Edge cloud collaborative low-dose CT imaging method

Abstract

The invention discloses an edge cloud collaborative low-dose CT imaging method. The method is specifically implemented according to the following steps that 1, an edge end initiates a scanning request to a cloud end; step 2, the cloud end generates a positioning scanning protocol after receiving the scanning request, and the edge end forwards the received protocol to the CT terminal to generate a positioning image; 3, transmitting the positioning image to an edge node of an edge end by the CT terminal; after being packaged by the edge node, the data are uploaded to a cloud to generate a fine scanning protocol; 4, the CT end executes formal scanning according to a fine scanning protocol to obtain a low-dose image and uploads the low-dose image to the cloud end; and step 5, the cloud end executes high-quality image reconstruction and analysis after confirming the image pre-reconstruction result, generates a standard image, and feeds back the standard image and an analysis result to the edge end. According to the side cloud collaborative low-dose CT imaging method disclosed by the invention, the problem that the imaging process of a traditional local CT system cannot be interacted in the prior art is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical computer photography methods, and specifically relates to an edge-cloud collaborative low-dose CT imaging method. Background Art

[0002] With the mature application of "edge computing" in fields such as the Industrial Internet, its proximity to data sources, low-latency processing, and reduced central load are gradually attracting attention in the medical field. In particular, in medical imaging scenarios, the large size and high transmission sensitivity of CT raw data make it an ideal application for edge-cloud collaborative architectures. By offloading some image processing or intermediate result preprocessing capabilities to edge nodes, not only can cloud bandwidth pressure be alleviated, but real-time control, anomaly detection, and rapid response can also be achieved during the imaging process.

[0003] First, existing CT imaging methods are generally divided into two common modes: one is fully localized imaging, where both scanning and imaging are performed locally on a highly integrated CT device; the other is distributed CT cloud imaging. Application number CN202310506674.8 discloses a distributed CT imaging and intelligent diagnosis and treatment system and method, in which the raw scanned data is uploaded to the cloud, where the cloud centrally schedules imaging and subsequent processing. Both current methods have their advantages and disadvantages: local imaging is limited by hardware configuration, has high update costs, and the equipment is highly integrated, with poor deployment flexibility, making it unsuitable for promotion in grassroots hospitals. Distributed CT imaging, on the other hand, breaks down the hardware into sensor devices responsible only for scanning, with all computation performed in the cloud. However, the large volume of raw data generated by a single scan (typically approximately 1.8GB) must be uploaded to a remote cluster, resulting in significant network bandwidth consumption, significantly increasing the system's reliance on network transmission stability and speed, and the high transmission costs. This also limits its widespread application in remote areas and low-resource environments. Secondly, although current distributed CT systems have proven capable of achieving low-dose or ultra-low-dose imaging with the support of the cloud's powerful computing resources, they are still not suitable for widespread use in remote areas and low-resource environments. However, the entire imaging process suffers from the problem of "entire cloud processing with no real-time feedback." After the scan is complete, the raw data is uploaded to the cloud, where image reconstruction, processing, and display are performed. In this mode, frontline technicians and physicians cannot obtain intermediate imaging results in real time, nor can they dynamically adjust or interact with the imaging process. In the event of an imaging failure (such as improper parameter settings or reconstruction failure), the only option is to wait until the imaging is completed to detect the error and reconstruct the entire process, resulting in time delays and wasted computing resources, seriously affecting the timeliness of diagnosis and treatment.

[0004] In summary, due to the lack of "edge-cloud" collaboration capabilities in the existing imaging process, it is neither possible to achieve real-time interaction and dynamic control during the imaging process, nor can it be evolved intelligently at the parameter optimization level. Whether it is local imaging or distributed cloud imaging, "non-interactivity" and "insufficient intelligence" have become the core technical barriers restricting the development of CT technology towards high efficiency and low dose. Once problems such as unclear images, parameter disorders, and algorithm abnormalities occur, it is only possible to discover the problems after the entire reconstruction is completed. Often, it is necessary to re-upload the data and restart the imaging process, resulting in diagnostic delays and waste of computing resources. Existing systems generally do not have dynamic control capabilities such as edge preprocessing, real-time interruption, and stage reconstruction. And in clinical practice, the settings of CT imaging parameters such as voltage, current, pitch, and scan mode have an important impact on image quality and radiation safety. However, at present, the vast majority of systems rely on technicians to manually configure based on experience, lacking intelligent assistance or a parameter recommendation mechanism driven by historical data, which is prone to human errors and difficult to fully explore the low-dose imaging potential in existing data. Summary of the Invention

[0005] The object of the present invention is to provide a low-dose CT imaging method with edge-cloud collaboration, which solves the problem of non-interactivity during the imaging process of traditional local CT systems in the prior art.

[0006] The technical solution adopted by the present invention is that the low-dose CT imaging method with edge-cloud collaboration is specifically implemented according to the following steps: Step 1, the edge end initiates a scan request to the cloud and starts the scout image process; Step 2, after receiving the scan request, the cloud generates a scout scan protocol and sends it back to the edge end. The edge end forwards the received protocol to the CT terminal to perform the scout image scan and generate a scout image; Step 3, the scout image is transmitted from the CT terminal to the edge node at the edge end; after being encapsulated by the edge node, it is uploaded to the cloud to generate a fine scan protocol; Step 4, the CT end performs the formal scan according to the fine scan protocol, and the edge node performs pre-reconstruction of the low-dose image to obtain a low-dose image and upload it to the cloud; Step 5, after the cloud confirms the result of the image pre-reconstruction, it performs high-quality image reconstruction and analysis to generate a standard DICOM image, and feeds back the standard DICOM image and the analysis result to the edge end.

[0007] The characteristics of the technical solution of the present invention also lie in that Step 1 is specifically as follows: After the patient is placed on the scanning bed, the edge device automatically sends a scanning request to the cloud. The request contains the device number, patient identity information, and key data fields of the examination site. After receiving the request, the cloud interface prompts that there is a pending request for the user to review, accept, or for the cloud to automatically process.

[0008] Step 2 is specifically as follows: Based on the received patient identity information and the protocol templates existing in the historical database, the cloud automatically generates a positioning scan protocol in a way based on parameter mean regression. The positioning scan protocol controls the start and end points of the scanning bed and the initial image window width and window level, and transmits the positioning scan protocol back to the edge device in the form of a standard structured data packet. The edge device forwards the received protocol to the CT terminal, triggering the CT device to perform a positioning image scan and generate a positioning image.

[0009] Step 3 is specifically as follows: The generated positioning image is transmitted from the CT terminal to the edge node of the edge device, and after being encapsulated by the edge node, it is uploaded to the cloud. The cloud analyzes the positioning image, identifies the scanning site, body posture, and target structure, and combines the patient's height, weight, age, and task type to generate a fine scan protocol through a protocol optimization model. The fine scan protocol includes key fields such as scanning voltage, tube current, slice thickness, slice interval, Pitch, and dose modulation parameters, which control the radiation dose and ensure the imaging quality.

[0010] Step 4 is specifically as follows: The CT terminal performs a formal scan according to the fine scan protocol, generating raw projection data; at the same time, the edge device calls a lightweight pre-reconstruction model to complete the preliminary reconstruction of the positioning image locally, initially generating a low-dose image; the initially generated low-dose image and basic quality parameters are uploaded to the cloud for the fine generation of the image.

[0011] Step 5 is specifically as follows: The cloud evaluates the received low-dose image to confirm whether it meets the current diagnostic requirements. If not, the cloud generates a correction protocol and transmits it back to the edge device for re-scanning; if it meets the requirements, image reconstruction is performed. The cloud calls the imaging container through a high-performance GPU cluster and schedules the corresponding reconstruction algorithm model according to the task type to complete the generation of the standard DICOM image.

[0012] It also includes Step 6, which is specifically as follows: After the standard DICOM image reconstruction is completed, it is automatically pushed to the cloud imaging service platform for diagnosis to generate an analysis result; after the cloud completes the reconstruction and diagnosis, the standard DICOM image and the analysis result are fed back to the edge terminal.

[0013] The cloud imaging service platform includes 2D and 3D image browsing and rendering, AI model for identifying target organs, lesions, abnormal density regions, automatic report writing and structured output, and synchronization and archiving with the PACS system.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) The edge-cloud collaborative low-dose CT imaging method provided by the present invention introduces an edge collaborative terminal, sinking part of the CT image pre-reconstruction ability to the edge side, enabling the system to operate without relying on a high-end GPU host. At the same time, it reduces the rigid demand for a large bandwidth network for directly uploading raw data in distributed CT, facilitating deployment and application in grass-roots, community, and environments with limited network conditions.

[0015] (2) The edge-cloud collaborative low-dose CT imaging method provided by the present invention supports a mechanism for rapid feedback of pre-reconstructed images of raw data scanned at the edge and confirmation by the cloud. It can judge whether the scanning process is qualified during the scanning process, solving the problem that distributed CT completely places the entire reconstruction process in the cloud, avoiding unnecessary full-process uploading and reconstruction, and reducing the waiting time for diagnosis and treatment and resource waste.

[0016] (3) The edge-cloud collaborative low-dose CT imaging method provided by the present invention integrates patient individual parameters, positioning image parameters, and historical image quality data, and automatically generates a low-dose scanning protocol that meets the diagnostic requirements in an interactive form in the cloud and sends it to the edge side for technicians to confirm. It not only solves the problem of constant parameter settings and uneven technician levels, but also improves the consistency of image quality, effectively controls the radiation dose, and contributes to the construction of a green and safe imaging diagnosis and treatment system.

[0017] (4) The edge-cloud collaborative low-dose CT imaging method provided by the present invention, through a task-driven mechanism, the system realizes full-process management from scanning protocol generation, data acquisition, image pre-reconstruction, image fine reconstruction to result feedback and PACS archiving. Images and reports can flow seamlessly at the edge, in the cloud, and within the hospital, effectively supporting the needs of hierarchical diagnosis and treatment and remote consultation. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is a schematic flow chart of the edge-cloud collaborative low-dose CT imaging method of the present invention; Figure 2 is a schematic structural diagram of the cloud imaging center in the embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0019] The present invention will be described in detail below with reference to the drawings and specific embodiments.

[0020] Embodiment 1 The present invention provides an edge-cloud collaborative low-dose CT imaging method, as Figure 1-2 shown, which is specifically implemented according to the following steps: Step 1, the edge side initiates a scanning request to the cloud and starts the positioning image process; Step 2: After the cloud receives the scanning request, it generates a positioning scan protocol and sends it back to the edge side. The edge side forwards the received protocol to the CT terminal to perform a positioning image scan and generate a positioning image. Step 3: The positioning image is transmitted from the CT terminal to the edge node on the edge side; after being encapsulated by the edge node, it is uploaded to the cloud to generate a fine scan protocol. Step 4: The CT side performs a formal scan according to the fine scan protocol, and the edge node performs pre-reconstruction of the low-dose image, obtains the low-dose image and uploads it to the cloud. Step 5: After the cloud confirms the result of the image pre-reconstruction, it performs high-quality image reconstruction and analysis to generate a standard DICOM image, and feeds back the standard DICOM image and the analysis result to the edge side.

[0021] Embodiment 2 Based on Embodiment 1, Step 1 is specifically as follows: After the CT technician places the patient on the scanning bed and starts the operation of the terminal interface, the edge collaborative terminal automatically sends a scanning request to the cloud imaging center. This request includes key data fields such as the device number, patient identity information, and examination site. After the cloud system receives the request, the interface will prompt that there is a pending request for the user to review, receive, or the system to automatically process, thus starting the protocol generation in the next stage.

[0022] Embodiment 3 Based on Embodiment 2, Step 2 is specifically as follows: The cloud automatically generates a positioning scan protocol based on the received patient identity information and the protocol templates existing in the historical database in the way of parameter mean regression. The positioning scan protocol controls the start and end points of the scanning bed and the initial image window width and window level. The positioning scan protocol is sent back to the edge side in the form of a standard structured data packet. The edge side forwards the received protocol to the CT terminal to trigger the CT device to perform a positioning image scan and generate a positioning image.

[0023] Embodiment 4 Based on Embodiment 3, Step 3 is specifically as follows: The generated positioning image is transmitted from the CT terminal to the edge node on the edge side. After being encapsulated by the edge node, it is uploaded to the cloud. The cloud analyzes the positioning image, identifies the scanning site, body posture, and target structure, combines the patient's height, weight, age, and task type, and generates a fine scan protocol through a protocol optimization model. The fine scan protocol includes key fields such as scanning voltage, tube current, slice thickness, slice interval, Pitch, and dose modulation parameters, which control the radiation dose and ensure the imaging quality.

[0024] The specific process of generating the fine scan protocol through the protocol optimization model here is as follows: Use a Convolutional Neural Network (CNN) to extract the features of the scanned part, body posture, and target structure of the localization image, normalize information such as the patient's height, weight, and age to adapt to the model input, concatenate the feature vectors of the localization image and patient information (height, weight, age), and extract the influence of patient individual differences on the scanning parameters through a multi-layer Feed Forward Neural Network (FFNN). Train and optimize the model through historical data (including the patient's scan records and corresponding image quality assessment results) so that it can generate the best scanning protocol based on the input image features, patient information, and task type.

[0025] Example 5 Based on Example 4, step 4 is specifically as follows: The CT terminal performs a formal scan according to the fine scanning protocol to generate raw projection data; at the same time, the edge terminal calls a lightweight pre-reconstruction model to preliminarily reconstruct the localization image locally and preliminarily generate a low-dose image; upload the preliminarily generated low-dose image and basic quality parameters to the cloud for fine image generation.

[0026] Example 6 Based on Example 5, step 5 is specifically as follows: The cloud evaluates the received low-dose image to confirm whether it meets the current diagnostic requirements. If not, the cloud generates a correction protocol and sends it back to the edge terminal for re-scanning; if it meets the requirements, image reconstruction is performed. The cloud calls the imaging container through a high-performance GPU cluster and schedules the corresponding reconstruction algorithm model according to the task type. The task type refers to calling different deep learning-based low-dose reconstruction algorithms (DLR) according to the scanned part of the patient, such as the lungs, abdomen, brain, etc., to perform image reconstruction, so as to generate a standard DICOM image.

[0027] Example 7 Based on Example 6, it further includes step 6, and step 6 is specifically as follows: After the standard DICOM image reconstruction is completed, it is automatically pushed to the cloud imaging service platform for diagnosis to generate an analysis result; after the cloud completes the reconstruction and diagnosis, the standard DICOM image and the analysis result are fed back to the edge terminal. The cloud imaging service platform includes 2D and 3D image browsing and rendering, AI model for identifying target organs, lesions, abnormal density regions, automatic report writing and structured output, and synchronization and archiving with the PACS system.

[0028] The low-dose CT imaging method with edge-cloud collaboration provided by the present invention optimizes the distributed CT process and proposes a new CT imaging paradigm of "edge-cloud collaboration": while ensuring imaging performance and image quality, it fully combines the advantages of the local and cloud sides, optimizes the phased imaging of low-dose CT through an edge computing + cloud collaboration architecture, thereby optimizing the data transmission size, enhancing the real-time interaction ability, introducing an imaging protocol recommendation mechanism, and supporting the calculation and feedback of low-dose scanning strategies for patient individual characteristics, so as to achieve an overall improvement in the aspects of diagnosis and treatment efficiency, system scalability, and imaging.

Claims

1. Edge-cloud collaborative low-dose CT imaging method, characterized in that The implementation is specifically carried out according to the following steps: Step 1: The edge device sends a scanning request to the cloud, and starts the positioning image process; Step 2: After receiving the scanning request, the cloud generates a positioning scanning protocol and sends it back to the edge device. The edge device forwards the received protocol to the CT terminal to perform the positioning image scan and generate a positioning image; Step 3: Transfer the positioning image from the CT terminal to the edge node of the edge device; After being encapsulated by the edge node, it is uploaded to the cloud to generate a fine scanning protocol; Step 4: The CT terminal performs the formal scan according to the fine scanning protocol. The edge node performs low-dose image pre-reconstruction, obtains the low-dose image and uploads it to the cloud; Step 5: After the cloud confirms the result of the image pre-reconstruction, it performs high-quality image reconstruction and analysis, generates a standard DICOM image, and feeds back the standard DICOM image and the analysis result to the edge device.

2. The edge-cloud collaborative low-dose CT imaging method according to claim 1, wherein The specific content of Step 1 is as follows: After the patient is placed on the scanning bed, the edge device automatically sends a scanning request to the cloud. The request includes the device number, patient identity information, and key data fields of the examination site. After receiving the request, the cloud interface prompts that there is a pending request for the user to review, receive, or the cloud to automatically process.

3. The edge-cloud collaborative low-dose CT imaging method according to claim 1, wherein The specific content of Step 2 is as follows: The cloud automatically generates a positioning scanning protocol based on the received patient identity information and the protocol templates existing in the historical database in the way of parameter mean regression. The positioning scanning protocol controls the start and end points of the scanning bed and the initial image window width and window level. The positioning scanning protocol is sent back to the edge device in the form of a standard structured data packet. The edge device forwards the received protocol to the CT terminal to trigger the CT device to perform the positioning image scan and generate a positioning image.

4. The edge-cloud collaborative low-dose CT imaging method according to claim 1, wherein The specific content of Step 3 is as follows: The generated positioning image is transferred from the CT terminal to the edge node of the edge device. After being encapsulated by the edge node, it is uploaded to the cloud. The cloud analyzes the positioning image, identifies the scanning site, body posture, and target structure, combines the patient's height, weight, age, and task type, and generates a fine scanning protocol through a protocol optimization model. The fine scanning protocol includes key fields such as scanning voltage, tube current, slice thickness, slice interval, Pitch, and dose modulation parameters, which control the radiation dose and ensure the imaging quality.

5. The edge-cloud collaborative low-dose CT imaging method according to claim 1, wherein The specific content of Step 4 is as follows: The CT terminal performs the formal scan according to the fine scanning protocol to generate the original projection data. At the same time, the edge device calls a lightweight pre-reconstruction model to complete the preliminary reconstruction of the positioning image locally and initially generate a low-dose image. The initially generated low-dose image and basic quality parameters are uploaded to the cloud for fine image generation.

6. The edge-cloud collaborative low-dose CT imaging method according to claim 1, characterized in that, The specific content of Step 5 is as follows: The cloud evaluates the received low-dose image to confirm whether it meets the diagnostic requirements of this time. If it does not meet the requirements, the cloud generates a correction protocol and sends it back to the edge device for re-scanning. If it meets the requirements, image reconstruction is performed. The cloud calls the imaging container through a high-performance GPU cluster and schedules the corresponding reconstruction algorithm model according to the task type to complete the generation of the standard DICOM image.

7. The edge-cloud collaborative low-dose CT imaging method according to claim 1, wherein It further includes step 6, specifically: after the standard DICOM image reconstruction is completed, it is automatically pushed to the cloud imaging service platform for diagnosis to generate an analysis result; after the cloud completes the reconstruction and diagnosis, the standard DICOM image and the analysis result are fed back to the edge terminal.

8. The edge-cloud collaborative low-dose CT imaging method according to claim 1, characterized in that, The cloud imaging service platform includes 2D and 3D image browsing and rendering, AI model for identifying target organs, lesions, abnormal density areas, automatic report writing and structured output, and PACS system synchronous archiving.

Images