KR102414299B1 - System and method for enhancing quality of CT(computed tomography) scan using AI(artificial intelligence) - Google Patents

Abstract

The present invention relates to a system and method for improving the quality of a CT scan using artificial intelligence.
According to the present invention, there is provided a method for improving the quality of a CT scan using artificial intelligence, comprising: receiving a CT scan of a predetermined thickness and interval and applying a thickness reduction AI model to each CT slice of the CT scan; acquiring a CT scan with reduced thickness by applying the thickness reduction AI model; obtaining a spaced-interpolated CT scan by applying an interval interpolation AI model to the obtained CT scan with reduced thickness; realigning the thickness-reduced, spaced-interpolated CT slices; and finally acquiring a CT scan of a predetermined thickness and spacing with improved quality based on the rearranged CT slices.
According to the present invention, it is possible to improve the quality of a low-resolution CT scan to a high-resolution quality by receiving a low-resolution CT scan and performing thickness reduction and interval interpolation using artificial intelligence technology.

Description

System and method for enhancing quality of CT (computed tomography) scan using AI (artificial intelligence)}

The present invention relates to a system and method for improving the quality of a CT scan, and more particularly, to a system and method for improving the quality of a CT scan using artificial intelligence that can improve a low-resolution CT scan to a high-resolution quality using artificial intelligence technology is about

In a CT scan, the resolution means the thickness of the CT slices constituting the CT scan and the spacing between the slices.

In general, the film thickness and the slice interval have the same value. In a 1mm scan, information corresponding to a thickness of 1mm is taken on one slice, and the interval between slices is also 1mm. If only information in a narrow space (1mm) is included in the same data (slice) size, it can be viewed in high resolution, and if information in a wide space (3mm) is included, it can be viewed in low resolution.

Except for large hospitals, most primary and secondary medical institutions use low-resolution CT. In such low-resolution CT, it is difficult to confirm detailed information because a wider range of information is compressed (squeeze) in one slice. Therefore, when there is a fine fracture in the bone, it may be difficult to identify such a fine fracture through a low-resolution CT scan.

On the other hand, Korean Patent Application Laid-Open No. 10-2019-0119550 (Patent Document 1) discloses "a method and apparatus for improving the resolution of an image". loading image data including an image and metadata for the image data; analyzing metadata including information related to an artificial neural network for image processing to be applied to the low-resolution image of the image data; selecting an artificial neural network for image processing to be applied to the low-resolution image from a group of artificial neural networks for image processing including a plurality of artificial neural networks for image processing, based on the metadata; and generating a high-resolution image by processing the low-resolution image according to the selected artificial neural network for image processing.

In the case of Patent Document 1 as described above, although there is an advantage that resolution quality can be improved when converting different images to high resolution by using different neural networks for image processing for each still image or moving image frame image, the low resolution of image data It is configured to use different neural networks for image processing for each image, so there is a problem in that a vast number of neural networks for image processing must be provided for different image processing for each low-resolution image.

Korean Patent Publication No. 10-2019-0119550 (published on October 22, 2019)

The present invention was created to improve the problems of the prior art as described above, and by using artificial intelligence technology to receive a low-resolution CT scan as input and perform thickness reduction and interval interpolation, the low-resolution CT scan is improved to high-resolution quality The purpose is to provide a system and method for improving the quality of a CT scan using artificial intelligence that can do it.

In order to achieve the above object, the system for improving the quality of CT scan using artificial intelligence according to the present invention,

an AI model generator for generating a thickness reduction AI (artificial intelligence) model for reducing the thickness of a slice of a CT scan and an interval interpolation AI model for interpolating an empty space between the slices;

a CT scan input unit receiving a CT scan of a predetermined thickness and interval;

A CT scan with reduced thickness is obtained by applying the thickness reduction AI model to each CT slice of the CT scan input through the CT scan input unit, and the interval is interpolated by applying the interval interpolation AI model to the reduced thickness CT scan. an AI model application unit that acquires a CT scan;

a CT slice reordering unit for rearranging the CT slices obtained by the AI model application unit with reduced thickness and interpolated intervals;

a final CT scan acquisition unit that finally acquires a CT scan of a predetermined thickness and spacing with improved quality based on the CT slice rearranged by the CT slice realigning unit; and

Controls the state check and operation of the AI model generation unit, CT scan input unit, AI model application unit, CT slice reordering unit, and final CT scan acquisition unit, and a predetermined thickness and improved quality obtained by the final CT scan acquisition unit; It is characterized in that it includes a control unit that displays the CT scan at intervals on the display screen and stores it in the database (DB).

Here, preferably, it further comprises a communication unit for transmitting the data of the CT scan of a predetermined thickness and interval with improved quality obtained by the final CT scan acquisition unit to an external device, and receiving data or commands from the external device. can do.

In addition, at least one of the thickness reduction AI model and the interval interpolation AI model generated by the AI model generator may be configured as a Convolutional Neural Network (CNN)-based deep-learning model.

In addition, when the AI model application unit obtains the spaced interpolated CT scan by applying the interval interpolation AI model to the CT scan with reduced thickness, a plurality of different interval interpolation AI models are applied to interpolate different intervals, respectively A CT scan can be obtained.

In addition, when the AI model application unit obtains the spaced interpolated CT scan by applying the gap interpolation AI model to the reduced-thickness CT scan, the pre- and post-slice of the reduced-thickness CT scan are used as input. After , it is possible to obtain an interpolated CT scan by interpolating the interval by creating an empty slice between the slices.

In addition, in order to achieve the above object, the method for improving the quality of a CT scan using artificial intelligence according to the present invention,

a) receiving a CT scan of a predetermined thickness and interval and applying a thickness reduction AI model to each CT slice of the CT scan;

b) acquiring a reduced-thickness CT scan by application of the reduced-thickness AI model;

c) obtaining a spaced-interpolated CT scan by applying an interval interpolation AI model to the obtained CT scan with reduced thickness;

d) rearranging the CT slices reduced in thickness by step b) and interpolated at intervals by step c); and

e) finally, based on the rearranged CT slices, acquiring a CT scan of a predetermined thickness and spacing with improved quality.

Here, the method may further include generating a thickness reduction AI (artificial intelligence) model that reduces the thickness of a slice of a CT scan before step a) and an interval interpolation AI model that interpolates an empty space between the slices. have.

In this case, at least one of the thickness reduction AI model and the interval interpolation AI model may be configured as a Convolutional Neural Network (CNN)-based deep-learning model.

In addition, in obtaining the spaced interpolated CT scan by applying the interval interpolation AI model in step C), a plurality of different interval interpolation AI models may be applied to obtain CT scans each having different intervals interpolated. have.

In addition, in obtaining the spaced interpolated CT scan by applying the interval interpolation AI model in step c), an empty slice between the before and after slices using the before and after slices of the CT scan with reduced thickness as input By interpolating the intervals by generating

According to the present invention, there is an advantage in that a low-resolution CT scan can be improved to a high-resolution quality by receiving a low-resolution CT scan and performing thickness reduction and interval interpolation using artificial intelligence technology.

1 is a diagram schematically showing the configuration of a system for improving the quality of a CT scan using artificial intelligence according to the present invention.
2 is a flowchart illustrating an execution process of a method for improving the quality of a CT scan using artificial intelligence according to the present invention.
3 is a diagram illustrating an outline of a slice thickness and a scan interval of a CT scan.
4 is a view showing a slice thickness and a scan interval of a CT scan for a model of a hand.
5 is a view showing a slice thickness and a scan interval of a three-dimensional CT scan of the wrist joint.
6 is a diagram illustrating an outline of performing thickness reduction and interval interpolation for quality improvement on a CT scan taken at a low resolution.
7 is a diagram illustrating thickness reduction by applying a thickness reduction model to each slice of a CT scan taken at a low resolution.
8 is a diagram illustrating interpolation of intervals by applying an interval interpolation model to a CT scan whose thickness is reduced by application of the thickness reduction model.
9 is a view showing that the thickness is reduced by applying the thickness reduction model and the interval interpolation model, and the final CT scan with improved quality is obtained by interpolating the interval.
FIG. 10 is a view showing the result of improving a three-dimensional 3 mm thick, 3 mm spaced CT scan original with a 1 mm thick and 1 mm spaced CT scan.

The terms or words used in the present specification and claims should not be construed as being limited to their ordinary or dictionary meanings, and the inventor may properly define the concept of the term in order to best describe his invention. Based on the principle, it should be interpreted as meaning and concept consistent with the technical idea of the present invention.

Throughout the specification, when a part "includes" a certain element, it means that other elements may be further included, rather than excluding other elements, unless otherwise stated. In addition, terms such as “…unit”, “…group”, “module”, and “device” described in the specification mean a unit that processes at least one function or operation, which is hardware or software or a combination of hardware and software. can be implemented as

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

Here, before describing the embodiment of the present invention in earnest, the slice thickness and scan interval of the CT scan mentioned in the present invention will be briefly described first.

3 is a diagram illustrating an outline of a slice thickness and a scan interval of a CT scan.

Referring to FIG. 3 , the slice thickness refers to a length (ie, z-axis in a 3D image) included in one slice taken during a CT scan. Therefore, as the thickness increases, a wide range of information is included in one slice, and thus detailed expression is deteriorated.

The scan interval means the distance between the slice and the center of the slice. Usually, it has the same value as the slice thickness described above, and is photographed without an empty space between the slices.

3, (A) is a case of a CT scan with a slice thickness of 1 mm and a scan interval of 1 mm, (B) is a case of a CT scan with a slice thickness of 3 mm and a scan interval of 3 mm, (C) is a case of a slice thickness of 1 mm and a scan interval of 2 mm In the case of a CT scan, (D) is a case of a CT scan with a slice thickness of 3 mm and a scan interval of 6 mm.

4 is a view showing a slice thickness and a scan interval of a CT scan for a model of a hand.

Referring to FIG. 4 , in a CT scan of a model of a hand, (A) shows a case of a slice thickness of 1 mm and a scan interval of 1 mm, and (B) shows a case of a slice thickness of 3 mm and a scan interval of 3 mm.

5 is a diagram illustrating a slice thickness and a scan interval of a three-dimensional CT scan of the wrist joint.

Referring to FIG. 5 , in a 3D CT scan of the wrist joint, (A) shows a case of a slice thickness of 1 mm and a scan interval of 1 mm, and (B) shows a case of a slice thickness of 3 mm and a scan interval of 3 mm.

As can be seen from the image comparison of (A) and (B), it can be seen that the case of (A), which has a thinner slice thickness, shows a clearer and more detailed image than the case of (B).

Then, in the following, an embodiment of the present invention will be described in detail based on the concepts of the slice thickness and the scan interval of the CT scan as described above.

1 is a diagram schematically showing the configuration of a system for improving the quality of a CT scan using artificial intelligence according to an embodiment of the present invention.

Referring to FIG. 1 , the CT scan quality improvement system 100 using artificial intelligence according to the present invention includes an AI model generator 110 , a CT scan input unit 120 , an AI model application unit 130 , and a CT slice rearrangement. It is configured to include a unit 160 , a final CT scan acquisition unit 170 , and a control unit 180 .

The AI model generator 110 generates a thickness reduction AI (artificial intelligence) model 140 that reduces the thickness of a slice of a CT scan and an interval interpolation AI model 150 that interpolates an empty space between the slices. . Here, at least one of the thickness reduction AI model 140 and the interval interpolation AI model 150 generated by the AI model generator 110 is a Convolutional Neural Network (CNN)-based deep-learning model. can be

The CT scan input unit 120 receives a CT scan of a predetermined thickness and interval (eg, a slice thickness of 3 mm and a scan interval of 3 mm) that is a target for quality improvement (improvement).

The AI model application unit 130 applies the thickness reduction AI model 140 to each CT slice of the CT scan input through the CT scan input unit 120 to obtain a CT scan with reduced thickness, and An interval-interpolated CT scan is obtained by applying the interval interpolation AI model 150 to the CT scan. Here, when the AI model application unit 130 applies the interval interpolation AI model 150 to the CT scan with reduced thickness to obtain the interval interpolated CT scan, a plurality of different interval interpolation AI models are used. It is possible to obtain CT scans in which different intervals are interpolated, respectively. In addition, when the AI model application unit 130 applies the interval interpolation AI model 150 to the CT scan with reduced thickness to obtain the CT scan with the interval interpolated, before and after the CT scan with the reduced thickness A CT scan with interpolated intervals can be obtained by interpolating the interval by generating an empty slice between the front and rear slices using the slice as an input. The above will be described again later.

The CT slice reordering unit 160 rearranges the CT slices obtained by the AI model application unit 130 with reduced thickness and interpolated intervals.

The final CT scan acquisition unit 170 finally acquires a CT scan of a predetermined thickness and spacing with improved quality based on the CT slices rearranged by the CT slice realignment unit 160 .

The control unit 180 checks and operates the state of the AI model generator 110 , the CT scan input unit 120 , the AI model application unit 130 , the CT slice reordering unit 160 , and the final CT scan acquisition unit 170 . control, and the CT scan of a predetermined thickness and interval with improved quality acquired by the final CT scan acquisition unit 170 is displayed on the display screen and stored in the database (DB) 185).

Here, the CT scan quality improvement system 100 using artificial intelligence according to the present invention as described above is preferably a CT scan of a predetermined thickness and interval in which the quality acquired by the final CT scan acquisition unit 170 is improved. It may further include a communication unit 190 for transmitting the data of the external device and receiving data or commands from the external device.

In addition, the CT scan quality improvement system 100 using artificial intelligence as described above may be integrated into one computer system.

Hereinafter, a method for improving the quality of a CT scan using artificial intelligence based on the system 100 for improving the quality of a CT scan using artificial intelligence according to the present invention having the above configuration will be described.

2 is a flowchart illustrating an execution process of a method for improving the quality of a CT scan using artificial intelligence according to an embodiment of the present invention.

2, according to the method for improving the quality of a CT scan using artificial intelligence according to the present invention, first, a CT scan of a predetermined thickness and an interval by the CT scan input unit 120, for example, in FIGS. 6 and 7 As shown, a CT scan having a slice thickness of 3 mm and a scan interval of 3 mm is received, and the AI model applying unit 130 applies the thickness reduction AI model 140 to each CT slice of the CT scan (step S202). Here, before step S202, a thickness reduction AI (artificial intelligence) model 140 that reduces the thickness of a slice of a CT scan and an interval interpolation AI model 150 that interpolates an empty space between the slices. It may further include step S201. In this case, at least one of the thickness reduction AI model 140 and the interval interpolation AI model 150 may be configured as a Convolutional Neural Network (CNN)-based deep-learning model.

As described above, a CT scan (a CT scan with a slice thickness of 1 mm and a scan interval of 3 mm) having a reduced thickness by the application of the AI model 140 to reduce the thickness of the AI model application unit 130 is obtained (step S203).

Thereafter, the AI model application unit 130 applies the interval interpolation AI model 150 to the obtained CT scan with reduced thickness, as shown in FIGS. 6 and 8 , and the interval is interpolated CT scan (that is, A CT scan with a slice thickness of 1 mm and a scan interval of 1 mm) is acquired. Here, in obtaining an interval-interpolated CT scan by applying the interval interpolation AI model 150, a plurality of different interval interpolation AI models 150a and 150b are applied (steps S204 and S205), CT scans with each interpolated interval may be obtained (steps S206 and S207).

In addition, in obtaining the spaced interpolated CT scan by applying the interval interpolation AI model 150, as shown in FIG. 8, using the before and after slices of the CT scan with reduced thickness as input By interpolating the interval by creating an empty slice between the subsequent slices (refer to the figure at the upper right of FIG. 8), a CT scan with interpolated intervals can be obtained.

In this way, when the slice thickness is reduced and the interpolation of the scan interval is completed, the CT slice reordering unit 160 rearranges the reduced thickness and the interpolated CT slice (step S208).

Then, based on the rearranged CT slices by the final CT scan acquisition unit 170, a CT scan of a predetermined thickness and interval with improved quality (eg, a CT scan with a slice thickness of 1 mm and a scan interval of 1 mm) is finally acquired (step S209).

Meanwhile, FIG. 9 is a view showing that the thickness is reduced by applying the thickness reduction model and the interval interpolation model, and the final CT scan with improved quality is obtained by interpolating the interval.

Referring to FIG. 9 , this schematically shows a method for improving the quality of a CT scan using artificial intelligence according to the present invention described above, and as described above, a CT scan of a predetermined thickness and interval (eg, a slice thickness of 3 mm, a scan interval) When a CT scan of 3 mm) is input, the thickness of the slice is reduced from the initial 3 mm to 1 mm by applying the thickness reduction AI model 140 . Then, by applying a plurality of interval interpolation AI models 150a and 150b to the CT scan with the reduced thickness of the slice as described above, the intervals of different points (eg, +1 mm point, +2 mm point) are respectively After interpolating, realigning the reduced thickness and interpolated (thickness and spacing reduced and interpolated from the first 3 mm to 1 mm, respectively) CT slices, and finally based on the rearranged CT slices, a given thickness and A CT scan of the interval (eg, a CT scan with a slice thickness of 1 mm and a scan interval of 1 mm) is obtained.

FIG. 10 is a view showing the result of improving a three-dimensional 3 mm thick, 3 mm spaced CT scan original with a 1 mm thick and 1 mm spaced CT scan.

Referring to FIG. 10, (A) is a three-dimensional 3mm thick, 3mm spaced CT scan original, (B) is a 3mm thick, 3mm spaced CT scan original by reducing the thickness of the slice according to the method of the present invention, and scanning This is the result of interpolating the interval to improve the CT scan with a slice thickness of 1 mm and a scan interval of 1 mm.

As can be confirmed through the image comparison of (A) and (B), the case of (B) (the case of reducing the slice thickness and interpolating the scan interval according to the method of the present invention) is the case of (A) It can be seen that it shows a clearer and more detailed image than the case (original CT scan).

As described above, the system and method for improving the quality of a CT scan using artificial intelligence according to the present invention uses artificial intelligence technology to receive a low-resolution CT scan and perform thickness reduction and interval interpolation to obtain a high-resolution CT scan, It is possible to detect microscopic lesions that cannot be observed at low resolution, and thus has the effect of lowering the lesion overlook rate.

In addition, there is an advantage in that a high-quality CT scan can be obtained from a low-resolution CT scan without replacing the CT imaging equipment.

As mentioned above, although the present invention has been described in detail through preferred embodiments, the present invention is not limited thereto, and it is common knowledge in the art that various changes and applications can be made without departing from the technical spirit of the present invention. self-explanatory to the technician. Accordingly, the true protection scope of the present invention should be construed by the following claims, and all technical ideas within the equivalent scope should be construed as being included in the scope of the present invention.

100: (the present invention) CT scan quality improvement system using artificial intelligence
110: AI model generation unit 120: CT scan input unit
130: AI model application unit 140: thickness reduction AI model
150: interval interpolation AI model 160: CT slice rearrangement unit
170: final CT scan acquisition unit 180: control unit
185: database 190: communication department

Claims (10)

an AI model generator for generating a thickness reduction AI (artificial intelligence) model for reducing the thickness of a slice of a CT scan and an interval interpolation AI model for interpolating an empty space between the slices;
a CT scan input unit receiving a CT scan of a predetermined thickness and interval;
A CT scan with reduced thickness is obtained by applying the thickness reduction AI model to each CT slice of the CT scan input through the CT scan input unit, and the interval is interpolated by applying the interval interpolation AI model to the CT scan with reduced thickness. an AI model application unit that acquires a CT scan;
a CT slice reordering unit for rearranging the CT slices obtained by the AI model application unit with reduced thickness and interpolated intervals;
a final CT scan acquisition unit that finally acquires a CT scan of a predetermined thickness and interval with improved quality based on the CT slice rearranged by the CT slice realigning unit; and
Controls the state check and operation of the AI model generation unit, the CT scan input unit, the AI model application unit, the CT slice rearrangement unit, and the final CT scan acquisition unit, and a predetermined thickness obtained by the final CT scan acquisition unit has improved quality and Displaying the CT scan of the interval on the display screen while including a control unit for storing in the database (DB),
When the AI model application unit applies the interval interpolation AI model to the reduced-thickness CT scan to obtain a spaced-interpolated CT scan, the before and after slices of the reduced-thickness CT scan are used as inputs. A CT scan quality improvement system using artificial intelligence to obtain an interpolated CT scan with an interval by interpolating the interval by creating an empty slice between the slices.
The method of claim 1,
Transmitting the CT scan data of a predetermined thickness and interval with improved quality acquired by the final CT scan acquisition unit to an external device, and using artificial intelligence further comprising a communication unit for receiving data or commands from the external device CT scan quality improvement system.
The method of claim 1,
At least one of the thickness reduction AI model and the interval interpolation AI model generated by the AI model generator is a CNN (Convolutional Neural Network)-based deep-learning model. A CT scan quality improvement system using artificial intelligence.
The method of claim 1,
When the AI model application unit obtains the spaced interpolated CT scan by applying the spaced interpolation AI model to the reduced-thickness CT scan, a plurality of different spaced interpolation AI models are applied to each interpolated CT A system for improving the quality of CT scans using artificial intelligence to acquire scans.
delete a) receiving a CT scan of a predetermined thickness and interval and applying a thickness reduction AI model to each CT slice of the CT scan;
b) obtaining a reduced-thickness CT scan by application of the thickness-reduced AI model;
c) obtaining a spaced-interpolated CT scan by applying an interval interpolation AI model to the obtained CT scan with reduced thickness;
d) rearranging the CT slices reduced in thickness by step b) and interpolated at intervals by step c); and
e) finally acquiring a CT scan of a predetermined thickness and spacing with improved quality based on the rearranged CT slices,
In step c), when the interval interpolated CT scan is obtained by applying the interval interpolation AI model, an empty slice between the before and after slices is generated using the before and after slices of the reduced-thickness CT scan as input. A method of improving the quality of a CT scan using artificial intelligence to obtain an interpolated CT scan by interpolating the interval. 7. The method of claim 6,
The step a) further comprising the step of generating a thickness reduction AI (artificial intelligence) model that reduces the thickness of the slice of the CT scan prior to the step and an interval interpolation AI model that interpolates the empty space between the slices. A method for improving the quality of CT scans using
8. The method of claim 7,
At least one of the thickness reduction AI model and the interval interpolation AI model is a Convolutional Neural Network (CNN)-based deep-learning model.
7. The method of claim 6,
In the step C), in obtaining the spaced interpolated CT scan by applying the interval interpolation AI model, a plurality of different interval interpolation AI models are applied to obtain an interpolated CT scan with different intervals. A method for improving the quality of CT scans using
delete

Images