KR101718159B1 - Method for Extracting Region of Interest Value From Medical Image and Computer Readable Recording Medium Recorded with Program for Performing the Same Method - Google Patents

Abstract

The present invention relates to a method for automatically extracting a ROI value from a medical image provided by a medical image storage system (PACS) and a computer readable recording medium on which a program for performing the method is recorded. According to the method of the present invention, the ROI values from the medical image can be extracted by an automated method, so that the efficiency and accuracy are greatly improved as compared with the case where ROI values are manually extracted. The method of the present invention is excellent in general accessibility by using optical character recognition (OCR) software and a macro program which are open-source software. The present invention is expected to be applied to a medical image storage system (PACS) to greatly improve the efficiency of this system.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention This invention relates to a method for extracting a ROI value from a medical image and a computer-readable recording medium on which a program for performing the method is recorded. Same Method}

The present invention relates to a method for extracting a ROI value from a medical image and a computer-readable recording medium on which a program for performing the method is recorded.

In the conventional film-based imaging medicine field, setting the region of interest (ROI) has been limited to a CT-console or an image-reconstruction workstation. As Picture Archiving and Communicating System (PACS) is used in digital imaging medicine, there is a growing demand for setting ROIs on PACS viewers for various purposes. (ROI) including the area value, the mean value, the standard deviation value, the max value and the min value, and the like in the setting of the ROI, Are displayed on the screen (see Figure 1). The radiologist reads the figures on the PACS screen and types them on the required radiation report. Sometimes, especially for research purposes, radiologists must set up a large number of ROIs and manually enter relevant values on the spreadsheet. However, this manual processing is repetitive and cumbersome, and errors may occur in this process.

Optical character recognition (OCR) technology is a technique that mechanically or electronically converts a typed, scanned, or printed text image into a digitally encoded text format [1]. The OCR engine can convert image values into text data and is widely used for pattern recognition and character recognition. In digitized radiological environments, applications in radiology using OCR have been reported [2-4].

Numerous papers and patent documents are referenced and cited throughout this specification. The disclosures of the cited papers and patent documents are incorporated herein by reference in their entirety to better understand the state of the art to which the present invention pertains and the content of the present invention.

Korean Patent No. 10-1287382 Korean Patent No. 10-1099446 U.S. Published Patent Application No. 2013/0251233 A1

1. Optical character recognition. Wikipedia. Available at http://en.wikipedia.org/wiki/Optical_character_recognition. Accessed August 24, 2013. 2. Cook TS, Zimmerman S, Maidment AD, Kim W, Boonn WW. J Am Coll Radiol 2010; 7: 871-877. 3. Li X, Zhang D, Liu B. AJR Am J Roentgenol 2011; 196: W781-783. 4. Lee YH, Song HT, Suh JS. Journal of digital imaging 2012; 25: 815-818. 5. AutoHotkey Web site. Available at http://www.autohotkey.com. Accessed August 24, 2013. 6. GOCR Web site. Available at http://jocr.sourceforge.net/api Accessed August 24, 2013. 7. Goyal N, Jain N, Rachapalli V. Clin Radiol 2009; 64: 119-126. 8. Harisinghani MG, Blake MA, Saksena M, et al. Radiographics 2004; 24: 615-627.

The present inventors have sought to develop a method for automatically extracting statistical values of a ROI from a medical image provided by a PACS (Picture Archiving and Communicating System). As a result, an area where the values of the ROI are displayed in the image of the medical image storage system (PACS) is captured as a graphics file, and then the characters and numerals of the captured image are captured by optical character recognition (OCR) To text, and to extract the statistical values of the ROI from the converted text file successfully, thereby completing the present invention.

Accordingly, it is an object of the present invention to provide a method for extracting a ROI value from a medical image.

Another object of the present invention is to provide a computer readable recording medium on which a program for automatically performing the above method is recorded.

The objects and advantages of the present invention will become more apparent from the following detailed description of the invention, claims and drawings.

According to one aspect of the present invention, the present invention provides a method of extracting a ROI value from a medical image comprising the steps of:

(a) capturing, as a graphic file, an area in which values of a ROI are displayed from an image of a PACS (Picture Archiving and Communicating System);

(b) converting values of a region of interest (ROI) of the captured graphic file image into a text file using optical character recognition (OCR) software;

(c) correcting errors in the text file converted by the optical character recognition (OCR);

(d) extracting statistical values of the ROI from the error corrected text file;

(e) constructing the extracted statistical values in a temporary string format; And

(f) copying the temporary string format into a spreadsheet.

Hereinafter, the present invention will be described in detail with reference to the respective steps.

(A) capturing a region of a region of interest (ROI) from a video of a PACS (Picture Archiving and Communicating System) into a graphic file;

As used herein, the term " Picture Archiving and Communicating System (PACS) " refers to a system that digitally acquires, stores, transmits, and queries medical images in an integrated manner. More specifically, the present invention refers to a system that enables medical imaging professionals to read medical images by converting medical images taken by various devices into digital images, storing them in a mass storage device, transmitting and inquiring them through a monitor.

In the present invention, the image of the medical image storage system (PACS) is preferably an X-ray image, a computed tomography image, an MRI (magnetic resonance imaging) image, a PET (positron emission tomography) computed tomography) images.

As used herein, the term "region-of-interest " (ROI) refers to an area set for a particular medical purpose in the displayed medical image. For example, to measure the size of a tumor, the tumor refers to a region defined by a border containing the tumor in the imaged image.

As used herein, the term "value of ROI" refers to a specific value that represents information related to a region of interest, and is preferably an area value, an average value, a standard deviation (ROI) values, such as a value of a region of interest (ROI), a maximum value, and a minimum value.

An area including values of ROIs from an image displayed on the PACS viewer is captured and temporarily stored in a graphic file.

As used herein, the term "capture" means to separate necessary portions of the original image or image information for editing or storage. That is, it means to select some information displayed on the screen for editing or saving such as converting to a text string or a graphic file.

The capture may be performed using a variety of non-commercially accessible open source programs.

The stored graphic file should include values of ROI that are converted into a text file by a subsequent optical character recognition (OCR). The graphics file is not limited to a specific format, but can be a file format that can be recognized by Optical Character Recognition (OCR) and converted to text format.

(B) converting the values of the ROI of the captured graphic file image into a text file using optical character recognition (OCR) software;

Converts the values of ROIs into an image file in an image of the captured graphic file by optical character recognition (OCR) software. The converted text file can be stored in a temporary folder.

Step c): correcting the error in the text file converted by the optical character recognition (OCR)

And corrects the error in the text file converted by the optical character recognition (OCR). The error refers to values recognized as characters or digits of an incorrect format that are erroneously recognized when the values of ROIs are converted into text files by OCR. Representative patterns of these errors can be segmentation errors, extra-spaces errors, and errors that recognize numbers as characters. For example, "1, 441" is recognized as "1, 44 1", and correction of such a division error is performed by removing space between numbers and characters . The oversampling error is an error generated when the space between recognized characters is excessively set and recognized. The oversampling error can be corrected by removing the unnecessary double space by a single space or space. An error recognized as a character of the number is a case where the number is mistakenly recognized as a character of a similar type, for example, when "8" is recognized as "B".

In the present invention, the recognition error due to the optical character recognition (OCR) can be corrected by a macro program.

As used herein, the term "macro program" has the same meaning as a macro generating program. In the present specification, a macro program is a program in which a program in a series of related commands is regarded as one macro instruction, it means.

(D) extracting statistical values of the ROI from the error corrected text file;

And extracts statistical values of the ROI from the string of the corrected text file. The statistical values of the extracted ROI are preferably an area value, an average value, a standard deviation value, a max value, and a min value But is not limited thereto.

Step (e): constructing the extracted statistical values in a temporary string format

An area value, an average value, a standard deviation value, a maximum value, and a minimum value, which are statistical values of the extracted ROI, are formed in the temporary string format of the order. Preferably, the interval between the statistical values is divided by a tab.

Step (f): copying the temporary string format into a spreadsheet

To export the prepared temporary string format to a spreadsheet of Excel software (Microsoft), copy it and paste it into a spreadsheet.

According to another aspect of the present invention, the present invention provides a computer-readable recording medium having recorded thereon a program for extracting a ROI value from a medical image comprising the steps of:

(a) capturing, as a graphic file, an area in which values of a ROI are displayed from an image of a PACS (Picture Archiving and Communicating System);

(b) converting values of a region of interest (ROI) of the captured graphic file image into a text file using optical character recognition (OCR) software;

(c) correcting errors in the text file converted by the optical character recognition (OCR);

(d) extracting statistical values of the ROI from the error corrected text file;

(e) constructing the extracted statistical values in a temporary string format; And

(f) copying the temporary string format into a spreadsheet.

The description of each of the above steps in the computer-readable recording medium on which the program of the present invention is recorded will be described with reference to a method of extracting ROI values from an image of a medical image storage system (PACS) The description is not duplicated.

The features and advantages of the present invention are summarized as follows:

(i) The present invention relates to a method for automatically extracting a ROI value from a medical image provided by a medical image storage system (PACS) and a computer readable recording medium on which a program for performing the method is recorded .

(Ii) According to the method of the present invention, the ROI values can be extracted from the medical image in an automated manner, so that efficiency and accuracy are greatly improved compared to extracting by ROI values manually do.

(Iii) The method of the present invention is excellent in universal accessibility by using optical character recognition (OCR) software and a macro program which are open-source software.

(Iv) The present invention is expected to be applied to a medical image storage system (PACS) to greatly improve the efficiency of this system.

Figure 1 shows a screen shot of a PACS. Panel A of Figure 1 is the ROI values displayed on the PACS screen. As a large image, ROI values overlap with background signals. Panel B of FIG. 1 shows that ROI values are displayed without background, by zoom-out.
Figure 2 is a flow chart of a ROI extraction module. PACS is a Picture Archiving and Communicating System; ROI is region-of-interest; OCR represents optical character recognition.
Figure 3 shows the OCR results of the region of interest (ROI). Panel A of FIG. 3 is an image that captures ROI values. Panel B in Figure 3 is raw text that has not been processed from the OCR. Panel C in Fig. 3 is a result of error correction and space removal of the original text.
Figure 4 shows a spreadsheet of Excel software. The extracted ROI values were successfully inserted onto the spreadsheet with suitable cell-separation.

Hereinafter, the present invention will be described in more detail with reference to Examples. It is to be understood by those skilled in the art that these embodiments are only for describing the present invention in more detail and that the scope of the present invention is not limited by these embodiments in accordance with the gist of the present invention .

Example

Tools and Methods

1. Hardware and Software

The PACS workstation (DB-P600, Samsung, Korea) consists of 3.3-GHz i5-2500 (Intel, Santa Clara, CA), 4096 MB random access memory (RAM), internal video graphics array HD- And an external video graphics array NVIDIA Quadro FX 1400 (NVIDIA, Santa Clara, Calif.) And a 5-megapixel flat-panel LCD (liquid-crystal display) monitor (Totoku Electric, Japan). The PACS software was a Windows-based PACS viewer, Centricity® PACS RA1000 (GE Healthcare, Barrington, IL). The operating system (OS) program used Microsoft Windows 7. I designed the main module using AutoHotkey (version 1.0.48). This open source macro program can be downloaded from the official web page (http://www.autohotkey.com) [5]. We selected GOCR version 0.49 as OCR function, and this open source OCR software can be downloaded from official web page (http://jocr.sourceforge.net/api) [6].

2. Main module of ROI value extraction

By pressing a predefined computer key once (keystroke), the main module is activated as follows to convert ROI values on the screen to ROI values on the text. The detailed conversion process is as follows (see FIG. 2).

(i) Capture area of region of interest (ROI) values as a graphic file :

OCR-targeted captured screen shots with ROI values for OCR were stored as JPG-formatted graphics files in a temporary folder on a local drive (see FIG. 1). We designed the OCR to work within a specific area where ROI values are expected to be located for efficient image screen shots.

(Ii) OCR of the captured image :

The OCR engine loaded the JPG-formatted graphics file (see panel A of FIG. 3) and converted the characters of the ROI values in the image file into a text formatted file (see panel B of FIG. 3). The converted text file was saved in a temporary folder.

(Iii) Correction of errors in optically recognized ROI values :

Correction of errors in Optical Molecognition (OCR) process was automatically performed by software written by macro-string replacement operations. Typical recognition error patterns in OCR engines in modules are segmentation errors, extra-spaces errors, and errors that recognize numbers as characters. For example, a partition error is an error that recognizes "1,441" as "1, 44 1". To correct this error, a macro program was used to remove the space between numbers. An unnecessary double space was replaced by a single space or space removal (see panel C of FIG. 3). Because the ROI statistical values are usually numbers, in the case of an error in which a number is recognized as a character in the ROI value, for example, when "O" is recognized as "Ο" or "o" , "8" was recognized as "B", and "7" was mistakenly recognized as "Z".

(Iv) Extraction of statistical values of ROIs:

The statistical values of the ROI including the area value, the average value, the standard deviation value, the max value and the min value are stored in an OCR-recognized string . Within the PACS viewer (Centricity, GE Healthcare), the ROI value formats are A [Area Value], Average [Average Value], SD [Standard Deviation Value], Max [Maximum Value], and Min [Minimum Value] to be.

(V) Making temporary strings spaced by tabs :

Temporary formats are tab-spaced values. An example of a predefined template of a temporary string was: 101.9 [Tab] 1885.8 [Tab] 67.9 [Tab] 2080 [Tab] 1714 [Tab]. Here, [tab] means that spacing between characters is spaced by a tab. These strings can be used to paste into Excel (Microsoft) spreadsheets for data separation.

(Vi) copy temporary strings to the clipboard

Temporary strings are copied for export to a spreadsheet. The string can be paste and cell-separated on a spreadsheet by sending a [Control-V] key for paste (see FIG. 4).

3. Evaluating and comparing program accuracy

The accuracy of the OCR of selected ROI values was evaluated from the recognition of 1040 ROIs of 280 ROIs from magnetic resonance images obtained from PACS server data between March 2010 and December 2011. One ROI measurement includes five ROI values: an area value, an average value, a standard deviation value, a maximum value, and a minimum value. The OCR-recognized ROI values from the extraction module were compared to the ROI values visually identified by one radiologist. If a recognition error or failure occurred, the extraction system was repeated after zooming out to avoid unnecessary background images from disturbing the OCR recognition process. To compare with the manual input, the input time on the Excel spreadsheet was compared between the manual input and the extraction module help input. Manual method and automated extraction module - To compare the methods of help, one radiologist recorded the input time in each method. The images on the PACS are zoomed out so that the ROI values are not obscured. As a reference, the accuracy of the results by the visual recognition manual method was also evaluated. A total of 12 ROIs were compared on a PACS workstation as three sets of four ROIs.

Experiment result

The extraction module of ROI values has been successfully performed by OCR and macro program. Five values of the area value, the average value, the standard deviation value, the maximum value, and the minimum value could be recognized by OCR. Also, ROI values are extracted and error corrected by an AutoHotkey-coded module, and each ROI value can be separated into tab-space. Accuracy was assessed for 1040 perceptions of 280 ROIs of a magnetic resonance image, and the value extraction results were 100% accurate without background overlapping ROIs Respectively. The average input times between conventional manual input methods and automated extraction module assist methods were 34.97 seconds and 7.87 seconds, respectively. In this time-saving comparative test, the accuracy of the extraction module of the present invention was 100%.

Review

The use of computers has been increasing in the medical field and especially in the field of radiology [7]. In order to minimize physical stress caused by repetitive tasks, film-less radiology introduced language recognition software and electronic questionnaire formats, but it could not replace the task of setting ROIs. The setting of the ROI is one of the most tedious and sexually charged tasks in imaging and research. It is also well known that people who use computers every day can suffer from musculoskeletal diseases caused by repetitive physical stresses [7, 8]. When the ROI values are visually recognized on the screen and manually entered, errors due to recognition errors and typographical errors may occur. Such errors may cause very serious consequences. The present inventors have wondered whether a commercially available PACS view does not have a useful function of storing a ROI in a temporary file or copying ROI values to a clipboard, We examined whether these useful functions could be included in the PACS generation. For the development of extraction modules of ROI values, we used open source software that includes AutoHotkey for GOCR and macro functions for OCR and applied it to ROI settings (ROI) values with error correction functions using these modules, which are capable of automatically extracting ROI values. The method of the present invention ensures accuracy and minimizes the number of repeated visual-confirmation and typing operations. In our accuracy experiments, the accuracy of OCR was 100%. Since the values of the ROI are all numerical values, it is considered that there is little recognition error in the OCR. Error correction of the converted ROI values was performed by a macro program. Common recognition error patterns, including errors that recognize segmentation errors, extra-spaces, and characters, have been successfully corrected. The format of the ROI values may vary depending on the PACS viewer, and values may be extracted in a specific format. This format may be changed according to the ROI format. However, there is a limit to recognition due to the presence of background signal strength. If the ROI values are interfered by the background on the screen, a perceptual error may occur (see panels A and B of FIG. 1). In the study of the present invention, To remove the background signal by zooming out. In conclusion, a simple and efficient extraction module of ROI values could be established by using open source OCR software and macro programs. These modules and concepts will be applicable to next generation PACS.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the same is by way of illustration and example only and is not to be construed as limiting the scope of the present invention. Accordingly, the actual scope of the present invention will be defined by the appended claims and their equivalents.

Claims (14)

A method for extracting a region-of-interest (ROI) value from a medical image comprising the steps of:
(a) capturing a region of a region of interest (ROI) from a video of a PACS (Picture Archiving and Communicating System) into a graphic file, wherein the image of the medical image storage system is X (SPECT) image or a single photon emission computed tomography (SPECT) image, wherein the background region is a region of interest, a CT (computed tomography) image, a magnetic resonance imaging And removing the background signal by zooming out the image when a recognition error of the values of the image is generated;
(b) converting values of a region of interest (ROI) of the captured graphic file image into a text file using optical character recognition (OCR) software;
(c) correcting an error in a text file converted by the optical character recognition (OCR), wherein the error is a segmentation error, an extra-space error, Is an error;
(d) extracting statistical values of a ROI from the error corrected text file, wherein the statistical values of the ROI include an area value, an average value, a standard deviation deviation value, a max value, and a min value;
(e) constructing the extracted statistical values in a temporary string format; And
(f) copying the temporary string format into a spreadsheet.
delete delete The method of claim 1, wherein the error correction in step (c) is performed by a macro program.
delete The method of claim 1, wherein the temporary character string in step (e) is in the order of an area value, an average value, a standard deviation value, a maximum value, and a minimum value.
7. The method of claim 6, wherein a gap is formed between each statistical value in the temporary string by a tab.
A computer-readable medium having recorded thereon a program for extracting a region of interest (ROI) value from a medical image comprising the steps of:
(a) capturing a region of a region of interest (ROI) values from a video image of a PACS into a graphics file, the PACS The image may be an X-ray image, a computed tomography (CT) image, a magnetic resonance imaging (MRI) image, a positron emission tomography (PET) image, or a single photon emission computed tomography (SPECT) And removing a background signal by zooming out the image when a recognition error of the values of the ROI occurs;
(b) converting values of a region of interest (ROI) of the captured graphic file image into a text file using optical character recognition (OCR) software;
(c) correcting an error in a text file converted by the optical character recognition (OCR), wherein the error is a segmentation error, an extra-space error, Is an error;
(d) extracting statistical values of a ROI from the error corrected text file, wherein the statistical values of the ROI include an area value, an average value, a standard deviation deviation value, a max value, and a min value;
(e) constructing the extracted statistical values in a temporary string format; And
(f) copying the temporary string format into a spreadsheet.
delete delete 9. The method according to claim 8, wherein correction of the error in step (c) is performed by a macro program. A computer readable recording medium on which a program for extracting a ROI value from a medical image is recorded media.
delete 9. The method according to claim 8, wherein the temporary string in the step (e) is in the order of an area value, an average value, a standard deviation value, a maximum value and a minimum value. And a computer readable recording medium storing the program.
14. The computer-readable medium of claim 13, wherein a program is provided for extracting a ROI value from a medical image, wherein spacing between each statistical value in the temporary string is formed by a tab. Recording medium.

Images