Automated quality control of medical images

Abstract

System for automated quality control of medical images (B), comprising the following units: - Acquisition unit (AE), designed for CT acquisition (A) of an image (B) using an imaging device (1), wherein the imaging device (1) is a CT device, - a processing unit (PU) spatially separated from the recording unit (AE), designed for processing (V) the image (B), - an analysis unit (AyE) designed for the analysis (Ay) of the image (B), for the evaluation (Bu) of the analysis (Ay) and for the calculation of an evaluation signal (BuS), wherein the analysis unit (AyE) is assigned to the acquisition unit (AE), and wherein the analysis (Ay) includes at least part of the processing (V), where the processing unit (PU) is designed to perform more computationally intensive image processing steps than the analysis unit (AyE), the system is designed to transfer a result (Er) of the analysis (Ay) from the analysis unit (AyE) to the processing unit (VE), wherein the analysis (Ay) includes the detection (E) of at least one physical structure, wherein the result (Er) is a success in the detection (E) of the at least one physical structure.

Description

[0001] The invention relates to a system for automated quality control of medical images according to claim 1 and a method for automated quality control of medical images according to claim 7.

[0002] Imaging techniques such as magnetic resonance imaging (MRI), computed tomography (CT), and rotational angiography are essential components of medical diagnostics. Modern imaging techniques rely on digital acquisition technology, enabling computer-aided processing of the images. For many imaging procedures, such as tomography, computer-aided image processing is even necessary. This processing includes steps such as creating cross-sectional images, registering images, fusing multiple images, segmenting, and rendering to display three-dimensional datasets on a screen. Erroneous images, resulting from incorrect acquisition parameters (such as the imaging area, acquisition frequency, or radiation intensity), must be avoided.Firstly, repeating an image generally means increased time and expense, as well as – depending on the imaging technique used – additional radiation exposure. Furthermore, in conventional medical workflows, the processed images are only available after the patient has left the facility. A faulty image often only becomes apparent after processing, i.e., after the patient has left. This increases the time and expense required for a new image. Additionally, there are no automatic quality filters for the processed images, so the decision regarding image quality often rests with the treating physician / radiographer.

[0003] US 7 672 491 B2 discloses a computer-based method for automatic decision support in medical imaging, comprising the following steps: - Capturing an image with image data, - Extracting characteristic data from the image data, - Automatic assessment of an error in the image using the extracted characteristic data, wherein the error corresponds to a diagnostic feature of the image, wherein the automatic assessment of an error using the extracted characteristic data is performed by querying a database of templates derived from information on known cases to identify similar templates and using the information associated with the identified similar templates to determine the diagnostic feature.

[0004] US 7 298 876 B1 discloses a method for quality control of radiological images.

[0005] DE 10 2010 009 105 A1 discloses a method for modifying at least one recording parameter and / or at least one reconstruction parameter, wherein an evaluation of the quality of at least one generated image with respect to at least one evaluation criterion is provided.

[0006] DE 10 2009 015 007 A1 discloses a method for evaluating a time series of, in particular, two-dimensional images of a contrast agent flow in at least one blood vessel of the human body as part of a test bolus measurement.

[0007] US 2006 / 0193437A1 discloses a method for controlling an imaging modality, wherein object-specific data are acquired and an example raw data set is automatically selected from a number of example raw data sets based on the object-specific data.

[0008] US 2004 / 0227069A1 discloses a method for evaluating the image quality of a radiological image.

[0009] DE 10 2011 090 047 A1 concerns a control procedure for determining a quality indicator of medical-technical imaging result data from a contrast-enhanced tomography scan.

[0010] DE 10 2012 214 513 A1 relates to a method for taking an X-ray image, in particular a topogram, and its automated quality control.

[0011] The purpose of the invention is to assess medical images more quickly and safely.

[0012] The problem is solved by a system according to claim 1 and by a method according to claim 7.

[0013] The inventive solution to the problem is described below with respect to both the claimed system and the claimed method. Features, advantages, or alternative embodiments mentioned herein are also applicable to the other claimed subject matter and vice versa. In other words, the claims (which, for example, relate to a system) can also be further developed with the features described or claimed in connection with a method. The corresponding functional features of the method are thereby implemented by corresponding material modules.

[0014] The system according to the invention for automated quality control of medical images comprises a recording unit designed for CT image acquisition using an imaging device, wherein the imaging device is a CT device, and a processing unit spatially separated from the recording unit, designed for processing the image.The invention is based on the idea of ​​assigning an analysis unit, designed for analyzing the image, evaluating the analysis, and calculating an evaluation signal, to the acquisition unit, and further designing the analysis unit such that the analysis comprises at least a part of the processing, wherein the processing unit is designed to perform more computationally intensive image processing steps than the analysis unit, wherein the system is designed to transfer a result of the analysis from the analysis unit to the processing unit, wherein the analysis comprises the recognition of at least one physical structure, and wherein the result is a success in the recognition of the at least one physical structure.

[0015] Direct image analysis, which incorporates at least some conventional processing steps, enables direct assessment and the direct calculation of an assessment signal, resulting in faster and more reliable image evaluation. This faster and more reliable evaluation also leads to an overall accelerated diagnostic workflow. In particular, it becomes possible to identify a faulty image before the patient leaves the imaging facility, allowing for an immediate re-exposure. This avoids the need for the patient to return for a second examination in the event of a faulty initial image.

[0016] In another embodiment, the acquisition unit is designed to perform a series of CT scans, wherein the analysis and evaluation includes at least the analysis and evaluation of a portion of the acquired images in the series. This allows an entire series of images to be evaluated more quickly and reliably.

[0017] In another embodiment, the recording unit is designed to abort the CT scan of an image series depending on the assessment signal, thereby avoiding scans with incorrect settings.

[0018] In another embodiment, the image analysis includes segmentation of the image, which facilitates the interpretation of the image, particularly with regard to the recognition of structures.

[0019] According to the invention, the analysis includes the detection of at least one physical structure, thereby increasing the significance of the subsequent assessment.

[0020] In another embodiment, the assessment includes the calculation of improved parameters for the CT scan of an image, so that a new scan results in an improved image.

[0021] In a further embodiment, the assessment includes an output of the assessment signal in the form of a symbol and / or text by means of an output unit assigned to the receiving unit, so that the meaning of the assessment signal is immediately recognizable to the operator of the medical device.

[0022] According to the invention, the system is designed to transfer the result of an analysis from the analysis unit to the processing unit. Since the analysis unit has already performed some processing steps, transferring the result of the analysis simplifies and shortens the processing time.

[0023] The invention will now be described and explained in more detail with reference to the exemplary embodiments shown in the figures.

[0024] They show: Fig. 1. A system for automated quality control, Fig. 2 the scheme of a system for automated quality control, and Fig. 3. The flowchart of a procedure for automated quality control.

[0025] Fig. Figure 1 shows a system for automated quality control. The system shown here is designed as a device. The system comprises an imaging device 1, namely a CT scanner, which has an acquisition unit AE comprising a radiation emitter 8 and a radiation detector 9. The radiation emitter 8 is typically an X-ray tube. The radiation detector 9 for a CT scanner is typically a line or flat-panel detector, but it can also be a scintillator counter or CCD camera. During the acquisition A of a medical image B, the patient 5 lies on a patient table 6, which is connected to a table base 4 in such a way that it supports the patient table 6 with the patient 5. The patient table 6 moves the patient 5 along an acquisition direction through the opening 10 of the acquisition unit AE. During this movement, an image B of an examination area of ​​the patient 5 is acquired.

[0026] Image B can be purely two-dimensional, consisting of pixels, or three-dimensional, consisting of voxels, and it can exist in both the spatial and frequency domains. In particular, this image B may have already been pre-processed before the analysis Ay claimed here, for example, by creating a two-dimensional cross-sectional image based on a tomographic scan A.

[0027] The image B captured by the acquisition unit AE is transmitted to an analysis unit AyE, which is integrated into a computer. The analysis unit AyE is spatially assigned to the medical device 1 and thus to the acquisition unit AE; in particular, the analysis unit AyE and the acquisition unit AE can be located together in a first room 2. Furthermore, the analysis unit AyE is connected to an output unit 11 and an input unit 7.

[0028] Furthermore, the system includes a processing unit VE, which is spatially separated from the analysis unit AyE and the recording unit AE, for example, as shown here in a second room 3. The processing unit VE is integrated into a computer and connected to a second output unit 12 and a second input unit 13.

[0029] Output unit 11 and the second output unit 12 are, for example, each one (or more) LCD, plasma, or OLED screen(s). The output on output unit 11 or the second output unit 12 can include, for example, the output of images B as well as analyzed and processed images B. Input unit 7 and the second input unit 13 are, for example, each a keyboard, a mouse, a touchscreen, or a microphone for voice input.

[0030] Fig. Figure 2 shows the schematic of an automated quality control system. The acquisition unit AE of an imaging device 1, namely a CT scanner, is designed to acquire a CT scan (A) of image B. The image B acquired by the acquisition unit AE is transmitted to the analysis unit AyE, which is designed to analyze image B (Ay), evaluate the analysis (Bu), and calculate an evaluation signal (BuS). The analysis unit AyE is assigned to the acquisition unit AE, so that both are located in the first room 2. Also located in the first room 2 is an output unit 11, which can output the evaluation signal (BuS), image B, or an analyzed image B. The evaluation signal (BuS) can also be sent back to the acquisition unit AE to acquire another image B (A).

[0031] Furthermore, image B is transferred directly to processing unit VE, which is located in a second room 3. Processing unit VE is designed to process image B. Additionally, analysis unit AyE can directly transfer results Er from analysis Ay to processing unit VE. Both images B and processed images B, or the results Er, can be output on the second output unit 12, which is connected to processing unit VE.

[0032] The analysis unit AyE can be configured as a client, and the processing unit VE as a server. Both the analysis unit AyE and the processing unit VE are designed to transfer images B, or analyzed or processed images B, to a Picture Archiving and Communication System.

[0033] The processing V by the processing unit VE of an image B basically includes the processing steps known as image post-processing, i.e. all those steps that are carried out after the reconstruction of the image B from raw data and a simple preprocessing, for example filtering.

[0034] Processing V includes, for example, the segmentation S of images B using a thresholding method or a region-oriented method such as region growing or region splitting, or edge extraction. Furthermore, the automatic recognition E of a specific physical structure, such as an organ or blood vessels, is a typical component of processing V. Such recognition E can be achieved, in particular, through the use of neural networks and / or machine learning, as well as classification algorithms.

[0035] Furthermore, specific image representations (B) are the target of processing (V), often based on prior segmentation (S) and / or recognition (E). For example, image post-processing can highlight or hide certain physical structures; bones in a CT scan can be hidden. Processing (V) also includes a quantitative evaluation of the partially processed images (B). For instance, certain recognized structures can be automatically measured spatially, or parameters describing blood flow rates can be calculated from time series of images (B).

[0036] All these processing steps have in common that they are usually computationally intensive. Therefore, they are typically performed on a high-performance processing unit (PU) that is spatially separate from the acquisition unit (AU). However, this has the disadvantage that quality control often only takes place once the images (B) have been completely processed. In the event of a faulty acquisition (A), repeating the acquisition (A) requires considerable effort, e.g., the patient (5) having to travel to the facility again, administering contrast agent again, etc.

[0037] Therefore, essential processing steps V in the claimed method are carried out by the analysis unit AyE assigned to the acquisition unit AE, taking into account the technical characteristics of the analysis unit AyE. Essential in the context of the invention are those steps that serve the quality control of the acquired image B. Such steps are, in particular, the segmentation S of the image B and the recognition E of a physical structure.

[0038] In the embodiment described here, it is further provided that only one user has access to the analysis unit AyE at any given time, so that the individual processing and analysis steps can be executed relatively quickly. Typically, however, many users of medical devices 1 have access to a processing unit VE, resulting in a large number of requests for processing V being sent to the processing unit VE. Therefore, delays often occur when providing the processed images B. By shifting certain processing steps to the analysis unit AyE assigned to the acquisition unit AE, the analysis Ay of an image B or a series of images becomes possible immediately after or even during acquisition A.

[0039] The processing unit (VE) has a larger memory than the analysis unit (AyE) because the VE is designed to perform more computationally intensive image processing steps. Furthermore, the VE is designed to execute multiple processing steps in parallel, as it can receive multiple processing requests (V) from different users. The analysis unit (AyE), on the other hand, is designed to process the request from a single user.

[0040] To facilitate processing V, analysis Ay, and assessment Bu, the operator of the medical device 1 is permitted to specify, before image A is taken, which physical structure is to be depicted and, if applicable, identified, as well as which type of analysis Ay is to be performed. For example, the operator can select the "Coronary Artery Segmentation" mode from a list displayed via a graphical user interface. This informs the analysis unit AyE which algorithms it must apply for analysis Ay.

[0041] The assessment Bu of the analysis Ay also depends on the selected mode. For example, the underlying question for assessment Bu might be: "Were fewer or more than 10 arteries detected?" The assessment Bu could then be: "More than 10 arteries were detected," or "Segmentation failed, and therefore no arteries could be detected." Assessment Bu can also refer to whether certain structures, such as organs, are completely depicted in image B, or how much of an error a calculated, derived value, such as the blood flow rate, contains. The exact assessment criteria do not need to be defined before the start of the image acquisition; they can also be entered or selected by an operator after the acquisition has been completed.

[0042] Furthermore, the assessment Bu includes the calculation of improved acquisition parameters. An acquisition parameter is generally understood to be any parameter that influences the CT acquisition A of an image B, e.g., the X-ray tube voltage, the patient's position, or the scan area. For example, if the signal in a CT scan is too weak, the analysis unit AyE can suggest an improved, in this case higher, radiation dose as part of the assessment Bu. If a structure, such as the heart, is not fully depicted in a topogram used to calculate the radiation dose of a CT scan A, then the analysis unit AyE can suggest a shifted or longer acquisition area as part of the assessment Bu. These parameters can be output as an assessment signal BuS on output unit 11.For example, such an output might read: "Should the topogram be repeated and the scan area shifted 15 cm towards the head?" The calculation and output of improved recording parameters has the advantage that a renewed, improved recording is immediately possible.

[0043] Furthermore, the analysis unit AyE is designed to calculate an assessment signal BuS. An assessment signal BuS is a statement or question that contains quantitative or qualitative information about image B based on the assessment Bu. For example, the assessment signal BuS can refer to the image quality or the success or failure of certain steps in the analysis Ay. The assessment signal BuS is output on output unit 11 as text or as a symbol. A symbolic output Au could, for example, be a traffic light with color coding: "red" for failure, "green" for success. The calculation of an assessment signal BuS and its output Au thus leads to simplified quality control of image B by the operator and ultimately to improved recordings A of image B.Based on the generated assessment signal BuS, a decision is made as to whether or not a further assessment A is necessary. Furthermore, the assessment signal BuS can also contain information about improved parameters for a renewed assessment A.

[0044] The assessment signal BuS can be designed not only as information for the operator of the medical device 1, but also as information transmitted to the acquisition unit AE. If a series of images B is acquired, for example, to observe the influx of a contrast agent, then acquisition A of the series can be aborted based on the assessment Bu performed during acquisition A and the calculated assessment signal BuS. If, for example, the first images B of the series already show that the contrast agent has influx (due to a delayed start of acquisition A of the series), then the series is automatically aborted. The output Au of the assessment signal BuS then reads: "Series aborted due to premature contrast agent influx."“However, the system can also be designed so that the termination of a series is only suggested, and the operator of the medical device 1 must decide whether or not to terminate the series. Furthermore, the system can be designed so that the degree of automation regarding the termination of a series based on the assessment signal BuS can be selected by the operator before the start of the series' acquisition A. The (optional) interruption of a series depending on the assessment signal BuS offers the advantage of reducing the radiation exposure for the patient 5 as well as the duration of the entire acquisition process (including a repetition of acquisition A).

[0045] Furthermore, the system can be designed so that an analysis Ay and an assessment Bu are possible during the acquisition A of a single image B. For example, if a topogram is being acquired, the analysis Ay together with the assessment Bu may reveal, even during the acquisition A of the topogram, that the set radiation intensity is too high (or too low). In such a case, an automatic interruption of the acquisition A is also advisable to avoid unnecessarily high radiation exposure for the patient 5; because if the radiation intensity is too high (or too low), the topogram should be repeated anyway.

[0046] Furthermore, the result Er of the analysis Ay is transferred from the analysis unit AyE to the processing unit VE, where the result Er is a success in the recognition E of at least one physical structure. The transfer of the result Er reduces the time required by the processing unit VE for processing V. This is because the analysis A anticipates certain steps of processing V.

[0047] Fig.Figure 3 shows the flowchart of an automated quality control process. This involves the acquisition (A) of images (B), which are then transferred from the acquisition unit (AE) to an analysis unit (AyE) for analysis (Ay), comprising segmentation (S) and recognition (E). The results of the analysis (Eer) are transferred to a processing unit (VE) for processing (V). Furthermore, the images (B) undergo an evaluation (Bu), based on which an evaluation signal (BuS) is calculated. This evaluation signal (BuS) is then transferred to an output unit (Au) for output (Au).

[0048] In an advantageous embodiment, the recording process A is performed first, followed by the analysis Ay, wherein the segmentation S and then the recognition E are carried out first within the analysis Ay. After the recognition E, the evaluation Bu is performed, and (within the framework of the evaluation Bu) an evaluation signal BuS is calculated. As the final step of the embodiment, the processing V is carried out.

[0049] Although the invention has been further illustrated and described by the preferred embodiments, the invention is not limited by the disclosed examples and other variations can be derived by a person skilled in the art without departing from the scope of protection of the invention. In particular, process steps can be carried out in a different sequence than those specified.

Claims

[1] System for automated quality control of medical images (B), comprising the following units: - Acquisition unit (AE), designed for CT acquisition (A) of an image (B) using an imaging device (1), wherein the imaging device (1) is a CT device, - a processing unit (PU) spatially separated from the recording unit (AE), designed for processing (V) the image (B), - an analysis unit (AyE) designed for the analysis (Ay) of the image (B), for the evaluation (Bu) of the analysis (Ay) and for the calculation of an evaluation signal (BuS), wherein the analysis unit (AyE) is assigned to the acquisition unit (AE), and wherein the analysis (Ay) includes at least part of the processing (V), where the processing unit (PU) is designed to perform more computationally intensive image processing steps than the analysis unit (AyE), the system is designed to transfer a result (Er) of the analysis (Ay) from the analysis unit (AyE) to the processing unit (VE), wherein the analysis (Ay) includes the detection (E) of at least one physical structure, wherein the result (Er) is a success in the detection (E) of the at least one physical structure. [2] System according to claim 1, wherein the acquisition unit (AE) is designed to perform a series of CT scans (A), wherein the analysis (Ay) and evaluation (Bu) comprise at least the analysis (Ay) and evaluation (Bu) of a portion of the acquired images (B) of the series. [3] System according to claim 2, wherein the acquisition unit (AE) is designed to abort the CT acquisition (A) of an image series depending on the assessment signal (BuS). [4] System according to any one of claims 1 to 3, wherein the analysis (Ay) of the image (B) comprises a segmentation (S) of the image (B). [5] System according to any one of claims 1 to 4, wherein the assessment (B) comprises the calculation of at least one improved parameter for the CT acquisition (A) of an image (B). [6] System according to any one of claims 1 to 5, wherein the assessment (Bu) comprises an output (Au) of the assessment signal (BuS) in the form of a symbol and / or text by means of an output unit (11) associated with the recording unit (AE). [7] Methods for automated quality control of medical images (B), comprising the following steps: - CT scan (A) of an image (B) using a scanning unit (SU) of an imaging device (1), wherein the imaging device (1) is a CT scanner, - Processing (V) of the image (B) by means of a processing unit (VE) spatially separated from the recording unit (AE), - Analysis (Ay) of the image (B), evaluation (Bu) of the analysis (Ay) and calculation of an evaluation signal (BuS) using an analysis unit (AyE), wherein the analysis unit (AyE) is assigned to the acquisition unit (AuE), and wherein the analysis (Ay) includes at least part of the processing (V), where the processing unit (PU) is designed to perform more computationally intensive image processing steps than the analysis unit (AyE), - Transferring a result (Er) of the analysis (Ay) from the analysis unit (AyE) to the processing unit (VE), wherein the analysis (Ay) includes the recognition (E) of at least one physical structure, wherein the result (Er) is a success in the recognition (E) of the at least one physical structure. [8] The method of claim 7, comprising the following step: - CT scan (A) a series of images (B) using the acquisition unit (AE), wherein the analysis (Ay) and evaluation (Bu) include at least the analysis (Ay) and evaluation (Bu) of a portion of the acquired images (B) of the series. [9] The method of claim 8, comprising the following step: - Termination of the CT scan (A) of an image series depending on the assessment signal (BuS). [10] Method according to any one of claims 7 to 9, wherein the analysis (Ay) of the image (B) comprises a segmentation (S) of the image (B). [11] Method according to any one of claims 7 to 10, wherein the evaluation (Bu) comprises the calculation of at least one improved parameter for CT acquisition (Au) of an image (B). [12] Method according to any one of claims 7 to 11, wherein the assessment (Bu) comprises an output (Au) of the assessment signal (BuS) in the form of a symbol and / or text by means of an output unit (11) associated with the output unit (AE).

Images