TWI705459B - De-identification method and system thereof, method of generating templet data - Google Patents

Abstract

A de-identification method executed by a processor, the method includes: receiving a sample image data including sample image, the sample image includes text; recognizing the text of the sample image using OCR; choosing one of the recognized text which match a predetermined condition; generating de-identification template data which indicates de-identification area corresponds to the text matching the predetermined condition; saving the generated de-identification template data; receiving a target image data including target image; choosing one of a plurality of saved de-identification template data; de-identifying contents of the target image corresponds to the de-identification area of the chosen de-identification template data; and generating an output image data having the de-identified target image.

Description

De-identification method and system, method for generating template data

The present invention relates to a method for de-identifying images, in particular to a method for automatically creating templates to de-identify images.

In many industries, specific data has the opportunity to be shared from the original institution to another institution for use in tasks such as task outsourcing, academic research, etc. Therefore, there is an opportunity to expose the contents of the data that need to be kept secret to unauthorized organizations. For example, the Health Insurance Portability and Accountability Act (HIPAA) requires that the patient’s personal information not be disclosed, and medical image data usually contains the patient’s personal data (including name, ID number, Medical record numbers, etc.), so academic institutions will obtain the patient’s personal data when conducting research based on medical imaging data received from medical institutions, so that personal data that needs to be kept confidential is exposed to unauthorized academic institutions. For this reason, medical institutions must de-identify medical imaging data before sharing data with academic institutions. In some cases, the way to de-identify medical images is to identify and mask manually, which wastes manpower and is not efficient.

The present disclosure provides a de-identification method, including: a processing unit receives sample data with a sample image, the sample image has characters; the processing unit uses optical character recognition (Optical Character Recognition, OCR) to identify the data The text in the sample image; the processing unit selects and recognizes the text in the sample image which meets a predetermined condition The processing unit generates template data, the template data instructs to go out the recognition area, the de-recognition area corresponds to the text that meets the predetermined conditions, and stores the generated template data; the processing unit receives The target image data of the target image; the processing unit selects one of a plurality of stored template data; the processing unit removes the content of the de-identified region in the target image corresponding to the selected template data Identifying; and the processing unit generates output image data, the output image data having the de-identified target image.

The present disclosure also provides a method for generating template data, including: a processing unit receives sample data with a sample image, the sample image has text; the processing unit uses optical character recognition (OCR) to identify The text in the sample image; the processing unit selects and recognizes text from the text in the sample image that meets a predetermined condition; and the processing unit generates template data, the template data instructing to identify The de-identified area corresponds to a character that meets the predetermined condition.

The present disclosure provides a de-identification system including a storage module and a processing unit. The storage module stores a plurality of template data. The processing unit is in data connection with the storage module. The processing unit is configured to receive sample data with a sample image, the sample image has text; the processing unit is configured to use optical character recognition (Optical Character Recognition, OCR) to identify all of the sample images The text; the processing unit is configured to select the text recognized from the text in the sample image which meets a predetermined condition; the processing unit is configured to generate template data, the template data indicating the identification area , The de-identified area corresponds to the text that meets the predetermined condition, and the generated template data is stored in the storage module; the processing unit is configured to receive the target image data with the target image; the processing The unit is configured to select one of the plurality of template data stored in the storage module; the processing unit is configured to remove the content of the de-identified area corresponding to the selected template data in the target image Identification; and the processing unit is configured to generate output image data, the output image data having the de-identified target image.

As mentioned above, the processing unit uses the sample data to generate template data, and uses the template data to de-identify the target image and generate an output image data, which can effectively improve efficiency.

110: De-identification system

111: Communication module

112: storage module

113: Translation Module

114: Processing Unit

121: Equipment

122: Equipment

123: Equipment

130: communication network

D, D’: template data

T, T’: preset label

S21~S24: Program

S51~S55: Program

S521~S522: Program

310: Target image

311: text

410: De-identified target image

411: Mask

610, 710, 810: sample images

611: Image Information

612, 612’, 712’, 812’: text

913, 1003, 1103: reference area

914, 1004, 1104: comparison area

1201: template image

1202: De-identification area

S131~S139: Program

In order to understand the features described above in this case carefully, a more specific description of the above case can be provided with reference to the implementation aspects, and some implementation aspects are illustrated in the accompanying drawings. However, it should be noted that the accompanying drawings only illustrate the typical implementation aspects of this case and are therefore not regarded as limiting the scope of this case, because this case may recognize other equivalent implementation aspects.

Figure 1 shows a component block diagram and an exemplary operating environment of a de-identification system according to some embodiments of the present disclosure; Figure 2 shows a de-identification implemented by some exemplary de-identification systems according to the present disclosure Figure 3 shows a target image according to some embodiments of the present disclosure; Figure 4 shows a de-identified target image according to some embodiments of the present disclosure; Figure 5 shows A flowchart of a method for generating a de-identification template implemented by some exemplary de-identification systems according to the present disclosure; Fig. 6 to Fig. 8 respectively show three sample images according to some embodiments of the present disclosure; Figures 9 to 11 respectively show three sample images with de-identified regions defined according to some embodiments of the present disclosure; Figure 12 shows a de-identified region according to some embodiments of the present disclosure ; And Figure 13 shows a flowchart of a de-identification method implemented by some exemplary de-identification systems according to the present disclosure.

The following description will refer to the accompanying drawings to more fully describe the present disclosure. Shown in the attached picture This is an exemplary embodiment of the present disclosure. However, the present disclosure can be implemented in many different forms, and should not be construed as being limited to the exemplary embodiments set forth herein. These exemplary embodiments are provided to make the present disclosure thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art. Similar reference numerals indicate the same or similar elements.

The terms used herein are only for the purpose of describing specific exemplary embodiments, and are not intended to limit the present disclosure. As used herein, unless the context clearly dictates otherwise, the singular forms "a", "an" and "said" are intended to also include the plural forms. In addition, when used herein, "includes" and/or "includes" or "includes" and/or "includes" or "has" and/or "has", integers, steps, operations, elements and/or components , But does not exclude the presence or addition of one or more other features, regions, integers, steps, operations, elements, components and/or groups thereof.

Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those of ordinary skill in the art to which this disclosure belongs. In addition, unless clearly defined in the context, terms such as those defined in a general dictionary should be interpreted as having meanings consistent with their meanings in the related art and the present disclosure, and will not be interpreted as idealized or overly formal meanings.

The following content will describe exemplary embodiments with reference to the accompanying drawings. It should be noted that the elements depicted in the reference drawings are not necessarily shown to scale; and the same or similar elements will be given the same or similar reference numerals or similar technical terms.

FIG. 1 shows an exemplary operation scenario of the de-identification system 110 according to some embodiments of the present disclosure, which includes a block diagram of the components of the de-identification system 110 and data connection with the de-identification system 110的 device 121, device 122, device 123. In other operating environments, the number of devices connected to the de-identification system 110 is not limited to the content disclosed in FIG. 1.

In the exemplary operation scenario, the equipment 121, the equipment 122, and the equipment 123 belong to the image generating devices of the A hospital, the B hospital, and the C hospital, respectively. Device 121 is a tomographic device Prepared. In the exemplary operation scenario, the device 121, the device 122, and the device 123 are combined to generate image data with a file format definition, such as the image data conforming to the Digital Imaging and Medical Transmission Protocol (DICOM). Communications in Medicine) common standard image data. In the exemplary operation scenario, the device 121, the device 122, and the device 123 are respectively configured to generate image data conforming to the DICOM standard, such as computer tomography, computer X-ray photography, and echocardiography.

DICOM is a standard that regulates how to process, store, print, and transmit medical image data. The DICOM standard includes file format definitions and communication protocols. Image data conforming to the DICOM standard must be transmitted through a communication network that supports the TCP/IP protocol. Image data conforming to the DICOM standard has format definitions such as pixel data and image attribute information. Pixel data describes the value of each pixel to form an image. The attribute information has a plurality of tags and a plurality of attribute values respectively corresponding to the tags. The tag contains the group number and the element number, such as the tag (0010, 0010), where 0010 is the group number, and 0010 is the element number ). Each tag uniquely corresponds to an attribute value. Each attribute value is used to describe a type of information, the information can be the patient's name (such as "Huang Xiaoming"), patient ID, medical institution (such as "National Taiwan University Hospital"), instrument manufacturer (such as "Philips"), equipment model (such as "EPIQ 5"), image type (such as "US"), image size (such as "640 x 480"), image type (such as "Secondary"), etc. For example, the attribute value corresponding to the tag "(0010,0010)" is "Huang Xiaoming", which is used to describe the patient's name.

In this embodiment, the de-identification system 110 includes a storage module 112, a translation module 113, and a processing unit 114.

The storage module 112 may include one or more storage devices, and is configured to store a plurality of template data D, D'and a plurality of preset tags T, T'. In this embodiment, each template data D (D’) has at least one de-identification tag, at least one template attribute value, and a template with a specific format. The format of the board image and at least one de-identified area on the template image. In this embodiment, each de-identified area is located at a constant position of the template image. For example, the template data D'indicates a de-identified area (for example, the de-identified area 1202) and a de-identified label (for example, "(0010,0010)"), and has three template attribute values: "( ONIS, ONIS25)” (device model) and “430” (resolution in the length direction of the image), and “600” (resolution in the width direction of the image).

Fig. 12 exemplarily shows a template image 1201 and a de-identification area 1202 of the template data D'in some embodiments.

To give another simpler example, each template attribute value contained in template data D describes one of the following information: 1. Medical institution (such as National Taiwan University Hospital), 2. Instrument manufacturer (such as Philips), 3. Equipment model ( Such as EPIQ 5), 4. Image type (such as US), 5. Image size (such as 640 x 480), 6. Image type (such as Secondary). The de-identified area indicated by the template data D is a plurality of masked positions in the image.

When it is determined that the attribute value of the medical image is the same as the attribute values 1 to 6 of the template data D, the de-identified area indicated by the template data D can be directly used to de-identify the medical image.

In some cases, not all characters in medical images need to be de-identified. For example, a medical image has text corresponding to the name of the patient and the name of the imaging equipment manufacturer. Only the text corresponding to the patient’s name needs to be de-identified, while the text corresponding to the name of the imaging equipment manufacturer does not need to be de-identified.化. Therefore, when the template data D, D'are generated, the characters that need to be de-identified can be selected. The preset tags T, T'correspond to the characters that need to be de-identified, and the preset tags T, T'can be used in the method of generating template data described later. For example, the preset label T'is (0010, 0010), which corresponds to the patient's name. In other cases, the medical image may have text corresponding to the patient's name and corresponding to the patient's body temperature.

The translation module 113 may be an electronic device including necessary hardware, software, or firmware. Modules, such as servers. The translation module 133 is configured to translate text into a predetermined language. In this embodiment, the predetermined language is, for example, English, but not limited to this. In this embodiment, the text includes letters, words, numbers and other forms.

The processing unit 114, the storage module 112, and the translation module 113 are data connected. The processing unit 114 may be an electronic module including one or more hardware, software, or firmware, such as a server. These servers can adopt a centralized configuration or a distributed cluster arrangement. In other embodiments, the processing unit 114 may be a single computer or a high-speed computing computer. The processing unit 114 can generate template data D (D') from image data conforming to the DICOM standard received from the devices 121, 122, 123 by a communication module (not shown), and can use the template data D (D ') to de-identify the DICOM-compliant image data from the devices 121, 122, and 123 received through the communication module.

The communication module may be an electronic module including necessary hardware, software, or firmware, and is configured to establish a data connection with the devices 121, 122, and 123 through a suitable communication network. The suitable communication network may include at least one of wired and wireless media. The communication module can receive data from the devices 121, 122, 123 via a communication network. In this embodiment, the communication module supports the TCP/IP protocol, and can receive image data conforming to the DICOM standard from the device 121, the device 122, and the device 123. In other implementations, the communication module may also be omitted, and the image data generated by the device 121, the device 122, and the device 123 may be stored in the storage module 112 via other paths for the processing unit 114 to read.

Figure 2 shows a flowchart of a de-identification method implemented by some exemplary de-identification systems according to the present disclosure.

In step S21, the processing unit 114 receives a target image data with a file format definition from the device 121 through the communication module 111. In this embodiment, the target image data is, for example, image data conforming to the DICOM standard. The target image data includes attributes Information and to one or more target images, the attribute information includes a plurality of tags and a plurality of attribute values respectively corresponding to the tags. In this embodiment, in the target image data, the attribute value is at least related to the model of a device that generates the target image and the image size of the one or more target images. Generally speaking, the target The attribute value of the image data is also related to the personal information of the patient.

To facilitate understanding, an exemplary usage scenario is applied below to assist in explaining the subsequent procedures S22 to S24. In the usage scenario, the target image data has only one target image (for example, the target image 310 shown in Figure 3), and its attribute information has three attribute values and three tags, one attribute value "(ONIS,ONIS25) "Indicates the model of the device that generated the target image data, and corresponds to the label "(0008,1090)"; another attribute value "430" indicates the width resolution of the target image, and corresponds to Label "0028,0010"; another attribute value "600" indicates the resolution in length of the target image, and corresponds to the label "0028,0011"; another attribute "CHEN CHI WEN" indicates the The name of the patient and corresponds to the label "(0010,0010)".

FIG. 3 shows an exemplary target image 310 according to the present disclosure. The target image 310 has text 311 and other texts, and the text 311 is "CHEN CHI WEN".

In the procedure S22, the processing unit 114 selects one of the template data D, D'stored in the storage module 112 according to the attribute value of the target image data (as shown in FIG. 1). Specifically, the template attribute value of the template data D (D') selected by the processing unit 114 matches the attribute value of the target image data. Following the aforementioned use case, the processing unit 114 compares the attribute values "430", "600", and "(ONIS,ONIS25)" of the target image data to the three template attribute values of the template data D'( "430", "600", "(ONIS, ONIS25)"), that is, it is determined that the template attribute value of the template data D'matches the attribute value of the target image data, and then the template data D'is selected. In this way, the format of the target image in the target image data matches the format of the template image of the template data D', so the de-identification area 1202 indicated by the template data D'selected by the processing unit 114 is substantially The above corresponds to the text 311 of the target image 310. In some implementation aspects, Some template data D will also record that there is no data in some image pixels.

Take another example that is different from the previous usage. When the template attribute value of template data D contains the following information: 1. Medical institution (such as National Taiwan University Hospital), 2. Instrument manufacturer (such as Philips), 3. Device model (such as EPIQ) 5), 4. Image category (such as US), 5. Image size (such as 640 x 480), 6. Image type (such as Secondary). Then, when the processing unit 114 determines that the attribute value of the target image data is the same as the template attribute values such as 1 to 6 of the template data D in S22, it will select the template data D.

As in step S23, the processing unit 114 uses the selected template data D(D') to de-identify the content of the target image corresponding to the selected template data D(D') De-identification. In this embodiment, the processing unit 114 is configured to de-identify the content in the de-identified area by masking the content or part of the de-identified area. In other embodiments, the processing unit 114 may also be configured to blur the content or part of the de-identified area, remove the content or part of the de-identified area, and replace the de-identified area with other content. The content or part of the content in the de-identification area is de-identified. In some cases, the processing unit 114 will mask the information in the de-identified area in black or the same color as the surrounding color of the image, or process it in a mosaic manner.

In the target image data, in addition to the target image, its attribute values also reveal sensitive patient information. Therefore, in this embodiment, in the procedure S23, in addition to de-identifying the target image, the processing unit 114 also de-identifying at least one of the attribute values of the target image data. In other words, the de-identification processing of the program S23 includes two parts of the de-identification processing, one is the de-identification processing on the pixel data, and the other is the attribute de-identification processing (for DICOM tags). In some cases, the tag corresponding to at least one of the attribute values of the target image data that is de-identified matches the at least one of the template data D(D') selected by the processing unit 114 One to identify the label. Among the tags "(0008,1090)", "0028,0010", "0028,0011", and "(0010,0010)" in the target image data of the aforementioned use case, only "(0010,0010)" Match the handling order The de-identification label "(0010,0010)" of the template data D'selected by the element 114, therefore, the processing unit 114 removes the attribute value "CHEN CHI WEN" corresponding to "(0010,0010)" Identification. In this embodiment, the processing unit 114 configured to de-identify the attribute value includes deleting the content or part of the attribute value. In other implementation aspects, the processing unit 114 may be configured to replace the content or part of the attribute value. In other embodiments, the processing unit 114 can be configured to use special arithmetic logic to convert the content or part of the attribute value into another group of symbols that are different from the attribute value, for example, the The ID of the patient is converted into a group of numbers, so that when necessary, the content of the attribute value can be reversed by the arithmetic logic. In other implementations, the de-identification of attribute values can be processed globally, for example, directly specify which attribute values are to be identified, and then directly process all the image data, so that there is no need to specialize in the template data Specify which images should be specifically identified which attribute values.

Finally, in step S24, the processing unit 114 generates output image data based on the de-identified target image data, the output image data having the de-identified target image and the de-identified target image Attribute value. It should be noted that the generated de-identified target image is a single-layer image, instead of superimposing the de-template image and the target image in different layers.

FIG. 4 shows an exemplary de-identified target image 410 according to the present disclosure, in which the patient's name (ie, the content in the de-identified area) has been obscured by a mask 411.

It is worth noting that the target image data from a specific device model usually has a similar image resolution and layout. Therefore, the processing unit 114 uses the de-identification area indicated by the selected template data D(D') to be able to Quickly and effectively de-identify a large number of specific characters in the target image.

On the other hand, if the aforementioned de-identification method is not used, other de-identification methods can be used instead, for example, the use of optical character recognition (OCR) for each target image to identify the medical image Mask all the text after all the text on it, The time required is positively related to the size of the image, about 2 to 5 seconds. When a large number of medical images need to be de-identified, the working time is long. Take the processing of one million medical images in a single day as an example. A single thread requires 800 hours, which significantly increases the cost of wasting time and computing resources.

Figure 5 shows a flowchart of a method for generating template data implemented by some exemplary de-identification systems according to the present disclosure. The following will take the template data D'as an example for description.

First, in step S51, the processing unit 114 receives sample data with a file format definition. In this embodiment, the sample data is, for example, image data conforming to the DICOM standard. The sample data has one or more sample images and attribute information. The attribute information includes multiple tags and multiple attribute values respectively corresponding to the tags. Each sample image has image information and text. To facilitate understanding, an exemplary usage scenario is applied below to assist in explaining the subsequent procedures S52 to S55. In the example, the sample image data has only three sample images. In the example, the sample data has an attribute value "HUANG WEI DA" (patient name) and a corresponding tag "(0010, 0010)".

Figures 6 to 8 show some exemplary sample images 610, 710, 810 according to the present disclosure.

Fig. 6 exemplarily shows the image information 611 and text 612, 612' of the sample image 610. The image information 611 is an ultrasound image, the text 612 is related to the patient's body temperature, and the text 612' is the patient's name.

Then, as in procedure S52, the processing unit 114 executes a predetermined procedure for each sample image. The predetermined procedure includes procedure S521 and procedure S522.

In the procedure S521, the processing unit 114 uses Optical Character Recognition (OCR) to recognize the characters in the sample image. Taking the sample image 610 in FIG. 6 as an example, the characters 612, 612' recognized by the processing unit 114 include symbols such as "37.0C", "<37.0C", and "HUANG WEI DA".

In the procedure S522, the processing unit 11 selects a text that meets a predetermined condition among the text recognized from the sample image. In some cases, not all symbols need to be de-identified. For example, "HUANG WEI DA" needs to be de-identified, while "37.0C" and "<37.0C" do not need to be de-identified. Procedure S522 is used To select the characters that need to be de-identified (that is, those that meet the predetermined conditions) from all the characters recognized by the OCR. In this embodiment, the processing unit 114 uses the following procedure to determine whether each character meets the predetermined condition. For the convenience of description, the following takes the determination of whether the text 612' (ie "HUANG WEI DA") meets the predetermined conditions as an example for description.

First, the processing unit 114 determines whether the text matches one of the attribute values of the sample data. Following the previous example, the processing unit 114 compares the attribute value of the sample data and determines whether there is an attribute value that is the same as or similar to the text 612' (ie, "HUANG WEI DA").

Then, when the processing unit 114 determines that the character matches one of the attribute values of the sample data, it determines whether the label corresponding to the attribute value matching the character is the same as the stored one. One of the preset tags T and T′ in the storage module 112. Continuing the previous example, the processing unit 114 determines that the attribute value "HUANG WEI DA" of the sample data is indeed the same as the text 612' (ie "HUANG WEI DA"), and therefore determines whether there is a corresponding attribute value "HUANG WEI DA" The label (ie "(0010,0010)") is the same as the default label T(T').

Then, when the processing unit 114 determines that the tag corresponding to the attribute value matching the text is the same as one of the preset tags T, T', it determines that the text meets the predetermined condition. Continuing the previous example, the processing unit 114 determines that the default tag T'(ie "(0010,0010)") is the same as the tag corresponding to the attribute value "HUANG WEI DA" (ie "(0010,0010)") , So it is determined that the text 612' (ie, "HUANG WEI DA") meets the predetermined conditions.

In this embodiment, the processing unit 114 also uses the translation module 113 to translate the text recognized from the sample image, which does not meet the predetermined condition. It is a predetermined language, and it is determined whether the translated text meets the predetermined condition, and then the translated text meets the predetermined condition. In this embodiment, the predetermined language is, for example, English, but not limited to this. In other exemplary situations, if the characters recognized by the processing unit 114 are other language characters that are different from English, such as Chinese, it is possible to obtain the determination result that the Chinese characters do not meet the predetermined condition in the procedure S522. . The processing unit 114 uses the translation module 113 to translate the Chinese text into English, and again determines whether the translated text meets the predetermined condition, and selects the translated text Among them, those who meet the predetermined conditions. In this way, the chance of misjudgment caused by language differences is reduced.

The manner of determining whether the text meets the predetermined condition is not limited to this embodiment. Although the attribute value corresponding to the tag (0010, 0010) in the DICOM file should be the patient's name, in some cases, the content of the attribute value corresponding to the tag (0010, 0010) may vary depending on the source hospital or equipment It is different. It may fill in Chinese name, English translation name, medical record number, ID number, or even not fill in. Therefore, in other implementation modes, additional information from additional sources with higher accuracy (such as the National Health Insurance Agency) can be used to obtain their real name and other personal information. When the real name obtained is a Chinese name, it will be written in various English. The translation system converts the name into an English translation, so as to compare the name in the text recognized by optical character recognition, and then use the comparison result to create a mask. Take the National Health Insurance Agency as an example, the additional information can be the information declared together with the image (such as DICOM files).

6 to 8, in the foregoing example, at this stage, the processing unit 114 determines that the sample images 610, 710, and 810 each have a text 612', 712', 812 that meets the predetermined conditions '.

Then, next, as in procedure S53, the processing unit 114 generates template data, the template data instructs to identify the D(D') region, and the de-identification region corresponds to the text that meets the predetermined condition, and The template data D (D′) is stored in the storage module 111. In this embodiment, the processing unit 114 is based on the text that is recognized from the sample image and meets the predetermined condition. Among them, the one that occupies the largest area determines the coverage of the de-identified area, so that most length information can be effectively shielded. The details are as follows.

Referring to FIGS. 9 to 11, first, the processing unit 114 defines a reference area in the sample image, and the reference area respectively covers the area where the text that meets the predetermined condition is located. For example, the processing unit 114 defines a reference area 913 in the sample image 610, which covers the area where the text 612' (ie "HUANG WEI DA") is located; defines a reference area 1003 in the sample image 710, which covers The area where the text 712' (ie "CHEN CHI WEN") is located; the sample image 810 defines a reference area 1103, which covers the area where the text 812' (ie, "ALICE WANG") is located.

Next, referring to FIGS. 9 to 11, the processing unit 114 also uses the reference regions of the sample images that are located at the same position in each sample image as multiple comparison regions. Continuing the previous example, since the reference areas 913, 1003, and 1103 in the previous example are all located at the same position in each sample image 610 (710 or 810), the three reference areas 913, 1003, and 1103 are respectively regarded as three Two comparison areas 914, 1004, 1104.

Then, the processing unit 114 determines the coverage of the de-identified area corresponding to the comparison area according to one of the comparison areas that covers the largest area. Among the comparison areas 914, 1004, and 1104, the comparison area 914 occupies the largest range. Therefore, in FIG. 12, the de-identification area 1202 corresponds to the comparison areas 914, 1004, and 1104, and the de-identification area The size of the area occupied by 1202 is determined according to the area occupied by the comparison area 914. In this way, the de-identification area 1202 can effectively cover most of the length of names.

To supplement, in other cases, when the sample image data received by the processing unit 114 contains only one sample image, for each reference area, only one comparison area is obtained, and the comparison area is the largest occupied area. By.

Figure 13 shows some exemplary de-identification systems according to the present disclosure (e.g., de-identification The flow chart of the de-identification method implemented by the system 110).

S131: The processing unit (for example, the processing unit 114) receives sample data with sample images, the sample images having text.

S132: The processing unit uses Optical Character Recognition (OCR) to recognize the text in the sample image.

S133: The processing unit selects a word that meets a predetermined condition among the words recognized from the sample image.

S134: The processing unit generates template data, the template data indicates an identification area, and the de-identification area corresponds to a text that meets the predetermined condition, and stores the template data in a storage module (for example, storage Module 112).

S135: The processing unit receives target image data with a target image.

S136: The processing unit selects one of a plurality of template data stored in the storage module.

S137: The processing unit de-identifies the content of the de-identified area corresponding to the selected template data in the target image.

S138: The processing unit generates output image data, the output image data having the de-identified target image.

In summary, in the above-mentioned embodiment, the processing unit 114 can not only generate the corresponding template data D(D') based on the received sample image data and the preset label T(T') using OCR, but also according to the existing The template data D and D'stored in the storage module 112 quickly de-identify the received target image data. In this way, the time and computing resources required for de-identifying the image data can be effectively reduced.

However, the above are only examples of the present invention. When the scope of implementation of the present invention cannot be limited by this, all simple equivalent changes and modifications made in accordance with the scope of the patent application of the present invention and the content of the patent specification still belong to This invention patent covers the scope.

S131~S138: Program

Claims (13)

A de-identification method includes: a processing unit receives sample data with a sample image, the sample image has text; the processing unit uses optical character recognition (Optical Character Recognition, OCR) to identify the sample image The text; the processing unit selects and recognizes from the text in the sample image the text that meets a predetermined condition; the processing unit generates template data, the template data indicates the identification area, the de-identification The transformation area corresponds to the text that meets the predetermined conditions and stores the generated template data; the processing unit receives target image data with a target image; the processing unit selects one of the stored template data; The processing unit de-identifies the content of the de-identified area corresponding to the selected template data in the target image; and the processing unit generates output image data, the output image data having the de-identified The target image; wherein: the sample data includes a plurality of sample images; the processing unit determines the de-identification area according to the text that is recognized from the sample image and meets the predetermined condition, which occupies the largest range The scope of coverage. The de-identification method according to claim 1, wherein: the sample data further has a plurality of tags and a plurality of attribute values corresponding to the tags; and The processing unit determines whether the text meets the predetermined condition by the following procedure: determining whether the text matches one of the attribute values of the sample data, and after determining that the text matches the sample When one of the attribute values of the data is determined, it is determined whether the tag corresponding to the attribute value matching the text is the same as one of a plurality of preset tags, and when it is determined that the tag matching the text is When the tag corresponding to the attribute value is the same as one of the preset tags, it is determined that the text meets the predetermined condition. The de-identification method according to claim 1, further comprising: a translation module connected to the processing unit data to translate the text recognized from the sample image, and the text that does not meet the predetermined condition is translated Is a predetermined language; and the processing unit determines whether the translated text meets the predetermined condition. The de-identification method according to claim 1, wherein: each template data has at least one template attribute value; the target image data also has at least one attribute value; and the processing unit is selecting and storing in the storage module When one of the plurality of template data is selected, the at least one template attribute value of the selected template data matches at least one attribute value of the target image data. The de-identification method according to claim 4, wherein: The information described by each template attribute value includes one of a medical institution, an instrument manufacturer, a device model, an image type, an image size, and an image type. A method for generating template data includes: a processing unit receives sample data with a sample image, the sample image has text; the processing unit uses optical character recognition (OCR) to identify the sample image The text; the processing unit selects and recognizes the text in the sample image which meets a predetermined condition text; and the processing unit generates template data, the template data indicating the recognition area, the The de-identified area corresponds to the text that meets the predetermined condition; wherein: the sample data includes a plurality of sample images; the processing unit occupies the largest portion of the text that is identified from the sample image and meets the predetermined condition The scope person determines the scope of the de-identified area. The method for generating template data according to claim 6, wherein: the sample data further has a plurality of tags and a plurality of attribute values corresponding to the tags; and the processing unit determines whether the text is Meet the predetermined condition: determine whether the character matches one of the attribute values of the sample data, and when it is determined that the character matches one of the attribute values of the sample data, determine a match Whether the tag corresponding to the attribute value of the text is the same as one of a plurality of preset tags, and When it is determined that the tag corresponding to the attribute value matching the text is the same as one of the preset tags, it is determined that the text meets the predetermined condition. The method for generating template data according to claim 6, further comprising: a translation module connected to the processing unit data to recognize the text in the sample image, and the text does not meet the predetermined condition Translate into a predetermined language; and the processing unit determines whether the translated text meets the predetermined condition. A de-identification system includes: a storage module that stores a plurality of template data; and a processing unit connected to the storage module for data; the processing unit is configured to receive sample data with a sample image, the sample The image has text; the processing unit is configured to use optical character recognition (OCR) to recognize the text in the sample image; the processing unit is configured to select the text that is recognized from the sample image Among the texts, texts that meet a predetermined condition; the processing unit is configured to generate template data, the template data indicates the identification area, the de-identification area corresponds to the text that meets the predetermined condition, and The generated template data is stored in the storage module; the processing unit is configured to receive target image data with target images; the processing unit is configured to configure to select a plurality of template data stored in the storage module One of them The processing unit is configured to de-identify the content of the de-identified area corresponding to the selected template data in the target image; and the processing unit is configured to generate output image data, the output image The data has the de-identified target image; wherein: the sample data includes a plurality of sample images; the processing unit is configured to occupy the image according to the text identified from the sample image and meet the predetermined condition The person with the largest range determines the coverage of the de-identified area. The de-identification system according to claim 9, wherein: the storage module further stores a plurality of preset tags; the sample data further has a plurality of tags and a plurality of attribute values corresponding to the tags; the processing The unit is assembled to determine whether the text meets the predetermined condition by the following procedure: it is determined whether the text matches one of the attribute values of the sample data, and when it is determined that the text matches the sample When one of the attribute values of the data is determined, it is determined whether the tag corresponding to the attribute value matching the text is the same as one of a plurality of preset tags, and when it is determined that the tag matching the text is When the tag corresponding to the attribute value is the same as one of the preset tags, it is determined that the text meets the predetermined condition. The de-identification system as described in claim 9, further comprising: A translation module, which is connected to the processing unit for data, configured to translate the text that does not meet the predetermined condition among the text recognized in the sample image into a predetermined language; and the processing unit group Is configured to determine whether the translated text meets the predetermined condition. The de-identification system according to claim 9, wherein: each template data has at least one template attribute value; the target image data also has at least one attribute value; and the processing unit is selectively storing in the storage module When one of the plurality of template data is selected, the at least one template attribute value of the selected template data matches at least one attribute value of the target image data. The de-identification system according to claim 12, wherein: the information described by each template attribute value includes one of a medical institution, an instrument manufacturer, a device model, an image type, an image size, and an image type By.

Images