data may also include: directly inputting different types of scan data into the model of the deep learning network, or respectively inputting them into different channels of the model. In some embodiments, different types of scan data can be combined in the image channel dimension or the image space dimension, and the combined scan data can be used as input to the deep learning network. Combining in the image channel dimension means combining different types of scan data together according to the image channel dimension, while combining in the image space dimension means combining different types of scan data together according to the image space dimension. For example, scan data of the same energy spectrum can be input into the same channel of the model.

[0063] In one embodiment, before jointly processing different types of scan data using a preset deep learning network to obtain processing results for the different types of scan data, the method for processing scan data may further include: using the different types of scan data as inputs to the deep learning network. That is, the different types of scan data may be assigned to different inputs of the deep learning network, and the different types of scan data may be simultaneously input into the deep learning network, allowing the deep learning network to process the different scan data.

[0064] In one embodiment, the above-mentioned step S220 of jointly processing different types of scanning data using a preset deep learning network to obtain processing results of different types of scanning data may include: using a single extraction layer of the deep learning network to extract data features of each type of scanning data separately, and obtain single data features corresponding to each type of scanning data; using a common extraction layer of the deep learning network to jointly extract data features of different types of scanning data, and obtain common data features; and obtaining processing results of different types of scanning data based on single data features and common data features.

[0065] Both the single extraction layer and the common extraction layer are feature extraction layers in a deep learning network. The specific implementation of the structure of the feature extraction layer varies for different deep learning networks, and this embodiment does not specifically limit this. The data features extracted by the feature extraction layer refer to certain image features extracted by the convolution kernel in the neural network structure, such as image edges / small blood vessels and other image information. In the single extraction layer, it is necessary to distinguish the data sources of different types of scan data in order to extract the data features of each type of scan data separately. For example, for multi-spectral scan data, the data features of each type of scan data are extracted separately. The single extraction layer is mainly used to extract features from a single type of scan data to perform network training on the single type of scan data. Single data features refer to data features extracted from a single type of scan data by the single extraction layer. In the common extraction layer, the scan data is not distinguished, and data features of all scan data are extracted together. The common extraction layer is mainly used to extract common features of multiple types of scan data (for example, the multiple types of scan data included in a multi-spectral image) from these types of scan data to perform network training on these types of scan data. The common data features refer to data features common to multiple types of scan data extracted from these types of scan data by the common extraction layer.

[0066] The deep learning network can perform separate optimization processing on each type of scan data based on the above-mentioned single data features. For example, in the case where there are a first type of scan data and a second type of scan data, the first type of scan data can be optimized based on the first single data feature extracted from the first type of scan data, and the second type of scan data can be optimized based on the second single data feature extracted from the second type of scan data. In addition, the deep learning network can perform common optimization processing on different types of scan data based on the above-mentioned common data features, so that the optimization results of different types of scan data interact and influence each other. For example, in the case where there are a first type of scan data and a second type of scan data, the first type of scan data and the second type of scan data can be optimized separately based on the common data features extracted from the first type of scan data and the second type of scan data (including the first common data feature extracted from the first type of scan data and the second common data feature extracted from the second type of scan data), that is, the first type of scan data is optimized based on the first common data feature and the second common data feature, and the second type of scan data is optimized based on the first common data feature and the second common data feature. The above-mentioned individual optimization processing and joint optimization processing include but are not limited to the following optimization processing: optimizing the image quality of the scanned data (including noise reduction, edge smoothing, etc.); optimizing the segmentation of the region of interest in the scanned data; optimizing the identification of the region of interest in the scanned data, etc.

[0067] Based on this, in this embodiment, since the deep neural network includes a single extraction layer and a common extraction layer, it can process different types of scanning data in parallel while avoiding excessive fluctuations in processing results between different types of scanning data, which may affect the consistency of the structure of the scanning object reflected by different types of scanning data.

[0068] The following example illustrates noise reduction processing for multi-spectrum CT scan data. After obtaining CT scan data from two energy spectra, it is input into a deep learning network. The deep learning network extracts single data features from each of the two energy spectra and performs noise reduction on each CT scan data based on the single data features. It also extracts common data features from the two energy spectra and performs common noise reduction on these common data features. This allows the deep learning network to simultaneously reduce noise for both energy spectra and ensure structural consistency between the two energy spectra, preventing excessive fluctuations in CT values.

[0069] In particular, in one embodiment, the above-mentioned method for processing scanning data may further include: if the above-mentioned different types of scanning data are multi-energy spectrum data of at least two energy spectra that have not been materially decomposed, performing material decomposition on the output results of the deep learning network, and constraining the material decomposition process based on a preset loss function to obtain a material map corresponding to the multi-energy spectrum data; if the above-mentioned different types of scanning data are material decomposition data, converting the output results of the deep learning network into target energy spectrum data, and using the loss function structure in the deep learning network to calculate the loss function of the target energy spectrum data.

[0070] The above-mentioned material decomposition refers to the process of decomposing the CT scan data of the mixture into element maps according to the chemical composition, for example, decomposing it into a water map, an iodine map, etc. The specific type of material decomposition depends on the actual application requirements. The multi-energy spectrum data of the above-mentioned at least two energy spectra are at least two different types of CT scan data corresponding to at least two energy spectra respectively. In this embodiment, when the scan data is CT scan data, the output result after processing by the deep learning network is subjected to material decomposition, for example, after the multi-energy spectrum data is subjected to noise reduction processing using the deep learning network, the noise-reduced CT scan data is subjected to material decomposition. Based on the above-mentioned material decomposition, a more accurate and clear expression can be provided for the scan results, so as to facilitate subsequent analysis and judgment based on the scan results.

[0071] When using a deep learning network to process scan data, it is often necessary to verify the accuracy of the deep learning network's output against a gold standard (ground truth). For two types of energy spectrum scan data, for example, first energy spectrum scan data and second energy spectrum scan data, a gold standard can be set for each of the two types of energy spectrum scan data. The model based on the gold standard will output the results corresponding to the first energy spectrum scan data and the second energy spectrum scan data, respectively. For another example, a gold standard can be set only for the first energy spectrum scan data, in which case the model will only output the results corresponding to the first energy spectrum scan data; or a gold standard can be set only for the second energy spectrum scan data, in which case the model will only output the results corresponding to the second energy spectrum scan data. The gold standard can be mathematically represented as a matrix with the same size as the scan result. Therefore, when performing material decomposition on the output of the deep learning network, it is also necessary to perform material decomposition on the gold standard as well. To smooth the edges of the image after material decomposition and improve the quality of the material map, in this embodiment, a loss function is used to constrain the material decomposition process. Specifically, the loss function is the loss value between the decomposition result obtained by multiplying the output of the deep learning network by the material decomposition matrix and the decomposition result obtained by multiplying the gold standard by the material decomposition matrix. The loss function is as follows: Loss=L(M out,1 -M gold,1 ,M out,2 -M gold,2 ,...,M out,m -M gold,m ) (2)

[0072] Among them, M out,i M represents the decomposition result of the ith substance after the output of the deep learning network is decomposed. gold,i represents the decomposition result of the gold standard corresponding to the i-th substance after material decomposition, μ i,j Represents the decomposition coefficient matrix of base material i under energy spectrum j, Out j is the j-th energy spectrum data output by the deep learning network, Gold j is the energy spectrum j data of the corresponding gold standard. The above loss function can use the material decomposition results of only one material basis or multiple material basis groups. This material decomposition result forms a material graph, which can be used for material identification and medical diagnosis. Based on this, this embodiment constrains the material decomposition process based on the loss function, resulting in smoother edges on the processed material graph, improving its quality and facilitating subsequent manipulation of the graph.

[0073] As described above, if the different types of scanning data input into the deep learning network are multi-energy spectrum data of at least two energy spectra without material decomposition, a loss function structure of the material decomposition data can be added to the deep learning network. After the output result of the deep learning network (i.e., the target energy spectrum data) is converted into material decomposition data, the loss function structure of the material decomposition data is used to calculate the loss function of the material decomposition data. By adding a loss function structure of a data type (e.g., material decomposition data) different from the input data type (e.g., energy spectrum data) to the deep learning network, the loss functions of different data types can be used to further optimize the model and parameters used in the deep learning network, thereby making the edges of the processed material map smoother and improving the quality of the material map.

[0074] Furthermore, if the different types of scan data input into the deep learning network are material decomposition data that has been decomposed, i.e., material maps, a loss function structure for the target energy spectrum data can be incorporated into the deep learning network. After converting the output of the deep learning network (i.e., the target material map) into target energy spectrum data, the loss function structure for the target energy spectrum data can be used to calculate the loss function for the target energy spectrum data. This can also smooth the edges of the processed material map, improving its quality.

[0075] In one embodiment, the deep learning network's joint processing of different types of scan data includes at least one of the following: noise reduction processing of the different types of scan data; extraction of regions of interest from the different types of scan data; and analysis of scan results from the different types of scan data. For example, the input scan data to the deep neural network may be a noisy image, and the output may be a noise-reduced image. The input to the deep neural network may be a scan image containing a region of interest, and the output may be a segmented region of interest, or a scan result analysis of the medical analysis performed on the scan image.

[0076] The method for processing scan data provided by the present disclosure is further described and illustrated below through optional implementations.

[0077] FIG3 is a flow chart of a method for processing scan data according to an embodiment of the present disclosure. As shown in FIG3 , the method for processing scan data includes steps S301 to S304 .

[0078] In step S301, multi-spectral data of at least two different energy spectra is acquired. The energy spectrum data acquired by photon counting CT can be divided into multiple energy spectrum segments to obtain multi-spectral data. Alternatively, multi-spectral data can be obtained by performing spectral CT scans using different light sources or different energies. This multi-spectral data can be either forward-projected raw data or filtered back-projected images.

[0079] In step S302, the multi-energy spectrum data is input into a preset deep neural network for processing.

[0080] In step S303, the deep neural network distinguishes the input multi-spectral data by different energy spectra, extracts data features of each energy spectrum data separately, and extracts data features of all energy spectrum data together. The multi-spectral data is then processed based on the extracted data features. Specific processing may include noise reduction, region of interest segmentation, or scan result analysis.

[0081] In step S304, the output result of the deep neural network is subjected to material decomposition, and the process of material decomposition is constrained by using a loss function to obtain a material graph.

[0082] In steps S301 through S304, combined noise reduction, image segmentation, or scan result analysis of the multi-spectral data can improve both processing efficiency and performance. Furthermore, constraining the material decomposition process based on a loss function to generate a material map can smooth the edges of the processed material map, thereby improving its quality.

[0083] In another embodiment, a method for denoising multi-spectral scanning data is provided. As shown in FIG4 , the denoising method may specifically include step S401 and step S402 .

[0084] In step S401, a set of multi-energy spectrum scan data is acquired. The multi-energy spectrum scan data may be at least two types of multi-energy spectrum scan data obtained by scanning a scanned object using at least two different energy spectra during a CT scan. For example, multi-energy spectrum scan data including multiple energy spectrum segments obtained by photon counting CT scanning, multi-energy spectrum scan data obtained by performing CT scans using different light sources, or multi-energy spectrum scan data obtained by performing CT scans using different energies.

[0085] In step S402, a preset deep learning network is used to perform joint noise reduction on the multi-energy spectrum scan data to obtain a noise reduction result for the multi-energy spectrum scan data. The deep learning network also includes a single extraction layer for performing individual feature extraction on the single-energy spectrum scan data, and a common extraction layer for performing common feature extraction on the multi-energy spectrum scan data. The deep learning network can be any neural network used for medical image noise reduction, such as a convolutional neural network (CNN), a deep convolutional neural network (DCNN), etc., and this embodiment does not specifically limit this.

[0086] In the above steps S401 to S402, noise reduction processing can be simultaneously performed on the scanning data of multiple energy spectra, while improving the noise reduction efficiency. Based on the common characteristics of the scanning data of multiple energy spectra, the noise reduction performance is further optimized, thereby improving the noise reduction effect.

[0087] In another embodiment, a method for performing image segmentation on multi-energy spectrum scanning data is provided. As shown in FIG5 , the image segmentation method may specifically include step S501 and step S502.

[0088] In step S501, a set of multi-energy spectrum scanning data is acquired; the multi-energy spectrum scanning data may be different types of multi-energy spectrum scanning data obtained by scanning the scan object based on at least two different energy spectra during a CT scanning process.

[0089] In step S502, a preset deep learning network is used to perform joint region of interest extraction on the scanning data of the multi-energy spectrum to obtain the region of interest of the scanning data of the multi-energy spectrum. The deep learning network also includes a single extraction layer for performing individual feature extraction on the scanning data of a single energy spectrum, and a common extraction layer for performing common feature extraction on the scanning data of multiple energy spectra. The deep learning network also includes a single extraction layer for performing individual feature extraction on the scanning data of a single energy spectrum, and a common extraction layer for performing common feature extraction on the scanning data of multiple energy spectra. The deep learning network can be any neural network used for medical image segmentation, such as medical image segmentation models Unet, Unet++, Vnet, etc., which is not specifically limited in this embodiment.

[0090] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0091] Another aspect of the present disclosure also provides a device for processing scanned data. The implementation solution provided by the device to solve the problem is similar to the implementation solution described in the above method, so the specific definitions in one or more device embodiments provided below can refer to the above definitions of the method for processing scanned data, and will not be repeated here. The terms "module", "unit", "sub-unit", etc. used below can be a combination of software and / or hardware that can implement predetermined functions. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware, is also possible and conceived.

[0092] FIG6 is a block diagram of a device 60 for processing scan data according to an embodiment of the present disclosure. As shown in FIG6 , the device 60 for processing scan data includes an acquisition module 62 and an output module 64. The acquisition module 62 is used to acquire scan data. The output module 64 is used to process the scan data using a preset deep learning network to obtain a processing result of the scan data. The scan data includes scan data of multiple different energy spectra. The deep learning network is configured to perform feature extraction on the scan data of each energy spectrum among the multiple different energy spectrum scan data, and to perform feature extraction on the scan data of various energy spectra among the multiple different energy spectrum scan data. Specifically, the deep learning network includes both a single extraction layer for extracting features from the scan data of a single energy spectrum and a common extraction layer for extracting features from the scan data of multiple energy spectra. The features extracted from the scan data of a single energy spectrum by the single extraction layer are used only by the scan data of the single energy spectrum; the features extracted from the scan data of multiple energy spectra by the common extraction layer are shared by the scan data of the multiple energy spectra.

[0093] In some embodiments, the acquisition module 62 is further configured to acquire initial scanning data obtained by photon counting-based computed tomography; divide the initial scanning data into multiple different energy spectrum segments to obtain scanning data of multiple different energy spectra corresponding to the multiple energy spectrum segments.

[0094] In some embodiments, the acquisition module 62 is further configured to acquire scanning data of a plurality of different energy spectra respectively corresponding to a plurality of energy spectra based on spectral computed tomography.

[0095] In some embodiments, the acquisition module 62 is further configured to: perform computed tomography scanning using at least two different energies to obtain scanning data corresponding to different energy spectra; or perform computed tomography scanning based on two different light sources to obtain scanning data corresponding to different energy spectra.

[0096] In some embodiments, the acquisition module 62 is further configured to use the scanning data of the different energy spectra as inputs of the deep learning network respectively.

[0097] In some embodiments, the output module 64 is further configured to: use a single extraction layer of the deep learning network to extract data features of the scanning data of each energy spectrum respectively, and obtain single data features corresponding to the scanning data of different energy spectra; use a common extraction layer of the deep learning network to jointly extract data features of different types of scanning data, and obtain common data features; and obtain processing results of the scanning data of different energy spectra based on the single data features and the common data features.

[0098] In some embodiments, the output module 64 is further configured to: based on the single data feature, use the deep learning network to perform separate optimization processing on the scanning data of each energy spectrum; based on the common data feature, use the deep learning network to perform joint optimization processing on the scanning data of different energy spectra.

[0099] In some embodiments, the output module 64 is further configured as follows: if the scanning data is multi-energy spectrum data that has not been materially decomposed, the output result of the deep learning network is materially decomposed, and the material decomposition process is constrained based on a preset loss function to obtain a material map corresponding to the multi-energy spectrum data; or if the scanning data is material decomposition data, the output result of the deep learning network is converted into target energy spectrum data, and the loss function structure in the deep learning network is used to calculate the loss function of the target energy spectrum data.

[0100] In some embodiments, the output module 64 is also configured to perform at least one of the following processing: noise reduction processing of the scanning data of the multiple different energy spectra; extraction of regions of interest in the scanning data of the multiple different energy spectra; and scanning result analysis of the scanning data of the multiple different energy spectra.

[0101] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0102] It should be noted that, for specific examples in this embodiment, reference may be made to the examples described in the above embodiments and optional implementation modes, and will not be repeated in this embodiment.

[0103] Another aspect of the present disclosure provides a computer device, which may be a terminal as shown in FIG1 . The terminal's memory 104 stores a computer program for processing scanned data. When the processor executes the computer program, the method for processing scanned data according to any one of the above embodiments is implemented.

[0104] In one example embodiment, when a processor executes a computer program, the program is configured to: acquire scan data; and jointly process the scan data using a preset deep learning network to obtain a processing result for the scan data. The scan data includes scan data of multiple different energy spectra. The deep learning network is configured to perform feature extraction on the scan data of each energy spectrum in the multiple different energy spectra, and to perform feature extraction on the scan data of all energy spectra in the multiple different energy spectra.

[0105] In some embodiments, the deep learning network includes both a single extraction layer for performing feature extraction on the scanning data of each energy spectrum and a common extraction layer for performing feature extraction on the scanning data of various energy spectra.

[0106] In some embodiments, the features obtained by extracting features from the scanning data of each energy spectrum are only used by the scanning data of the energy spectrum; the features obtained by extracting features from the scanning data of various energy spectra are shared by the scanning data of the multiple energy spectra.

[0107] In some embodiments, the acquiring of scanning data includes: acquiring initial scanning data obtained by photon counting-based computed tomography; dividing the initial scanning data into a plurality of different energy spectrum segments to obtain scanning data of a plurality of different energy spectra corresponding to the plurality of energy spectrum segments.

[0108] In some embodiments, acquiring the scanning data includes: acquiring scanning data of a plurality of different energy spectra corresponding to the plurality of energy spectra based on spectral computed tomography.

[0109] In some embodiments, the acquisition of scanning data of multiple different energy spectra corresponding to multiple energy spectra based on energy spectrum computed tomography includes: performing computed tomography using at least two different energies to obtain scanning data corresponding to different energy spectra; or performing computed tomography based on two different light sources to obtain scanning data corresponding to different energy spectra.

[0110] In some embodiments, before using a preset deep learning network to process the scan data to obtain a processing result of the scan data, the method further includes: using the scan data as input to the deep learning network respectively.

[0111] In some embodiments, the scanning data is processed using a preset deep learning network to obtain the processing results of the scanning data, including: using a single extraction layer of the deep learning network to extract data features of the scanning data of each energy spectrum respectively, and obtaining single data features corresponding to the scanning data of different energy spectra; using a common extraction layer of the deep learning network to jointly extract data features of the scanning data of different energy spectra to obtain common data features; and obtaining the processing results of the scanning data of different energy spectra based on the single data features and the common data features.

[0112] In some embodiments, the scanning data is jointly processed using a preset deep learning network to obtain the processing results of the scanning data, and further includes: based on the single data features, using the deep learning network to perform separate optimization processing on the scanning data of each energy spectrum; based on the common data features, using the deep learning network to perform joint optimization processing on the scanning data of different energy spectra.

[0113] In some embodiments, when the processor executes the computer program, it also implements: if the scanning data is multi-energy spectrum data that has not been materially decomposed, the output result of the deep learning network is materially decomposed, and the material decomposition process is constrained based on a preset loss function to obtain a material map corresponding to the multi-energy spectrum data; or if the scanning data is material decomposition data, the output result of the deep learning network is converted into target energy spectrum data, and the loss function structure in the deep learning network is used to calculate the loss function of the target energy spectrum data.

[0114] In some embodiments, when the processor executes the computer program, it also implements at least one of the following processes: noise reduction processing of the scanning data of the multiple different energy spectra; extraction of regions of interest in the scanning data of the multiple different energy spectra; and analysis of scanning results of the scanning data of the multiple different energy spectra.

[0115] Another aspect of the present disclosure provides a computer-readable storage medium having a computer program stored thereon. When executed by a processor, the computer program prompts the processor to implement the method for processing scan data according to any one of the above embodiments.

[0116] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided by the present disclosure may include at least one of non-volatile memory and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this disclosure may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processor involved in each embodiment provided in this disclosure may be, but are not limited to, a general-purpose processor, a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processor (DSP), a programmable logic unit (PLC), a data processing logic unit based on quantum computing, an artificial intelligence (AI) processor, etc.

[0117] The above-described embodiments merely represent several implementation methods of the present disclosure. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of patent protection. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present disclosure, all of which fall within the scope of protection of the present disclosure. Therefore, the scope of protection of the present disclosure shall be determined by the appended claims.

Claims

1. A method for processing scan data, characterized in that include: Get scan data; as well as Processing the scan data using a preset deep learning network to obtain a processing result of the scan data; Wherein, the scanning data includes scanning data of multiple different energy spectra; Wherein, the deep learning network is configured to perform feature extraction on the scanning data of each energy spectrum among the scanning data of the multiple different energy spectra, and to perform feature extraction on the scanning data of various energy spectra among the scanning data of the multiple different energy spectra.

2. The method for processing scan data according to claim 1, characterized in that: The deep learning network also includes a single extraction layer for performing feature extraction on the scanning data of each energy spectrum, and a common extraction layer for performing feature extraction on the scanning data of various energy spectra.

3. The method for processing scan data according to claim 1, characterized in that: The features obtained by extracting the features from the scanning data of each energy spectrum are only used by the scanning data of the energy spectrum; the features obtained by extracting the features from the scanning data of various energy spectra are shared by the scanning data of the multiple energy spectra.

4. The method for processing scan data according to claim 1, characterized in that: The scanning data is obtained through a medical scanning device; the medical scanning device includes at least one of a photon counting computed tomography device, an energy spectrum computed tomography device, a direct digital photography device, a single photon emission computed tomography device, and a positron emission tomography device.

5. The method for processing scan data according to claim 1, characterized in that: The obtaining of scanning data comprises: acquiring initial scan data obtained by photon counting-based computed tomography; The initial scanning data is divided into a plurality of different energy spectrum segments to obtain scanning data of a plurality of different energy spectra respectively corresponding to the plurality of energy spectrum segments.

6. The method for processing scan data according to claim 1, characterized in that: The obtaining of scanning data comprises: Based on spectral computed tomography, scanning data of a plurality of different energy spectra respectively corresponding to the plurality of energy spectra are acquired.

7. The method for processing scan data according to claim 6, characterized in that: The method of acquiring scanning data of a plurality of different energy spectra corresponding to a plurality of energy spectra respectively based on energy spectrum computer tomography includes: Performing computer tomography scans using at least two different energies to obtain scan data corresponding to different energy spectra; or Computer tomography is performed based on two different light sources to obtain scanning data corresponding to different energy spectra.

8. The method for processing scan data according to any one of claims 1 to 7, characterized in that: Before using a preset deep learning network to process the scan data to obtain a processing result of the scan data, the method further includes: The scan data is used as input to the deep learning network.

9. The method for processing scan data according to any one of claims 2 to 8, characterized in that: The using a preset deep learning network to process the scan data to obtain a processing result of the scan data includes: The single extraction layer of the deep learning network is used to extract the data features of the scanning data of each energy spectrum, and obtain the single data features corresponding to the scanning data of different energy spectra; Using the common extraction layer of the deep learning network to jointly extract data features of scanning data of different energy spectra to obtain common data features; and Based on the single data feature and the common data feature, a processing result of the scanning data of the different energy spectra is obtained.

10. The method for processing scan data according to claim 9, characterized in that: The method of processing the scan data using a preset deep learning network to obtain a processing result of the scan data further includes: Based on the single data feature, the deep learning network is used to perform separate optimization processing on the scanning data of each energy spectrum; Based on the common data features, the deep learning network is used to jointly optimize the scanning data of different energy spectra.

11. The method for processing scan data according to any one of claims 1 to 10, characterized in that: The method further comprises: If the scan data is multi-energy spectrum data without material decomposition, material decomposition is performed on the output result of the deep learning network, and the material decomposition process is constrained based on a preset loss function to obtain a material map corresponding to the multi-energy spectrum data; or If the scanning data is material decomposition data, the output result of the deep learning network is converted into target energy spectrum data, and the loss function of the target energy spectrum data is calculated using the loss function structure in the deep learning network.

12. The method for processing scan data according to any one of claims 1 to 11, characterized in that: The processing of the scan data by the deep learning network includes at least one of the following processing: Denoising the scanning data of the plurality of different energy spectra; Extracting regions of interest from the scanning data of the plurality of different energy spectra; and The scanning results of the scanning data of the multiple different energy spectra are analyzed.

13. A device for processing scan data, characterized in that: include: an acquisition module configured to acquire scan data; as well as An output module, configured to process the scan data using a preset deep learning network to obtain a processing result of the scan data; Wherein, the scanning data includes scanning data of multiple different energy spectra; Wherein, the deep learning network is configured to perform feature extraction on the scanning data of each energy spectrum among the scanning data of the multiple different energy spectra, and to perform feature extraction on the scanning data of various energy spectra among the scanning data of the multiple different energy spectra.

14. The device for processing scan data according to claim 13, characterized in that: The deep learning network also includes a single extraction layer for performing feature extraction on the scanning data of each energy spectrum, and a common extraction layer for performing feature extraction on the scanning data of various energy spectra.

15. The method for processing scan data according to claim 14, characterized in that: The features obtained by extracting the features from the scanning data of each energy spectrum are only used by the scanning data of the energy spectrum; the features obtained by extracting the features from the scanning data of various energy spectra are shared by the scanning data of the multiple energy spectra.

16. The device for processing scan data according to claim 13, characterized in that: The acquisition module is further configured as follows: acquiring initial scan data obtained by photon counting-based computed tomography; The initial scanning data is divided into a plurality of different energy spectrum segments to obtain scanning data of a plurality of different energy spectra respectively corresponding to the plurality of energy spectrum segments.

17. The device for processing scan data according to claim 13, characterized in that: The acquisition module is further configured as follows: Based on spectral computed tomography, scanning data of a plurality of different energy spectra respectively corresponding to the plurality of energy spectra are acquired.

18. The device for processing scan data according to claim 17, characterized in that: The acquisition module is further configured as: Performing computer tomography scans using at least two different energies to obtain scan data corresponding to different energy spectra; or Computer tomography is performed based on two different light sources to obtain scanning data corresponding to different energy spectra.

19. The device for processing scan data according to any one of claims 13 to 18, characterized in that: The acquisition module is further configured as: The scanning data of the different energy spectra are respectively used as inputs of the deep learning network.

20. The device for processing scan data according to any one of claims 13 to 19, characterized in that: The output module is further configured as: The single extraction layer of the deep learning network is used to extract the data features of the scanning data of each energy spectrum, and obtain the single data features corresponding to the scanning data of different energy spectra; Extracting data features of different types of scan data using a common extraction layer of the deep learning network to obtain common data features; and Based on the single data feature and the common data feature, a processing result of the scanning data of the different energy spectra is obtained.

21. The apparatus for processing scan data according to claim 20, characterized in that: The output module is further configured as: Based on the single data feature, the deep learning network is used to perform separate optimization processing on the scanning data of each energy spectrum; Based on the common data features, the deep learning network is used to jointly optimize the scanning data of different energy spectra. deal with.

22. The device for processing scan data according to any one of claims 13 to 21, characterized in that: The output module is further configured as: If the scan data is multi-energy spectrum data without material decomposition, material decomposition is performed on the output result of the deep learning network, and the material decomposition process is constrained based on a preset loss function to obtain a material map corresponding to the multi-energy spectrum data; or If the scanning data is material decomposition data, the output result of the deep learning network is converted into target energy spectrum data, and the loss function of the target energy spectrum data is calculated using the loss function structure in the deep learning network.

23. The device for processing scan data according to any one of claims 13 to 22, characterized in that: The output module is further configured to perform at least one of the following processes: Denoising the scanning data of the plurality of different energy spectra; Extracting regions of interest from the scanning data of the plurality of different energy spectra; and The scanning results of the scanning data of the multiple different energy spectra are analyzed.

24. A computer device comprising a processor and a memory, wherein the memory stores a computer program executable by the processor, and when the processor executes the computer program, the method for processing scan data according to any one of claims 1 to 12 is implemented.

25. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the computer program causes the processor to implement the method for processing scan data according to any one of claims 1 to 12.

Images