Medical diagnosis and treatment nuclide split charging system and method based on activity correction

Abstract

The invention discloses a medical diagnosis and treatment nuclide split charging system and method based on activity correction. The system comprises: a stock solution dilution container; the nuclide sub-packaging device is used for sub-packaging the nuclide liquid medicine in the stock solution diluting container into a medicine dispensing container; the activity measuring device is used for collecting the activity measured value of the nuclide liquid medicine in the medicine dispensing container; the nuclide image acquisition module is used for acquiring a nuclide liquid medicine image in the medicine dispensing container; the nuclide activity correction coefficient measuring and calculating module is used for inputting the nuclide liquid medicine image into the trained activity correction coefficient measuring and calculating model and outputting a corresponding activity correction coefficient predicted value; wherein the activity correction coefficient measuring and calculating model is constructed based on a deep learning algorithm; and the nuclide activity correction module is used for correcting the activity measurement value of the corresponding nuclide liquid medicine according to the activity correction coefficient prediction value to obtain an activity true value of the nuclide liquid medicine in the medicine dispensing container. The accuracy, safety and efficiency of medical diagnosis and treatment nuclide split charging can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of medical diagnosis and treatment nuclide packaging, and in particular to a medical diagnosis and treatment nuclide packaging system and method based on activity correction. Background Art

[0002] Medical diagnostic nuclides refer to a series of radioactive isotopes that are widely used in the medical field for the diagnosis and treatment of diseases. These nuclides can be absorbed by specific tissues due to their unique radioactive energy to achieve the purpose of treatment or diagnosis. Different nuclides have different half-lives and radioactive energies, so they are suitable for different medical scenarios. For example, iodine-131 is often used to treat thyroid diseases, while phosphorus-32 is mostly used for radiotherapy of skin diseases.

[0003] Medical diagnostic radionuclides play an indispensable role in modern medicine. In terms of diagnosis, radionuclides can be used as tracers to help doctors observe and understand the functional status and pathological changes of organs in patients. In terms of treatment, by precisely controlling the dose and placement of radionuclides, diseased cells can be effectively killed or their growth can be inhibited, thereby achieving the purpose of treating tumors and other diseases.

[0004] Although the application of medical diagnostic radionuclides in the medical field has been quite mature, there are still certain challenges in the packaging process. At present, most medical institutions still use manual methods to dilute and package radionuclides. This method not only has the problems of backward technology and low efficiency, but also the packaging accuracy is difficult to guarantee. And due to the lack of effective protective measures, operators may face the risk of radiation injury. In addition, the activity of medical diagnostic radionuclides directly affects their effects in medical diagnosis and treatment. Accurate activity can ensure that radioactive drugs play the expected role in the patient's body. Although the radioactivity intensity is proportional to the detection distance, when the activity measurement device (detector) measures the activity of the radionuclide solution, it is affected by the volume of the drug distribution container and the shape of the drug solution (non-planar state), so that the activity measurement value has a certain deviation from the true value, and the activity measurement value needs to be corrected at this time. Therefore, how to design a medical diagnostic radionuclide packaging system based on activity correction is a technical problem that needs to be solved urgently. Summary of the invention

[0005] In view of the deficiencies of the above-mentioned prior art, the technical problem to be solved by the present invention is: how to provide a medical diagnostic radionuclide packaging system and method based on activity correction, so as to realize accurate, efficient and safe medical diagnostic radionuclide packaging through an automated packaging system; at the same time, the activity correction coefficient predicted value of the radionuclide solution in the dispenser is predicted through a deep learning algorithm, and then the activity measurement value is corrected, thereby improving the accuracy, safety and efficiency of medical diagnostic radionuclide packaging.

[0006] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0007] A medical diagnostic radionuclide packaging system based on activity correction, comprising:

[0008] The stock solution dilution container is used to contain the radionuclide solution;

[0009] A radionuclide dispensing device is used to dispense the radionuclide solution in the stock solution dilution container into a dispensing container;

[0010] Activity measuring device, used to collect activity measurement values ​​of radionuclide solution in the dispensing container;

[0011] A radionuclide image acquisition module is used to acquire images of radionuclide liquid medicine in the medicine dispensing container;

[0012] The nuclide activity correction coefficient calculation module is used to input the nuclide solution image into the trained activity correction coefficient calculation model and output the corresponding activity correction coefficient prediction value; wherein the activity correction coefficient calculation model is constructed based on the deep learning algorithm;

[0013] The nuclide activity correction module is used to correct the activity measurement value of the corresponding nuclide solution through the activity correction coefficient prediction value to obtain the true value of the activity of the nuclide solution in the drug distribution container.

[0014] Preferably, before the radionuclide dispensing device dispenses the radionuclide solution into the dispensing container, it first uses visual image technology to determine whether the dispensing container is correctly placed in the designated position.

[0015] Preferably, after the radionuclide image acquisition module acquires the radionuclide medicine solution image, the radionuclide medicine solution image is preprocessed to reduce the influence of the medicine dispensing container.

[0016] The present invention also discloses a method for packing medical diagnostic radionuclides based on activity correction, which is implemented based on a medical diagnostic radionuclides packing system and comprises:

[0017] S1: Dispense the radionuclide solution into the dispensing container through the radionuclide dispensing device;

[0018] S2: Collect the activity measurement value of the radionuclide solution in the drug distribution container through the activity measuring device;

[0019] S3: Collecting images of radionuclide liquid in the drug dispensing container;

[0020] S4: Inputting the radionuclide solution image into the trained activity correction coefficient calculation model, and outputting the corresponding activity correction coefficient prediction value; wherein the activity correction coefficient calculation model is constructed based on a deep learning algorithm;

[0021] S5: Correcting the activity measurement value of the corresponding nuclide solution by the predicted value of the activity correction coefficient to obtain the true value of the activity of the nuclide solution in the dispensing container;

[0022] S6: Repeat steps S1 to S5 until the actual value of the activity of the radionuclide solution in the dispensing container is consistent with the preset value.

[0023] Preferably, in step S1, before dispensing the radionuclide solution into the medicine dispensing container, visual image technology is used to determine whether the medicine dispensing container is correctly placed in the designated position.

[0024] Preferably, in step S1, the following steps are performed to determine whether the medicine dispensing container is correctly placed at the designated position:

[0025] S101: Acquire a scene image at a specified location;

[0026] S102: extracting image features from the scene image by using a computer vision algorithm;

[0027] S103: identifying a bounding box of a medicine dispensing container in the scene image based on the extracted image features using an object detection algorithm;

[0028] S104: Determine a designated position in the scene image;

[0029] S105: Calculate the Euclidean distance between the bounding box of the medicine dispensing container and the designated position; when the calculated Euclidean distance is less than a set threshold, determine that the medicine dispensing container is correctly placed at the designated position; otherwise, the medicine dispensing container is not correctly placed at the designated position.

[0030] Preferably, in step S3, after the radionuclide medicine solution image is collected, the radionuclide medicine solution image is first preprocessed to reduce the influence of the medicine dispensing container.

[0031] Preferably, in step S3, the specific steps of preprocessing the radionuclide solution image are as follows:

[0032] S301: Acquire the original radionuclide solution image, and convert the radionuclide solution image into a grayscale image by using a grayscale conversion algorithm;

[0033] S302: enhancing the contrast of the grayscale image by a histogram equalization method to obtain an enhanced grayscale image;

[0034] S303: determining a threshold value, and converting the enhanced grayscale image into a binary image through a thresholding operation;

[0035] S304: Perform morphological processing on the binary image;

[0036] S305: performing edge detection on the binary image after morphological processing by using an edge detection algorithm to identify the boundary between the radionuclide solution and the drug dispensing container;

[0037] S306: Smoothing the binary image after edge detection by using a filtering algorithm to obtain a pre-processed radionuclide solution image.

[0038] Preferably, in step S4, the activity correction coefficient calculation model is trained by the following steps:

[0039] S401: Obtaining a radionuclide solution image sample and a corresponding activity true value in a drug dispensing container;

[0040] S402: collecting activity measurement values ​​of the nuclide solution in the drug dispensing container through an activity measurement device; calculating the true value of the activity correction coefficient according to the activity measurement value of the nuclide solution and the true value of the activity; and marking the nuclide solution image sample according to the true value of the activity correction coefficient to form training data;

[0041] S403: constructing an activity correction coefficient calculation model based on a neural network model;

[0042] S404: inputting the training data into the activity correction coefficient calculation model, and outputting the corresponding activity correction coefficient prediction value;

[0043] S405: Calculating a loss function based on the activity correction coefficient prediction value and the corresponding activity correction coefficient true value, and reversely optimizing the model parameters of the activity correction coefficient calculation model;

[0044] S406: Repeat steps S404 to S405 until the activity correction coefficient calculation model converges.

[0045] Preferably, in step S5, the true value of the activity of the nuclide solution=the measured value of the activity of the nuclide solution*the predicted value of the activity correction coefficient.

[0046] Compared with the prior art, the medical diagnostic radionuclide packaging system and method based on activity correction in the present invention has the following beneficial effects:

[0047] The present invention realizes the automation of dilution, subpackaging and activity correction of medical diagnosis and treatment nuclides through a nuclide injection device, a nuclide dilution device, a nuclide subpackaging device, an activity measurement device, a nuclide image acquisition module, a nuclide activity correction coefficient calculation module and a nuclide activity correction module. First, the automation system can ensure the accuracy of the amount of nuclide liquid medicine in each subpackaging through precise mechanical control and sensor monitoring, avoiding errors and inconsistencies that may be caused by manual operation, thereby improving the accuracy of medical diagnosis and treatment nuclide subpackaging. Secondly, the automated medicine dispensing system can complete the medicine dispensing task continuously and quickly without manual intervention, thereby significantly improving the efficiency of the medicine dispensing operation and reducing waiting time. Finally, since nuclides are radioactive, operators may face the risk of radiation exposure during manual medicine dispensing, and the use of an automated system can reduce or avoid such direct contact, thereby improving the safety of medical diagnosis and treatment nuclide subpackaging.

[0048] In the process of automatic packaging of medical diagnosis and treatment nuclides, the present invention uses a deep learning algorithm to construct an activity correction coefficient measurement model to calculate the activity correction coefficient prediction value, and then corrects the activity measurement value of the corresponding nuclide liquid by the activity correction coefficient prediction value to obtain the true value of the activity of the nuclide liquid in the medicine dispensing container. First, the deep learning algorithm can learn the complex nonlinear relationship between the volume and solution morphology of the medicine dispensing container and the activity correction coefficient prediction value from a large amount of training data, so that the model can more accurately predict the activity correction coefficient, providing a reliable basis for subsequent activity correction. Secondly, by measuring and correcting the activity of the nuclide liquid in real time, the proportion or amount of the nuclide liquid can be dynamically adjusted according to the actual situation to ensure that the activity of the packaged nuclide liquid meets the preset requirements. This dynamic correction capability effectively improves the flexibility and adaptability of the medicine dispensing process. Then accurate prediction and correction of the activity correction coefficient can avoid the waste of nuclides caused by the activity not meeting the requirements, and can maximize the use of nuclide resources and reduce medical costs while ensuring the accuracy of the activity. Finally, the accuracy of radionuclide activity is crucial for medical diagnosis and treatment. Predicting and correcting the activity correction coefficient through deep learning algorithms can ensure that the radionuclide treatment dose is accurate and safe, thereby helping to improve the diagnosis and treatment effect and patient safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to make the purpose, technical solution and advantages of the invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings, in which:

[0050] Figure 1 This is the logic block diagram of the medical diagnostic radionuclide packaging system based on activity correction.

[0051] Figure 2 The figure is a logical block diagram of the method for packaging medical radionuclides based on activity correction. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but only represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work belong to the scope of protection of the present invention.

[0053] It should be noted that similar numbers and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. In the description of the present invention, it should be noted that the orientation or position relationship indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inside", "outside", etc. is based on the orientation or position relationship shown in the drawings, or the orientation or position relationship in which the invention product is usually placed when used, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present invention. In addition, the terms "first", "second", "third", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance. In addition, the terms "horizontal", "vertical", etc. do not mean that the components are required to be absolutely horizontal or suspended, but can be slightly tilted. For example, "horizontal" only means that its direction is more horizontal than "vertical", and does not mean that the structure must be completely horizontal, but can be slightly tilted. In the description of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "set", "install", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0054] The following is a further detailed description through specific implementation methods:

[0055] Embodiment 1:

[0056] This embodiment discloses a medical diagnostic radionuclide packaging system based on activity correction.

[0057] like Figure 1As shown, a medical diagnostic radionuclide packaging system based on activity correction includes:

[0058] The stock solution dilution container is used to contain the radionuclide solution;

[0059] A nuclide injection device, used for injecting nuclide stock solution into the stock solution container;

[0060] In this embodiment, the nuclide stock solution is injected into the stock solution dilution container through an electrically controlled needle.

[0061] A nuclide dilution device, used for injecting diluent into the original liquid dilution container;

[0062] In this embodiment, the diluent is injected into the original liquid dilution container by an electrically controlled needle, wherein the diluent is pure water.

[0063] A radionuclide dispensing device is used to dispense the radionuclide solution in the stock solution dilution container into a dispensing container;

[0064] Activity measuring device, used to collect activity measurement values ​​of radionuclide solution in the dispensing container;

[0065] A radionuclide image acquisition module is used to acquire images of radionuclide liquid medicine in the medicine dispensing container;

[0066] The nuclide activity correction coefficient calculation module is used to input the nuclide solution image into the trained activity correction coefficient calculation model and output the corresponding activity correction coefficient prediction value; wherein the activity correction coefficient calculation model is constructed based on the deep learning algorithm;

[0067] The nuclide activity correction module is used to correct the activity measurement value of the corresponding nuclide solution through the activity correction coefficient prediction value to obtain the true value of the activity of the nuclide solution in the drug distribution container.

[0068] The present invention realizes the automation of dilution, subpackaging and activity correction of medical diagnosis and treatment nuclides through a nuclide injection device, a nuclide dilution device, a nuclide subpackaging device, an activity measurement device, a nuclide image acquisition module, a nuclide activity correction coefficient calculation module and a nuclide activity correction module. First, the automation system can ensure the accuracy of the amount of nuclide liquid medicine in each subpackaging through precise mechanical control and sensor monitoring, avoiding errors and inconsistencies that may be caused by manual operation, thereby improving the accuracy of medical diagnosis and treatment nuclide subpackaging. Secondly, the automated medicine dispensing system can complete the medicine dispensing task continuously and quickly without manual intervention, thereby significantly improving the efficiency of the medicine dispensing operation and reducing waiting time. Finally, since nuclides are radioactive, operators may face the risk of radiation exposure during manual medicine dispensing, and the use of an automated system can reduce or avoid such direct contact, thereby improving the safety of medical diagnosis and treatment nuclide subpackaging.

[0069] In the process of automatic packaging of medical diagnosis and treatment nuclides, the present invention uses a deep learning algorithm to construct an activity correction coefficient measurement model to calculate the activity correction coefficient prediction value, and then corrects the activity measurement value of the corresponding nuclide liquid by the activity correction coefficient prediction value to obtain the true value of the activity of the nuclide liquid in the medicine dispensing container. First, the deep learning algorithm can learn the complex nonlinear relationship between the volume and solution morphology of the medicine dispensing container and the activity correction coefficient prediction value from a large amount of training data, so that the model can more accurately predict the activity correction coefficient, providing a reliable basis for subsequent activity correction. Secondly, by measuring and correcting the activity of the nuclide liquid in real time, the proportion or amount of the nuclide liquid can be dynamically adjusted according to the actual situation to ensure that the activity of the packaged nuclide liquid meets the preset requirements. This dynamic correction capability effectively improves the flexibility and adaptability of the medicine dispensing process. Then accurate prediction and correction of the activity correction coefficient can avoid the waste of nuclides caused by the activity not meeting the requirements, and can maximize the use of nuclide resources and reduce medical costs while ensuring the accuracy of the activity. Finally, the accuracy of radionuclide activity is crucial for medical diagnosis and treatment. Predicting and correcting the activity correction coefficient through deep learning algorithms can ensure that the radionuclide treatment dose is accurate and safe, thereby helping to improve the diagnosis and treatment effect and patient safety.

[0070] During the specific implementation process, before the radionuclide dispensing device dispenses the radionuclide solution into the dispensing container, it first uses visual image technology to determine whether the dispensing container is correctly placed in the designated position.

[0071] Specifically, the following steps are used to determine whether the medicine dispensing container is correctly placed in the designated position:

[0072] 1) Obtain a scene image at a specified location;

[0073] In this embodiment, the scene image needs to be processed by grayscale conversion, noise reduction, contrast enhancement, etc. to improve the image quality and feature recognizability.

[0074] 2) Extract image features from scene images through computer vision algorithms, including edge features, corner features, and texture features;

[0075] In this embodiment, features can be extracted through computer vision algorithms such as Canny edge detection, SIFT (Scale Invariant Feature Transform), SURF (Speeded Up Robust Features), etc.

[0076] 3) Identify the bounding box of the medicine dispensing container in the scene image based on the extracted image features through an object detection algorithm (such as a deep learning-based object detection model, such as YOLO, SSD, etc.);

[0077] In this embodiment, the target detection algorithm can output a bounding box of the medicine dispensing container in the scene image.

[0078] 4) determining a specified position in the scene image;

[0079] In this embodiment, the designated position may be determined by presetting an area or coordinate point in the image or by other sensor data.

[0080] 5) Calculate the Euclidean distance between the bounding box (center point or centroid) of the medicine dispensing container and the specified position; when the calculated Euclidean distance is less than the set threshold, it is determined that the medicine dispensing container is correctly placed at the specified position; otherwise, the medicine dispensing container is not correctly placed at the specified position.

[0081] The present invention uses visual image technology to capture the position information of the medicine dispensing container in real time, and compares it with the preset designated position. When the medicine dispensing container is correctly placed in the designated position, the system can accurately identify and continue to perform the packaging operation; if the position is incorrect, the system can issue a warning or stop the operation in time, thereby avoiding packaging errors, and significantly improving the accuracy of nuclide packaging through high-precision position recognition capabilities. At the same time, by judging the position of the medicine dispensing container through visual image technology, it can be ensured that the container is correctly placed before packaging, effectively avoiding safety hazards such as drug leakage, contamination or personal injury caused by container misalignment. And by judging the position of the medicine dispensing container through visual image technology, the time for manual inspection and confirmation can be shortened, making the packaging process smoother and more efficient.

[0082] In the specific implementation process, after the radionuclide image acquisition module acquires the radionuclide solution image, the radionuclide solution image is preprocessed to reduce the influence of the drug distribution container (glass) on the radionuclide solution image.

[0083] Specifically, the specific steps of radionuclide solution image preprocessing are as follows:

[0084] 1) Obtaining the original radionuclide solution image, and converting the radionuclide solution image into a grayscale image through a grayscale conversion algorithm;

[0085] In this embodiment, the radionuclide liquid image can be converted into a grayscale image using a weighted average method or a grayscale conversion function of an image processing library. In a grayscale image, the value of each pixel represents brightness information, and the range is usually 0 to 255.

[0086] 2) Analyze the histogram of the grayscale image, determine the distribution of pixel values, enhance the contrast of the grayscale image through the histogram equalization method (or contrast stretching technology), and obtain an enhanced grayscale image;

[0087] In this embodiment, histogram equalization expands the contrast range of an image by redistributing pixel values, while contrast stretching enhances the contrast of an image by linear or nonlinear transformation.

[0088] 3) Determine a threshold value and convert the enhanced grayscale image into a binary image through a thresholding operation;

[0089] In this embodiment, the threshold value can be determined by the grayscale histogram of the image or the Otsu threshold method. Pixel values ​​above the threshold value are set to white (or 1), and those below the threshold value are set to black (or 0).

[0090] 4) Perform morphological processing on binary images, including corrosion, expansion, opening or closing operations;

[0091] In this embodiment, erosion refers to removing small pixels at the edge of the image, which helps to remove noise. Dilation refers to making the white area in the image larger, which helps to fill the holes inside the characters. Opening (erosion followed by dilation) refers to removing small white areas (noise) while keeping the shape of the characters unchanged. Closing (dilation followed by erosion) refers to filling small black areas (holes) while keeping the separation between characters.

[0092] 5) Perform edge detection on the binary image after morphological processing by edge detection algorithm (such as Canny edge detection) to identify the boundary between the radionuclide solution and the glass in the dispensing container;

[0093] 6) Smoothing the binary image after edge detection by filtering (such as Gaussian filtering) to reduce the reflection and blurring effects caused by the glass, and obtaining the pre-processed radionuclide solution image.

[0094] The present invention can reduce the influence of the glass of the medicine dispensing container on the image of the nuclide medicine liquid by preprocessing the image of the nuclide medicine liquid. First, the interference factors such as reflection and refraction generated by the glass container can be effectively reduced by preprocessing, so that the image of the nuclide medicine liquid is clearer, and the preprocessing process can correct the blurring effect of the glass container on the image, thereby improving the resolution of the image and making the detail information in the image more prominent. Secondly, the quality of the preprocessed image as the input data of the deep learning algorithm has been significantly improved, which helps the deep learning algorithm to more accurately capture and identify the characteristic information of the nuclide medicine liquid image, and because the influence of the glass container on the image is effectively reduced, the error of the deep learning algorithm in predicting the activity correction coefficient will also be reduced accordingly, improving the accuracy of the prediction. Finally, the preprocessing technology can be adjusted according to different lighting conditions, shooting angles and other factors, so that the deep learning algorithm can better adapt to various complex shooting environments, and through preprocessing, the deep learning algorithm can learn more general image features, thereby showing a stronger generalization ability when predicting the activity correction coefficient of different nuclide medicine liquids.

[0095] Embodiment 2:

[0096] This embodiment discloses a method for packaging medical radionuclides based on activity correction, which is implemented based on the medical radionuclides packaging system in the first embodiment.

[0097] like Figure 2 As shown, the method for packaging medical radionuclides based on activity correction includes:

[0098] S1: Dispense the radionuclide solution into the dispensing container through the radionuclide dispensing device;

[0099] S2: Collect the activity measurement value of the radionuclide solution in the drug distribution container through the activity measuring device;

[0100] S3: Collecting images of radionuclide liquid in the drug dispensing container;

[0101] S4: Inputting the radionuclide solution image into the trained activity correction coefficient calculation model, and outputting the corresponding activity correction coefficient prediction value; wherein the activity correction coefficient calculation model is constructed based on a deep learning algorithm;

[0102] S5: Correcting the activity measurement value of the corresponding nuclide solution by the predicted value of the activity correction coefficient to obtain the true value of the activity of the nuclide solution in the dispensing container;

[0103] In this embodiment, the actual value of the activity of the nuclide solution=the measured value of the activity of the nuclide solution*the predicted value of the activity correction coefficient.

[0104] S6: Repeat steps S1 to S5 until the actual value of the activity of the radionuclide solution in the dispensing container is consistent with the preset value.

[0105] In the process of automatic packaging of medical diagnosis and treatment nuclides, the present invention uses a deep learning algorithm to construct an activity correction coefficient measurement model to calculate the activity correction coefficient prediction value, and then corrects the activity measurement value of the corresponding nuclide liquid by the activity correction coefficient prediction value to obtain the true value of the activity of the nuclide liquid in the medicine dispensing container. First, the deep learning algorithm can learn the complex nonlinear relationship between the volume and solution morphology of the medicine dispensing container and the activity correction coefficient prediction value from a large amount of training data, so that the model can more accurately predict the activity correction coefficient, providing a reliable basis for subsequent activity correction. Secondly, by measuring and correcting the activity of the nuclide liquid in real time, the proportion or amount of the nuclide liquid can be dynamically adjusted according to the actual situation to ensure that the activity of the packaged nuclide liquid meets the preset requirements. This dynamic correction capability effectively improves the flexibility and adaptability of the medicine dispensing process. Then accurate prediction and correction of the activity correction coefficient can avoid the waste of nuclides caused by the activity not meeting the requirements, and can maximize the use of nuclide resources and reduce medical costs while ensuring the accuracy of the activity. Finally, the accuracy of radionuclide activity is crucial for medical diagnosis and treatment. Predicting and correcting the activity correction coefficient through deep learning algorithms can ensure that the radionuclide treatment dose is accurate and safe, thereby helping to improve the diagnosis and treatment effect and patient safety.

[0106] During the specific implementation process, after the radionuclide solution image is collected, the radionuclide solution image is first preprocessed to reduce the influence of the drug distribution container; and then the preprocessed radionuclide solution image is input into the trained activity correction coefficient calculation model.

[0107] In the specific implementation process, before the radionuclide solution is dispensed into the dispensing container, visual image technology is first used to determine whether the dispensing container is correctly placed in the designated position.

[0108] Specifically, the following steps are used to determine whether the medicine dispensing container is correctly placed in the designated position:

[0109] S101: Acquire a scene image at a specified location;

[0110] In this embodiment, the scene image needs to be processed by grayscale conversion, noise reduction, contrast enhancement, etc. to improve the image quality and feature recognizability.

[0111] S102: extracting image features in the scene image through a computer vision algorithm, including edge features, corner features, and texture features;

[0112] In this embodiment, features can be extracted through computer vision algorithms such as Canny edge detection, SIFT (Scale Invariant Feature Transform), SURF (Speeded Up Robust Features), etc.

[0113] S103: identifying a bounding box of the medicine dispensing container in the scene image based on the extracted image features by using an object detection algorithm (such as an object detection model based on deep learning, such as YOLO, SSD, etc.);

[0114] In this embodiment, the target detection algorithm can output a bounding box of the medicine dispensing container in the scene image.

[0115] S104: Determine a designated position in the scene image;

[0116] In this embodiment, the designated position may be determined by presetting an area or coordinate point in the image or by other sensor data.

[0117] S105: Calculate the Euclidean distance between the bounding box (center point or centroid) of the medicine dispensing container and the specified position; when the calculated Euclidean distance is less than the set threshold, it is determined that the medicine dispensing container is correctly placed at the specified position; otherwise, the medicine dispensing container is not correctly placed at the specified position.

[0118] The present invention uses visual image technology to capture the position information of the medicine dispensing container in real time, and compares it with the preset designated position. When the medicine dispensing container is correctly placed in the designated position, the system can accurately identify and continue to perform the packaging operation; if the position is incorrect, the system can issue a warning or stop the operation in time, thereby avoiding packaging errors, and significantly improving the accuracy of nuclide packaging through high-precision position recognition capabilities. At the same time, by judging the position of the medicine dispensing container through visual image technology, it can be ensured that the container is correctly placed before packaging, effectively avoiding safety hazards such as drug leakage, contamination or personal injury caused by container misalignment. And by judging the position of the medicine dispensing container through visual image technology, the time for manual inspection and confirmation can be shortened, making the packaging process smoother and more efficient.

[0119] During the specific implementation process, after the radionuclide solution image is collected, the radionuclide solution image is first preprocessed to reduce the influence of the drug distribution container.

[0120] Specifically, the specific steps of radionuclide solution image preprocessing are as follows:

[0121] S301: Acquire the original radionuclide solution image, and convert the radionuclide solution image into a grayscale image by using a grayscale conversion algorithm;

[0122] In this embodiment, the radionuclide liquid image can be converted into a grayscale image using a weighted average method or a grayscale conversion function of an image processing library. In a grayscale image, the value of each pixel represents brightness information, and the range is usually 0 to 255.

[0123] S302: Analyze the histogram of the grayscale image to determine the distribution of pixel values, and enhance the contrast of the grayscale image by a histogram equalization method (or contrast stretching technology) to obtain an enhanced grayscale image;

[0124] In this embodiment, histogram equalization expands the contrast range of an image by redistributing pixel values, while contrast stretching enhances the contrast of an image by linear or nonlinear transformation.

[0125] S303: determining a threshold value, and converting the enhanced grayscale image into a binary image through a thresholding operation;

[0126] In this embodiment, the threshold value can be determined by the grayscale histogram of the image or the Otsu threshold method. Pixel values ​​above the threshold value are set to white (or 1), and those below the threshold value are set to black (or 0).

[0127] S304: performing morphological processing on the binary image, including corrosion, expansion, opening operation or closing operation;

[0128] In this embodiment, erosion refers to removing small pixels at the edge of the image, which helps to remove noise. Dilation refers to making the white area in the image larger, which helps to fill the holes inside the characters. Opening (erosion followed by dilation) refers to removing small white areas (noise) while keeping the shape of the characters unchanged. Closing (dilation followed by erosion) refers to filling small black areas (holes) while keeping the separation between characters.

[0129] S305: performing edge detection on the binary image after morphological processing by using an edge detection algorithm (such as Canny edge detection) to identify the boundary between the radionuclide solution and the glass in the drug dispensing container;

[0130] S306: Smoothing the binary image after edge detection by filtering (such as Gaussian filtering) to reduce the reflection and blurring effects caused by the glass, and obtaining a pre-processed radionuclide solution image.

[0131] The present invention can reduce the influence of the glass of the medicine dispensing container on the image of the nuclide medicine liquid by preprocessing the image of the nuclide medicine liquid. First, the interference factors such as reflection and refraction generated by the glass container can be effectively reduced by preprocessing, so that the image of the nuclide medicine liquid is clearer, and the preprocessing process can correct the blurring effect of the glass container on the image, thereby improving the resolution of the image and making the detail information in the image more prominent. Secondly, the quality of the preprocessed image as the input data of the deep learning algorithm has been significantly improved, which helps the deep learning algorithm to more accurately capture and identify the characteristic information of the nuclide medicine liquid image, and because the influence of the glass container on the image is effectively reduced, the error of the deep learning algorithm in predicting the activity correction coefficient will also be reduced accordingly, improving the accuracy of the prediction. Finally, the preprocessing technology can be adjusted according to different lighting conditions, shooting angles and other factors, so that the deep learning algorithm can better adapt to various complex shooting environments, and through preprocessing, the deep learning algorithm can learn more general image features, thereby showing a stronger generalization ability when predicting the activity correction coefficient of different nuclide medicine liquids.

[0132] During the specific implementation process, the activity correction coefficient calculation model is trained through the following steps:

[0133] S401: Obtaining a radionuclide solution image sample and a corresponding activity true value in a drug dispensing container;

[0134] In this embodiment, the radionuclide liquid image sample needs to be preprocessed. The true value of the activity is obtained by calibration or calculation.

[0135] S402: collecting activity measurement values ​​of the nuclide solution in the drug dispensing container through an activity measurement device; calculating the true value of the activity correction coefficient according to the activity measurement value of the nuclide solution and the true value of the activity; and marking the nuclide solution image sample according to the true value of the activity correction coefficient to form training data;

[0136] In this embodiment, the true value of the activity correction coefficient = the true value of the activity / the measured value of the activity of the nuclide solution.

[0137] S403: constructing an activity correction coefficient calculation model based on a neural network model;

[0138] In this embodiment, the number of input layer nodes is first determined, which is determined according to the number of pixels or features of the input image. Then the hidden layer is designed, and the appropriate number of hidden layer nodes and activation function are selected. Then the number of output layer nodes is determined. For the prediction of radionuclide liquid activity, the output layer only needs one node to represent the predicted activity value. Finally, the weights and biases of the neural network are initialized, which is completed by random initialization or other optimization methods.

[0139] S404: inputting the training data into the activity correction coefficient calculation model, and outputting the corresponding activity correction coefficient prediction value;

[0140] In this embodiment, the activity correction coefficient calculation model (neural network model) can predict the activity by extracting features such as the volume of the drug distribution container and the shape of the drug solution in the radionuclide drug solution image sample.

[0141] S405: Calculating a loss function based on the activity correction coefficient prediction value and the corresponding activity correction coefficient true value, and reversely optimizing the model parameters of the activity correction coefficient calculation model;

[0142] In this embodiment, the loss function adopts the cross entropy loss function.

[0143] S406: Repeat steps S404 to S405 until the activity correction coefficient calculation model converges.

[0144] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit the technical solution. Those skilled in the art should understand that those modifications or equivalent substitutions of the technical solution of the present invention that do not depart from the purpose and scope of the technical solution should be included in the scope of the claims of the present invention.

Claims

1. A medical diagnostic radionuclide packaging system based on activity correction, characterized in that: include: Stock solution dilution container, used to contain radionuclide solution; A radionuclide dispensing device is used to dispense the radionuclide solution in the stock solution dilution container into a dispensing container; Activity measuring device, used to collect activity measurement values ​​of radionuclide solution in the dispensing container; A radionuclide image acquisition module is used to acquire images of radionuclide liquid medicine in the medicine dispensing container; The nuclide activity correction coefficient calculation module is used to input the nuclide solution image into the trained activity correction coefficient calculation model and output the corresponding activity correction coefficient prediction value; The activity correction coefficient calculation model is built based on a deep learning algorithm; The nuclide activity correction module is used to correct the activity measurement value of the corresponding nuclide solution through the activity correction coefficient prediction value to obtain the true value of the activity of the nuclide solution in the drug distribution container.

2. The medical diagnostic radionuclide packaging system based on activity correction according to claim 1, characterized in that: Before the radionuclide dispensing device dispenses the radionuclide solution into the dispensing container, it first uses visual image technology to determine whether the dispensing container is correctly placed at the designated position.

3. The medical diagnostic radionuclide dispensing system based on activity correction according to claim 1, characterized in that: After the radionuclide image acquisition module acquires the radionuclide solution image, it pre-processes the radionuclide solution image to reduce the influence of the drug distribution container.

4. A method for packaging medical radionuclides based on activity correction, characterized in that: The medical diagnosis and treatment nuclide packaging system according to claim 1 is implemented, comprising: S1: Dispense the radionuclide solution into the dispensing container through the radionuclide dispensing device; S2: Collect the activity measurement value of the radionuclide solution in the drug distribution container through the activity measuring device; S3: Collecting images of radionuclide liquid in the drug dispensing container; S4: Inputting the radionuclide solution image into the trained activity correction coefficient calculation model, and outputting the corresponding activity correction coefficient prediction value; wherein the activity correction coefficient calculation model is constructed based on a deep learning algorithm; S5: Correcting the activity measurement value of the corresponding nuclide solution by the predicted value of the activity correction coefficient to obtain the true value of the activity of the nuclide solution in the dispensing container; S6: Repeat steps S1 to S5 until the actual value of the activity of the radionuclide solution in the dispensing container is consistent with the preset value.

5. The method for packaging medical diagnostic radionuclides based on activity correction according to claim 4, characterized in that: In step S1, before dispensing the radionuclide solution into the medicine dispensing container, it is first determined by visual image technology whether the medicine dispensing container is correctly placed at the designated position.

6. The method for packaging medical diagnostic radionuclides based on activity correction according to claim 5, characterized in that: In step S1, the following steps are used to determine whether the medicine dispensing container is correctly placed at the designated location: S101: Acquire a scene image at a specified location; S102: extracting image features from the scene image by using a computer vision algorithm; S103: identifying a bounding box of a medicine dispensing container in the scene image based on the extracted image features using an object detection algorithm; S104: Determine a designated position in the scene image; S105: Calculate the Euclidean distance between the bounding box of the medicine dispensing container and the designated position; when the calculated Euclidean distance is less than a set threshold, determine that the medicine dispensing container is correctly placed at the designated position; otherwise, the medicine dispensing container is not correctly placed at the designated position.

7. The method for packaging medical diagnostic radionuclides based on activity correction according to claim 4, characterized in that: In step S3, after the radionuclide liquid medicine image is collected, the radionuclide liquid medicine image is preprocessed to reduce the influence of the medicine dispensing container.

8. The method for packaging medical diagnostic radionuclides based on activity correction according to claim 7, characterized in that: In step S3, the specific steps of radionuclide solution image preprocessing are as follows: S301: Acquire the original radionuclide solution image, and convert the radionuclide solution image into a grayscale image by using a grayscale conversion algorithm; S302: enhancing the contrast of the grayscale image by a histogram equalization method to obtain an enhanced grayscale image; S303: determining a threshold value, and converting the enhanced grayscale image into a binary image through a thresholding operation; S304: Perform morphological processing on the binary image; S305: performing edge detection on the binary image after morphological processing by using an edge detection algorithm to identify the boundary between the radionuclide solution and the drug dispensing container; S306: Smoothing the binary image after edge detection by using a filtering algorithm to obtain a pre-processed radionuclide solution image.

9. The method for packaging medical diagnostic radionuclides based on activity correction according to claim 4, characterized in that: In step S4, the activity correction coefficient calculation model is trained by the following steps: S401: Obtaining a radionuclide solution image sample and a corresponding activity true value in a drug dispensing container; S402: collecting activity measurement values ​​of the radionuclide solution in the drug dispensing container through an activity measurement device; Calculate the true value of the activity correction coefficient according to the activity measurement value and the true value of the activity of the radionuclide solution; annotate the radionuclide solution image sample according to the true value of the activity correction coefficient to form training data; S403: constructing an activity correction coefficient calculation model based on a neural network model; S404: inputting the training data into the activity correction coefficient calculation model, and outputting the corresponding activity correction coefficient prediction value; S405: Calculating a loss function based on the activity correction coefficient prediction value and the corresponding activity correction coefficient true value, and reversely optimizing the model parameters of the activity correction coefficient calculation model; S406: Repeat steps S404 to S405 until the activity correction coefficient calculation model converges.

10. The method for packaging medical diagnostic radionuclides based on activity correction according to claim 4, characterized in that: In step S5, the actual value of the activity of the nuclide solution = the measured value of the activity of the nuclide solution * the predicted value of the activity correction coefficient.

Images