Methods, devices and equipment for assessing the quality of medical document images

Abstract

This invention discloses a method, apparatus, and device for evaluating the quality of medical document images, relating to the field of medical technology. The method includes: acquiring a medical document image to be evaluated; detecting text boxes containing text content within the image; inputting the text boxes into a text fuzzy prediction model to obtain a fuzzy prediction result for the text boxes; and determining the quality of the medical document image based on the fuzzy prediction result. This method can accurately evaluate the quality of medical documents that are partially clear and partially blurred.

Description

Technical Field

[0001] This invention relates to the field of medical technology, and in particular to a method, apparatus, and device for evaluating the quality of medical document images. Background Technology

[0002] In some business scenarios, it is necessary to automatically identify information from medical document images, such as medical insurance review and prescription compliance review. However, the quality of uploaded medical document images is too poor, that is, the text in the medical document images is too blurry, and the text content cannot be accurately extracted from them. Therefore, it is necessary to conduct quality assessment on medical document images.

[0003] Currently, the methods for assessing the quality of medical document images calculate the blur ratio globally using a full-image approach. When a medical document image is partially clear and partially blurry, the current methods cannot accurately assess the quality of the medical document. Summary of the Invention

[0004] In view of this, the present invention provides a method, apparatus and device for evaluating the quality of medical document images, which can accurately evaluate the quality of medical documents that are partially clear and partially blurred.

[0005] According to a first aspect of the present invention, a method for assessing the quality of images in medical documents is provided, the method comprising:

[0006] Obtain an image of the medical document to be evaluated, and detect the text box containing the text content in the image of the medical document to be evaluated;

[0007] The text box is input into the text fuzzy prediction model to obtain the fuzzy prediction result of the text box;

[0008] The quality of the medical document image to be evaluated is determined based on the fuzzy prediction results.

[0009] Preferably, acquiring the image of the medical document to be evaluated includes:

[0010] Obtain initial images of the medical documents to be evaluated;

[0011] The initial medical document image to be evaluated is represented by hue, saturation, and brightness. The contrast of the initial medical document image to be evaluated is increased to a preset contrast to obtain the medical document image to be evaluated.

[0012] Preferably, the fuzzy prediction result includes whether the text box is blurred or not, and determining the quality of the medical document image to be evaluated based on the fuzzy prediction result includes:

[0013] Based on the fuzzy prediction results, determine the fuzzy text boxes of all the text boxes, and calculate the fuzziness ratio of the medical document image to be evaluated based on the fuzzy text boxes and all the text boxes;

[0014] The quality of the medical document image to be evaluated is obtained based on the blur ratio and the preset blur level.

[0015] Preferably, before inputting the text box into the text fuzzy prediction model, the method further includes:

[0016] Obtain the first initial text fuzzy prediction model built based on a convolutional neural network;

[0017] Train the first initial text fuzzy prediction model until the loss function converges to obtain the trained text fuzzy prediction model.

[0018] Preferably, before inputting the text box into the text fuzzy prediction model, the method further includes:

[0019] Obtain a second initial text fuzzy prediction model, wherein the second initial text fuzzy prediction model includes an encoding module and a decoding module, the encoding module is constructed based on a convolutional neural network, and the decoding module is constructed based on a long short-term memory network with an attention mechanism;

[0020] Train the second initial text fuzzy prediction model until the loss function converges to obtain the trained text fuzzy prediction model.

[0021] Preferably, the method further includes:

[0022] If the quality of the medical document image to be evaluated meets the threshold for manual review, the medical document image to be evaluated is sent to the manual review module for manual review to determine whether to send a re-upload instruction to the medical document upload terminal.

[0023] Preferably, determining the quality of the medical document image to be evaluated based on the fuzzy prediction result includes:

[0024] Detect the initial text box containing the initial text content in the initial medical document image to be evaluated;

[0025] The initial text box is input into the text fuzzy prediction model to obtain the initial fuzzy prediction result of the initial text box, wherein the initial fuzzy prediction result is whether the initial text box is fuzzy or not;

[0026] Calculate the initial blurred text box of all the initial text boxes based on the initial blurred prediction results, and calculate the missing ink text box based on the initial blurred text box and the blurred text box;

[0027] The quality of the medical document image to be evaluated is determined based on the missing ink text box and the fuzzy prediction result.

[0028] According to a second aspect of the present invention, a medical document image quality assessment device is provided, the device comprising:

[0029] The detection module is used to acquire images of medical documents to be evaluated and to detect the text boxes containing the text content in the images of the medical documents to be evaluated.

[0030] The prediction module is used to input the text box into the text fuzzy prediction model to obtain the fuzzy prediction result of the text box;

[0031] A determination module is used to determine the quality of the medical document image to be evaluated based on the fuzzy prediction result.

[0032] According to a third aspect of this application, a storage medium is provided that stores a computer program thereon, which, when executed by a processor, implements the above-described medical document image quality assessment method.

[0033] According to a fourth aspect of this application, a computer device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor executes the program to implement the above-described medical document image quality assessment method.

[0034] By employing the above technical solution, this invention provides a method, apparatus, and device for quality assessment of medical document images. This invention first acquires a medical document image to be assessed and detects the text boxes containing the text content within the image. Then, the text boxes are input into a text fuzzy prediction model to obtain the fuzzy prediction results. Finally, the quality of the medical document image to be assessed is determined based on the fuzzy prediction results. Through this invention's technical solution, the medical document image to be assessed is segmented into individual text boxes, each in image format and containing text content. The quality assessment of the medical document image is thus transformed into a quality assessment of all text boxes. Therefore, the text fuzzy prediction model inputs all text boxes, rather than the entire medical document image. This avoids the influence of background information on the quality assessment. Furthermore, when the medical document image is partially clear and partially blurred, inputting all text boxes into the text fuzzy prediction model for quality assessment is more accurate than inputting the entire medical document image. Therefore, this application improves the accuracy of quality assessment for medical document images with partially clear and partially blurred images.

[0035] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0036] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of this application. In the drawings:

[0037] Figure 1 A flowchart illustrating a method for assessing the quality of medical document images provided in an embodiment of the present invention is shown.

[0038] Figure 2 A flowchart illustrating another method for assessing the quality of medical document images provided in an embodiment of the present invention is shown.

[0039] Figure 3 This diagram illustrates the structure of a medical document image quality assessment device according to an embodiment of the present invention.

[0040] Figure 4 This invention provides a schematic diagram of the structure of another medical document image quality assessment device according to an embodiment of the present invention.

[0041] Figure 5 A structural diagram of the second initial text fuzzy prediction model provided in an embodiment of the present invention is shown. Detailed Implementation

[0042] The present invention will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0043] This embodiment provides a method for assessing the quality of images in medical documents, such as... Figure 1 As shown, the method includes:

[0044] 101. Obtain the image of the medical document to be evaluated, and detect the text box containing the text content in the image of the medical document to be evaluated.

[0045] In this embodiment, the document images can be based on scanned images, photographed images, or video images obtained by scanning, photographing, or recording documents.

[0046] In specific application scenarios, the text information included in the medical document images to be evaluated can be medical insurance material information. When reviewing whether medical insurance material information is eligible for reimbursement, it is necessary to first assess whether the document image is clear. If it is blurry, a re-upload is requested, thereby improving review efficiency. It can also be doctor's prescription information. When reviewing doctor's prescription information for incompatibilities, etc., it is also necessary to first assess whether the document image is clear. If it is blurry, a re-upload is requested, thereby improving review efficiency.

[0047] The detection involves identifying text boxes containing the text content within the medical document image to be evaluated, where a text box refers to the area containing the text content in the medical document image. This detection can be performed using a Faster-RCNN network, and is not limited to this method.

[0048] 102. Input the text box into the text fuzzy prediction model to obtain the fuzzy prediction result of the text box.

[0049] In this embodiment, the fuzzy prediction result can be the fuzzy probability of the text content in the text box, thereby obtaining the fuzzy probability of the Chinese text content in each text box through a text fuzzy prediction model. The text fuzzy prediction model can be a convolutional neural network.

[0050] 103. Determine the quality of the medical document image to be evaluated based on the fuzzy prediction results.

[0051] In this embodiment, as one implementation method, the quality of the medical document image to be evaluated can be determined based on the minimum, maximum, and average values ​​of the blur probability of the text content in all text boxes.

[0052] Specifically, one method is to compare the minimum and maximum values ​​of the blur probability with a preset blur threshold. If the maximum value of the blur probability is less than the preset blur threshold, the image of the medical document to be evaluated has passed the review and does not need to be re-uploaded. If the minimum value of the blur probability is greater than the preset blur threshold, the image of the medical document to be evaluated has failed the review and needs to be re-uploaded. If neither the maximum value of the blur probability nor the minimum value of the blur probability is greater than the preset blur threshold is met, then the average blur probability of the text content in all text boxes is calculated, and this average value is compared with the preset blur threshold. If the average value is greater than the preset blur threshold, the image of the medical document to be evaluated has failed the review and needs to be re-uploaded. If the average value is less than or equal to the preset blur threshold, the image of the medical document to be evaluated has passed the review and does not need to be re-uploaded. Here, blur probability refers to the probability of blurriness, and the preset blur threshold is a threshold value. Values ​​greater than the threshold value indicate blurriness, and values ​​less than or equal to the threshold value indicate clarity.

[0053] Specifically, another method of determination is to directly compare the average value with the preset fuzzy threshold instead of comparing the maximum value, minimum value and preset fuzzy threshold. The specific comparison method is the same as the above embodiment, and will not be repeated here.

[0054] In this embodiment, as another implementation, the fuzzy prediction result can be a binary classification result, including whether the text box is blurred or not. Determining the quality of the medical document image to be evaluated based on the fuzzy prediction result includes: determining the blurred text boxes of all the text boxes based on the fuzzy prediction result; calculating the blur ratio of the medical document image to be evaluated based on the blurred text boxes and all the text boxes; and obtaining the quality of the medical document image to be evaluated based on the blur ratio and a preset blur level.

[0055] If a text box is blurred, it is called a blurred text box. For blurred text boxes that are determined according to the blurred prediction results, the blur ratio of the medical document image to be evaluated is calculated based on the blurred text boxes and all the text boxes. This includes: determining the number of blurred text boxes in the blurred prediction results as blurred text boxes, and then calculating the proportion of blurred text boxes to text boxes to obtain the blur ratio of the medical document image to be evaluated.

[0056] Here, the preset blur level refers to a preset blur ratio range corresponding to a preset blur level (for example, a preset blur ratio range of [0, 0.25] corresponds to a preset blur level of 1, a preset blur ratio range of (0.25, 0.5] corresponds to a preset blur level of 2, a preset blur ratio range of (0.5, 0.75] corresponds to a preset blur level of 3, and a preset blur ratio range of (0.75, 0.1] corresponds to a preset blur level of 4). To obtain the quality of the medical document image to be evaluated based on the blur ratio and preset blur level, the process includes: determining the target preset blur ratio range (0.6 within [0.5, 0.75]) of the medical document image to be evaluated, and then, based on the correspondence, obtaining the target preset blur level (a preset blur ratio range of (0.5, 0.75] corresponds to a preset blur level of 3). Therefore, the quality of the medical document image to be evaluated can be its target preset blur level (level 3).

[0057] This invention provides a method, apparatus, and device for quality assessment of medical document images. The invention first acquires a medical document image to be assessed and detects the text boxes containing the text content within the image. Then, the text boxes are input into a text fuzzy prediction model to obtain fuzzy prediction results. Finally, the quality of the medical document image is determined based on the fuzzy prediction results. Through this invention, the medical document image is segmented into individual text boxes, each in image format and containing text content. The quality assessment of the medical document image is thus transformed into an assessment of all text boxes. Therefore, the text fuzzy prediction model inputs all text boxes, rather than the entire medical document image. This avoids the influence of background information on the quality assessment. Furthermore, when the medical document image is partially clear and partially blurred, inputting all text boxes into the text fuzzy prediction model for quality assessment is more accurate than inputting the entire medical document image. Therefore, this application improves the accuracy of quality assessment for medical document images with partially clear and partially blurred images.

[0058] Furthermore, as a refinement and extension of the specific implementation methods described above, and to fully illustrate the specific implementation process in this embodiment, another method for assessing the quality of medical document images is provided, such as... Figure 2 As shown, the method includes:

[0059] 201. Obtain the initial medical document image to be evaluated, represent the initial medical document image to be evaluated according to hue, saturation and brightness, increase the contrast of the initial medical document image to be evaluated to a preset contrast, obtain the medical document image to be evaluated, and detect the text box where the text content in the medical document image to be evaluated is located.

[0060] The main implementation component of this solution is a medical document image quality assessment system. This system first acquires initial images of medical documents uploaded by the user. Since these initial images may have faded or insufficient ink, directly performing text box detection on them would result in missed text boxes, leading to inaccurate quality assessment results. Therefore, after acquiring the initial images, ink enhancement processing is applied. Then, text box detection is performed on the enhanced images. (Ink enhancement processing involves representing the initial image based on hue, saturation, and brightness, and increasing the contrast of the initial image to a preset contrast level.)

[0061] Text box detection can be performed using the Faster-RCNN network, and no specific method is specified here.

[0062] After completing step 201 of the embodiment, the text box (wherein the text box is in image form) can be input into the trained text fuzzy prediction model to determine whether the text content in the text box is fuzzy or the degree of fuzziness, and then perform quality assessment based on the judgment result. It should be noted that there is no need to identify the semantics of the text content. Before inputting the text box into the trained text fuzzy prediction model, it is also necessary to obtain the constructed initial text fuzzy prediction model and train it. This initial text fuzzy prediction model is divided into the first initial text fuzzy prediction model in step 202a of the embodiment, which is based on a convolutional neural network, and the second initial text fuzzy prediction model in step 202b of the embodiment, which is based on a convolutional neural network, a long short-term memory network, and an attention architecture.

[0063] 202a. Obtain the first initial text fuzzy prediction model based on the convolutional neural network, train the first initial text fuzzy prediction model until the loss function converges, and obtain the trained text fuzzy prediction model.

[0064] 202b. Obtain a second initial text fuzzy prediction model, wherein the second initial text fuzzy prediction model includes an encoding module and a decoding module. The encoding module is constructed based on a convolutional neural network, and the decoding module is constructed based on a long short-term memory network with an attention mechanism. Train the second initial text fuzzy prediction model until the loss function converges to obtain a trained text fuzzy prediction model.

[0065] For step 202a of the embodiment, the first initial text fuzzy prediction model built based on Convolutional Neural Networks (CNN) includes an input layer, multiple convolutional layers, pooling layers, fully connected layers, and an output layer. The input layer feeds the image into the CNN. The convolutional layers extract local features by sliding small filters (also called convolutional kernels) across the input layer. The pooling layers downsample the output of the convolutional layers, reducing the amount of data and lowering the dimensionality. The fully connected layers map the outputs of the previous layers to the output space. The output layer outputs the prediction result. To prevent overfitting due to an excessively deep network structure, a shallower network structure is preferred. For example, there are five convolutional layers, with the first layer being: ker... The first layer has a kernel of 7*7, a stride of 1, a kernel weight of 4×1×7×7, and a kernel bias of 4; the second layer has a kernel of 3*3, a stride of 1, a kernel weight of 8×4×4×3, and a kernel bias of 8; the third layer has a kernel of 4*3, a stride of 2*2, a kernel weight of 16×8×3×3, and a kernel bias of 16; the fourth layer has a kernel of 4*3, a stride of 2*2, a kernel weight of 32×16×4×3, and a kernel bias of 32; the fifth layer has a kernel of 8*3, a stride of 1, a kernel weight of 1×32×8×3, and a kernel bias of 1.

[0066] For step 202b of the embodiment, the second initial text fuzzy prediction model includes an encoding module and a decoding module. The encoding module is built based on a convolutional neural network, and the decoding module is built based on a long short-term memory network with an attention mechanism, such as... Figure 5As shown, the encoding module includes an input layer and a convolutional layer (which performs convolution, pooling, and node unrolling, i.e., dimensionality reduction operations, in sequence). The convolutional layer is used to extract local features. The decoding module includes a three-layer Long Short-Term Memory (LSTM) network. The output of the previous LSTM layer is the input of the next LSTM layer, and so on down the chain. The convolutional layers input these local features into the LSTM for sequence modeling. The LSTM can capture long-term dependencies in time-series data and generate hidden states to represent the sequence information of the data. Then, the last LSTM layer inputs the sequence information into the attention layer. The attention mechanism of the attention layer allocates more attention to the key parts of the sequence information that affect the output result, so as to better learn the information in the sequence information. After passing through the fully connected layer, the learned "distributed feature representation" is mapped to the sample label space, and then the output layer obtains the prediction result.

[0067] Taking step 202a of the embodiment as an example, the training of the first initial text fuzzy prediction model is carried out (the training process of step 202b of the embodiment is the same as that of step 202a of the embodiment, and will not be repeated here). The label document image (the label document image carries its true fuzziness) is used as the input of the first initial text fuzzy prediction model. The first initial text fuzzy prediction model calculates the predicted fuzziness probability of the label document image. The loss value of the loss function is calculated based on the true fuzziness and the predicted fuzziness probability. If the loss function does not converge, the parameters of the first initial text fuzzy prediction model are adjusted until the loss function converges. At this time, the trained text fuzzy prediction model is obtained.

[0068] 203. Input the text box into the text fuzzy prediction model to obtain the fuzzy prediction result of the text box.

[0069] Among them, the fuzzy prediction result is whether the text box is blurred or not.

[0070] 204. Detect the initial text box containing the initial text content in the initial medical document image to be evaluated, input the initial text box into the text fuzzy prediction model, and obtain the initial fuzzy prediction result of the initial text box, wherein the initial fuzzy prediction result is whether the initial text box is fuzzy or not.

[0071] 205. Calculate the initial blurred text boxes of all the initial text boxes based on the initial blurred prediction results, calculate the missing ink text boxes based on the initial blurred text boxes and the blurred text boxes, and determine the quality of the medical document image to be evaluated based on the missing ink text boxes and the blurred prediction results.

[0072] For step 204 of the embodiment, the specific implementation method is the same as that of steps 201-203 of the above embodiment, and will not be repeated here. The only difference is that in steps 201-203 of the above embodiment, the medical document image to be evaluated is after ink enhancement processing, while in step 204 of the embodiment, the initial medical document image to be evaluated is not after ink enhancement processing.

[0073] For step 205 of the embodiment, since the initial medical document image to be evaluated has not undergone ink enhancement processing, while the medical document image to be evaluated has undergone ink enhancement processing, as an implementation method, calculating the missing ink text boxes based on the initial blurred text boxes and the blurred text boxes, and determining the quality of the medical document image to be evaluated based on the missing ink text boxes and the fuzzy prediction result, includes: calculating the difference between the number of initial medical document images to be evaluated and the total number of medical document images to be evaluated, obtaining the number of missing ink text boxes, calculating a first proportion of the number of missing ink text boxes to the initial number of medical document images to be evaluated, calculating a second proportion of the blurred text boxes in the fuzzy prediction result to all text boxes, and obtaining the quality of the medical document image to be evaluated by multiplying the first proportion weight by the first proportion plus the second proportion weight by the second proportion. Thus, the impact of missing ink on quality evaluation is considered, improving the accuracy of quality evaluation.

[0074] 206. If the quality of the medical document image to be evaluated meets the threshold for manual review, the medical document image to be evaluated is sent to the manual review module for manual review to determine whether to send a re-upload instruction to the medical document upload terminal.

[0075] As one implementation method, after obtaining the quality of the medical document image to be evaluated in step 205 of the embodiment, the next step is to determine whether to send a re-upload instruction to the medical document upload terminal based on the quality of the medical document image to be evaluated. However, the quality of the medical document image to be evaluated is obtained using a text fuzzy prediction model, which is not completely accurate. In order to improve the user experience of the uploader, if the quality of the medical document image to be evaluated meets the manual review threshold (the quality of the medical document image to be evaluated is so bad that it needs to be manually reviewed, and the degree of badness is the manual review threshold), then the uploader is not directly asked to re-upload, but rather a manual review is conducted to confirm whether or not a re-upload is required.

[0076] This invention provides a method, apparatus, and device for quality assessment of medical document images. The invention first acquires a medical document image to be assessed and detects the text boxes containing the text content within the image. Then, the text boxes are input into a text fuzzy prediction model to obtain fuzzy prediction results. Finally, the quality of the medical document image is determined based on the fuzzy prediction results. Through this invention, the medical document image is segmented into individual text boxes, each in image format and containing text content. The quality assessment of the medical document image is thus transformed into an assessment of all text boxes. Therefore, the text fuzzy prediction model inputs all text boxes, rather than the entire medical document image. This avoids the influence of background information on the quality assessment. Furthermore, when the medical document image is partially clear and partially blurred, inputting all text boxes into the text fuzzy prediction model for quality assessment is more accurate than inputting the entire medical document image. Therefore, this application improves the accuracy of quality assessment for medical document images with partially clear and partially blurred images.

[0077] Furthermore, as Figure 1 and Figure 2 The specific implementation of the method shown in this invention provides a medical document image quality assessment device, such as... Figure 3 As shown, the device includes: a detection module 31, a prediction module 32, and a determination module 33;

[0078] Detection module 31 is used to acquire the image of the medical document to be evaluated and detect the text box containing the text content in the image of the medical document to be evaluated.

[0079] The prediction module 32 is used to input the text box into the text fuzzy prediction model to obtain the fuzzy prediction result of the text box;

[0080] The determination module 33 is used to determine the quality of the medical document image to be evaluated based on the fuzzy prediction result.

[0081] Accordingly, in order to obtain the image of the medical document to be evaluated, the detection module 31 can be used to obtain an initial image of the medical document to be evaluated; according to the hue, saturation and brightness of the initial image of the medical document to be evaluated, the contrast of the initial image of the medical document to be evaluated is increased to a preset contrast, thereby obtaining the image of the medical document to be evaluated.

[0082] Accordingly, the fuzzy prediction result includes whether the text box is blurred or not. In order to determine the quality of the medical document image to be evaluated based on the fuzzy prediction result, the determining module 33 can be used to determine the blurred text boxes of all the text boxes based on the fuzzy prediction result, calculate the blur ratio of the medical document image to be evaluated based on the blurred text boxes and all the text boxes, and obtain the quality of the medical document image to be evaluated based on the blur ratio and the preset blur level.

[0083] In specific application scenarios, such as Figure 4 As shown, a medical document image quality assessment device further includes: a first training module 34, specifically used to obtain a first initial text fuzzy prediction model constructed based on a convolutional neural network; training the first initial text fuzzy prediction model until the loss function converges to obtain a trained text fuzzy prediction model.

[0084] In specific application scenarios, such as Figure 4 As shown, a medical document image quality assessment device further includes: a second training module 35, which can be used to obtain a second initial text fuzzy prediction model, wherein the second initial text fuzzy prediction model includes an encoding module and a decoding module, the encoding module is constructed based on a convolutional neural network, and the decoding module is constructed based on a long short-term memory network with an attention mechanism; the second initial text fuzzy prediction model is trained until the loss function converges to obtain a trained text fuzzy prediction model.

[0085] In specific application scenarios, such as Figure 4 As shown, a medical document image quality assessment device further includes a manual review module 36, which is specifically used to send the medical document image to be assessed to the manual review module if the quality of the image meets the manual review threshold, so as to manually review the image and determine whether to send a re-upload instruction to the medical document upload terminal.

[0086] Accordingly, in order to determine the quality of the medical document image to be evaluated based on the fuzzy prediction result, the determining module 33 can also be used to detect the initial text box containing the initial text content in the initial medical document image to be evaluated; input the initial text box into the text fuzzy prediction model to obtain the initial fuzzy prediction result of the initial text box, wherein the initial fuzzy prediction result is whether the initial text box is fuzzy or not; calculate the initial fuzzy text box of all the initial text boxes based on the initial fuzzy prediction result; calculate the missing ink text box based on the initial fuzzy text box and the fuzzy text box; and determine the quality of the medical document image to be evaluated based on the missing ink text box and the fuzzy prediction result.

[0087] It should be noted that other corresponding descriptions of the functional units involved in the medical document image quality assessment device provided in this embodiment can be found in [reference needed]. Figures 1 to 2 The corresponding description will not be repeated here.

[0088] Based on the above, Figures 1 to 2 Accordingly, this embodiment also provides a storage medium, which may be volatile or non-volatile, storing a computer program that, when executed by a processor, implements the above-described method. Figures 1 to 2 The method for assessing the quality of images in medical documents is shown.

[0089] Based on this understanding, the technical solution of the present invention can be embodied in the form of a software product, which can be stored in a storage medium (such as a CD-ROM, USB flash drive, portable hard drive, etc.) and includes several instructions to cause a computer device (such as a personal computer, server, or network device, etc.) to execute the methods of various implementation scenarios of the present invention.

[0090] Based on the above, Figures 1 to 2 The method shown and Figure 3 , Figure 4 To achieve the above objectives, this embodiment also provides a computer device, which includes a storage medium and a processor; the storage medium is used to store a computer program; the processor is used to execute the computer program to implement the above-described virtual device embodiment. Figures 1 to 2 The method for assessing the quality of images in medical documents is shown.

[0091] Optionally, the computer device may also include a user interface, a network interface, a camera, radio frequency (RF) circuitry, sensors, audio circuitry, a Wi-Fi module, etc. The user interface may include a display screen, input units such as a keyboard, etc., and optional user interfaces may also include USB interfaces, card reader interfaces, etc. The network interface may optionally include standard wired interfaces, wireless interfaces (such as Wi-Fi interfaces), etc.

[0092] Those skilled in the art will understand that the computer device structure provided in this embodiment does not constitute a limitation on the physical device, and may include more or fewer components, or combine certain components, or have different component arrangements.

[0093] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the aforementioned computer device, supporting the operation of information processing programs and other software and / or programs. The network communication module is used for communication between the various components within the storage medium, as well as for communication with other hardware and software in the information processing device.

[0094] Through the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be implemented by means of software plus necessary general-purpose hardware platform, or it can be implemented by hardware.

[0095] This invention provides a method, apparatus, and device for quality assessment of medical document images. The invention first acquires a medical document image to be assessed and detects the text boxes containing the text content within the image. Then, the text boxes are input into a text fuzzy prediction model to obtain fuzzy prediction results. Finally, the quality of the medical document image is determined based on the fuzzy prediction results. Through this invention, the medical document image is segmented into individual text boxes, each in image format and containing text content. The quality assessment of the medical document image is thus transformed into an assessment of all text boxes. Therefore, the text fuzzy prediction model inputs all text boxes, rather than the entire medical document image. This avoids the influence of background information on the quality assessment. Furthermore, when the medical document image is partially clear and partially blurred, inputting all text boxes into the text fuzzy prediction model for quality assessment is more accurate than inputting the entire medical document image. Therefore, this application improves the accuracy of quality assessment for medical document images with partially clear and partially blurred images.

[0096] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of a preferred embodiment, and the modules or processes shown in the drawings are not necessarily essential for implementing the present invention. Those skilled in the art will understand that the modules in the apparatus of the embodiment can be distributed within the apparatus of the embodiment as described, or they can be located in one or more apparatuses different from this embodiment, with corresponding changes. The modules of the above-described embodiment can be combined into one module, or further divided into multiple sub-modules.

[0097] The serial numbers used above are for descriptive purposes only and do not represent the superiority or inferiority of the implementation scenarios. The above disclosures are merely a few specific implementation scenarios of the present invention; however, the present invention is not limited thereto, and any variations conceived by those skilled in the art should fall within the protection scope of the present invention.

Claims

1. A method for assessing the image quality of medical documents, characterized in that, The method includes: Obtain an image of a medical document to be evaluated, and detect the text box containing the text content in the image of the medical document to be evaluated. The image of the medical document to be evaluated is obtained by performing ink enhancement processing on an initial image of the medical document to be evaluated. The text box is input into the text fuzzy prediction model to obtain the fuzzy prediction result of the text box. The fuzzy prediction result includes whether the text box is fuzzy or not. If the text box is fuzzy, it is determined to be a fuzzy text box. The quality of the medical document image to be evaluated is determined based on the fuzzy prediction results. Determining the quality of the medical document image to be evaluated based on the fuzzy prediction result includes: Detect the initial text box containing the initial text content in the initial medical document image to be evaluated; The initial text box is input into the text fuzzy prediction model to obtain the initial fuzzy prediction result of the initial text box, wherein the initial fuzzy prediction result is whether the initial text box is fuzzy or not; Calculate the initial blurred text box of all the initial text boxes based on the initial blurred prediction results, and calculate the missing ink text box based on the initial blurred text box and the blurred text box; The quality of the medical document image to be evaluated is determined based on the missing ink text box and the fuzzy prediction result. The step of calculating missing ink text boxes based on the initial blurred text boxes and the blurred text boxes, and determining the quality of the medical document image to be evaluated based on the missing ink text boxes and the blurred prediction result, includes: calculating the difference between the number of initial blurred text boxes and the number of blurred text boxes to obtain the number of missing ink text boxes; calculating a first proportion of the number of missing ink text boxes to the total number of initial text boxes; calculating a second proportion of the number of blurred text boxes to the total number of text boxes; and obtaining the quality of the medical document image to be evaluated by multiplying the first proportion weight by the first proportion plus the second proportion weight by the second proportion.

2. The method according to claim 1, characterized in that, The acquisition of images of the medical documents to be evaluated includes: Obtain initial images of the medical documents to be evaluated; The initial medical document image to be evaluated is represented by hue, saturation, and brightness. The contrast of the initial medical document image to be evaluated is increased to a preset contrast to obtain the medical document image to be evaluated.

3. The method according to claim 1, characterized in that, Determining the quality of the medical document image to be evaluated based on the fuzzy prediction result includes: Based on the fuzzy prediction results, determine the fuzzy text boxes of all the text boxes, and calculate the fuzziness ratio of the medical document image to be evaluated based on the fuzzy text boxes and all the text boxes; The quality of the medical document image to be evaluated is obtained based on the blur ratio and the preset blur level.

4. The method according to claim 1, characterized in that, Before inputting the text box into the text fuzzy prediction model, the method further includes: Obtain the first initial text fuzzy prediction model built based on a convolutional neural network; Train the first initial text fuzzy prediction model until the loss function converges to obtain the trained text fuzzy prediction model.

5. The method according to claim 1, characterized in that, Before inputting the text box into the text fuzzy prediction model, the method further includes: Obtain a second initial text fuzzy prediction model, wherein the second initial text fuzzy prediction model includes an encoding module and a decoding module, the encoding module is constructed based on a convolutional neural network, and the decoding module is constructed based on a long short-term memory network with an attention mechanism; Train the second initial text fuzzy prediction model until the loss function converges to obtain the trained text fuzzy prediction model.

6. The method according to claim 1 or 3, characterized in that, The method further includes: If the quality of the medical document image to be evaluated meets the threshold for manual review, the medical document image to be evaluated is sent to the manual review module for manual review to determine whether to send a re-upload instruction to the medical document upload terminal.

7. A medical document image quality assessment device, characterized in that, The device includes: The detection module is used to acquire an image of a medical document to be evaluated and to detect the text box containing the text content in the image of the medical document to be evaluated. The image of the medical document to be evaluated is obtained by performing ink enhancement processing on an initial image of the medical document to be evaluated. The prediction module is used to input the text box into the text fuzzy prediction model to obtain the fuzzy prediction result of the text box, wherein the fuzzy prediction result includes whether the text box is fuzzy or not; if the text box is fuzzy, it is determined to be a fuzzy text box. A determination module is used to determine the quality of the medical document image to be evaluated based on the fuzzy prediction result; The determining module is used to detect the initial text box containing the initial text content in the initial medical document image to be evaluated; input the initial text box into the text fuzzy prediction model to obtain the initial fuzzy prediction result of the initial text box, wherein the initial fuzzy prediction result indicates whether the initial text box is blurred or not blurred; calculate the initial blurred text box of all the initial text boxes based on the initial fuzzy prediction result; calculate the missing ink text box based on the initial blurred text box and the blurred text box; and determine the quality of the medical document image to be evaluated based on the missing ink text box and the fuzzy prediction result. The determining module is used to calculate the difference between the initial number of blurred text boxes and the number of blurred text boxes to obtain the number of text boxes with missing ink, calculate the first ratio of the number of text boxes with missing ink to the total number of initial text boxes, calculate the second ratio of the number of blurred text boxes to the total number of text boxes, and obtain the quality of the medical document image to be evaluated by multiplying the first ratio by a preset first ratio weight and adding the second ratio by a preset second ratio weight.

8. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the medical document image quality assessment method according to any one of claims 1 to 6.

9. A computer device, comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, When the processor executes the computer program, it implements the medical document image quality assessment method according to any one of claims 1 to 6.

Images