WO2022164215A1

WO2022164215A1 - Device and method for diagnostic test using exclusion of atypical foreign material from test result image of artificial intelligence-based diagnostic kit

Info

Publication number: WO2022164215A1
Application number: PCT/KR2022/001446
Authority: WO
Inventors: 김원태; 강신욱; 이명재; 김동민; 홍호탁; 김소연; 한아름
Original assignee: (주) 제이엘케이; 피씨엘 (주)
Priority date: 2021-01-28
Filing date: 2022-01-27
Publication date: 2022-08-04

Abstract

The present invention relates to a device and method for a diagnostic test using exclusion of an atypical foreign material from a test result image of an artificial intelligence-based diagnostic kit. The device for a diagnostic test using exclusion of an atypical foreign material from a test result image of an artificial intelligence-based diagnostic kit comprises: a spot detection unit which extracts a region of interest from a diagnostic kit test result image obtained from a screening device and detects spots on the extracted region of interest; a spot classification unit which classifies the spots into ring spots and general spots by using a deep learning model; a foreign material detection unit which detects a foreign material inside or outside the general spots by using a foreign material detection algorithm; a negative/positive reading unit which excludes the foreign material detected from the general spots and quantitatively quantifies the degree of positive reaction in the general spots so as to make a negative or positive determination; and a reading result output unit which displays a result value of the determination.

Description

A diagnostic test apparatus and method by excluding atypical foreign substances from the test result image of an AI-based diagnostic kit

The present invention relates to a diagnostic test apparatus and method by excluding atypical foreign substances from a diagnostic kit test result image. It relates to a diagnostic test apparatus and method through the exclusion of atypical foreign substances from the test result image of an AI-based diagnostic kit.

In Vitro Diagnostics (IVD) refers to medical devices used outside the body for the purpose of providing information such as diagnosis, prognosis, observation, and judgment of blood or histocompatibility using samples derived from the human body.

In general, the diagnostic kit test result image consists of a reference point and a spot, and as shown in FIG. 1 , according to the diagnostic test method of automatically reading the result by acquiring the diagnostic kit image of the prior art, sample collection, sample It proceeds through the process of input, diagnosis kit test result image acquisition, spot identification with the naked eye, and infection determination process.

However, since the existing diagnostic test determines whether or not an infection is present by the inspector's visual inspection, the negative/positive judgment is made only qualitatively, and the accuracy of the test occurs depending on the skill of the inspector. There is a problem in that time and precision are lowered because the judgment must be made by excluding them.

Therefore, since the accuracy of diagnostic results is as important as the speed of reading results in in vitro diagnostic devices, technology is needed to prevent errors from occurring even when a general examiner other than a skilled manpower, including disease test items that require quick test results, performs it. .

The present invention has been devised to solve the above-described problem, and according to the present invention, foreign substances present on the diagnostic kit test result image acquired through the image scanner of the screening equipment are automatically detected using image processing and deep learning, and By excluding the disease, it is intended to make a more rapid and precise determination of the negative or positive diagnosis of the disease.

The diagnostic test apparatus through the exclusion of atypical foreign substances from the AI-based diagnostic kit test result image according to an embodiment of the present invention for solving the above-mentioned problem extracts a region of interest from the diagnostic kit , a spot detector for detecting spots on the extracted ROI; a spot classifier for classifying the spots into ring spots and normal spots using a deep learning model; a foreign material detection unit that detects a foreign material inside or outside the general spot using a foreign material detection algorithm; a negative/positive reading unit that excludes foreign substances detected in the general spot and quantitatively quantifies the degree of a positive reaction in the general spot to determine negative or positive; and a read result output unit for displaying the determined result value.

According to another embodiment of the present invention, the spot classifier may detect a foreign substance inside or outside the general spot using a pre-trained convolutional neural network (CNN).

According to another embodiment of the present invention, the foreign material detection unit may detect a foreign material inside or outside the general spot by using a Maximally Stable External Region (MSER) algorithm.

According to another embodiment of the present invention, the negative/positive reading unit calculates the number of white pixels by distinguishing white pixels that have reacted with foreign substances in the area of the spot, and calculates the area of the white pixel in the area of the spot. It can be used to quantitatively quantify the degree of a positive reaction to determine whether it is negative or positive.

According to another embodiment of the present invention, the spot detection unit acquires a diagnostic kit test result image through an image scanner of a screening device, detects a reference point on the acquired diagnostic kit test result image, and the detected reference point By aligning the test result image of the diagnostic kit based on , a region of interest (ROI) is extracted, and the extracted Spots may be detected on the region of interest.

In the diagnostic test method by excluding atypical foreign substances from the AI-based diagnostic kit test result image according to an embodiment of the present invention, the spot detection unit extracts a region of interest from the diagnostic kit galsa result image obtained from the screening equipment, and the extracted a spot detection step of detecting spots on the region of interest; a spot classification step in which a spot classifier classifies the spots into ring spots and normal spots using a deep learning model; a foreign material detection step in which a foreign material detection unit detects a foreign material inside or outside the general spot using a foreign material detection algorithm; a negative/positive reading step of excluding foreign substances detected in the general spot by a negative/positive reading unit and quantitatively quantifying the degree of positive reaction in the general spot to determine negative or positive; and a read result output step in which the read result output unit displays the determined result value.

According to another embodiment of the present invention, in the spot classification step, the spot classification unit can detect foreign substances inside or outside the general spot using a pre-trained convolutional neural network (CNN). .

According to another embodiment of the present invention, in the foreign material detecting step, the foreign material detecting unit may detect a foreign material inside or outside the general spot using a Maximally Stable External Region (MSER) algorithm.

According to another embodiment of the present invention, in the negative/positive reading step, the negative/positive reading unit calculates the number of white pixels by classifying white pixels that have reacted with foreign substances in the area of the spot, and calculates the number of white pixels in the area of the spot. Negative or positive can be determined by quantitatively quantifying the degree of a positive reaction by using the area of the white pixel occupied by the .

According to another embodiment of the present invention, in the spot detection step, the spot detection unit acquires a diagnostic kit test result image through an image scanner of a screening device, and detects a reference point on the acquired diagnostic kit test result image, , aligning the diagnostic kit test result image based on the detected reference point to extract a region of interest (ROI), image denoising technique and blob detection algorithm can be used to detect spots on the extracted ROI.

According to the present invention, by using image processing and deep learning to automatically detect and exclude foreign substances present in the diagnostic kit test result image acquired through the image scanner of the screening equipment, the negative or positive diagnosis of the disease can be diagnosed more quickly and precisely can judge

1 is a view for explaining a diagnostic test method of acquiring an image of a diagnostic kit according to the prior art and reading the result.

2 is a diagram illustrating an image of a diagnostic kit test result according to an embodiment of the present invention.

3 is a view showing foreign substances in a spot of a diagnostic kit test result image according to an embodiment of the present invention.

FIG. 4 is a diagram for explaining a diagnostic test apparatus by excluding atypical foreign substances from an AI-based diagnostic kit test result image.

5 is a diagram quantitatively quantifying the degree of a positive reaction according to an embodiment of the present invention.

6 is a flowchart illustrating a diagnostic test method by excluding atypical foreign substances from an AI-based diagnostic kit test result image according to an embodiment of the present invention.

Since the present invention can apply various transformations and can have various embodiments, specific embodiments are illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the present invention to specific embodiments, and should be understood to include all modifications, equivalents, and substitutes included in the spirit and scope of the present invention.

However, in describing the embodiment, if it is determined that a detailed description of a related known function or configuration may unnecessarily obscure the gist of the present invention, a detailed description thereof will be omitted. In addition, the size of each component in the drawings may be exaggerated for explanation, and does not mean the size actually applied.

In addition, throughout the specification, when an element is referred to as "connected" or "connected" with another element, the one element may be directly connected or directly connected to the other element, but in particular It should be understood that, unless there is a description to the contrary, it may be connected or connected through another element in the middle. In addition, throughout the specification, when a part "includes" a certain component, this means that other components may be further included, rather than excluding other components, unless otherwise stated.

The present invention relates to HIV 1/2 (Human immunodeficiency virus 1/2), HCV (Hepatitis C virus) and HTLV 1/2 (Human T-lymphotropic virus 1/2) in plasma or serum using the principle of fluorescence immunoassay. Antibodies formed when infected, HBV surface antigen (HBsAg) formed when infected with HTLV 1/2 (Human T-lymphotropic virus 1/2) and HBV (Hepatitic B virus), HIV p24 antigen, HCV core antigen The purpose of this study is to acquire an image of the test result of the diagnostic kit for an in vitro diagnostic kit that simultaneously detects , and automatically detect foreign substances so that the disease can be diagnosed faster and more accurately than the existing diagnostic method.

FIG. 2 is a view showing a diagnostic kit test result image according to an embodiment of the present invention, and FIG. 3 is a view showing foreign substances in a spot of the diagnostic kit test result image according to an embodiment of the present invention.

In addition, FIG. 4 is a view for explaining a diagnostic test apparatus through the exclusion of atypical foreign substances from an AI-based diagnostic kit test result image.

Referring to FIG. 2 , the diagnostic kit test result image according to an embodiment of the present invention is configured to include a reference point 101 , a spot 102 , a foreign material 103 , and a foreign material 104 in the spot. At this time, as in FIG. 3 , when a foreign material enters the diagnostic kit in the process of waiting for the result reaction after collecting and inputting the sample, the diagnostic test apparatus according to the present invention excludes the foreign material 104 on the spot 102 and It can read negative or positive.

The diagnostic test device through the exclusion of atypical foreign substances from the AI-based diagnostic kit test result image according to an embodiment of the present invention is a device used in a medical facility, and can be operated in an environment where an external network is not connected. -Premise) form. In other words, considering that the external Internet connection is not possible due to the characteristics of medical facilities, it is built in an on-premise form. It can be designed to be used only in In addition, the spot detection device in the diagnostic kit result image may be composed of a computer terminal, a server, or a dedicated device, and a component providing each function may be composed of hardware or software.

Referring to FIG. 4 , the diagnostic test apparatus through the exclusion of atypical foreign substances from the AI-based diagnostic kit test result image according to an embodiment of the present invention is a spot detection unit 110 , a spot classification unit 120 , and a foreign material detection unit 130 . , a negative/positive reading unit 140 and a reading result output unit 150 .

The spot detection unit 110 extracts a region of interest from the result image of the diagnostic kit obtained from the screening device, and detects spots on the extracted region of interest.

More specifically, the spot detection unit 110 acquires a diagnostic kit test result image through an image scanner of a screening equipment, detects a reference point on the acquired diagnostic kit test result image, and then based on the detected reference point. By aligning the test result image of the diagnostic kit with can be configured to detect spots on the

The spot classifier 120 classifies the spots into ring spots and general spots using a deep learning model.

The shape of the spot on the diagnostic kit test result image is largely divided into two types. According to an embodiment of the present invention, the spot can be divided into a ring spot and a general spot.

In the case of the ring spot, as a defect generated in the process of forming the spot, it is excluded from the negative or positive judgment, and only the normal spot is judged as negative or positive.

The spot classifier 120 uses a pre-trained convolutional neural network (CNN), and the convolutional neural network (CNN) uses the spot patch images obtained from the spot detector 110 . The training data is constructed by separating the ring spot and the general spot, and learning is performed using the training data constructed in this way.

The foreign material detection unit 130 detects a foreign material inside or outside the general spot by using a foreign material detection algorithm.

More specifically, the foreign material detection unit 130 detects a foreign material by using a Maximally Stable External Region (MSER) algorithm and an image processing technique, and the MSER algorithm is a set of pixels having an intensity different from that of neighboring pixels. By detecting the area, it can be detected as a foreign substance.

The negative/positive reading unit 140 determines negative or positive by excluding foreign substances detected in the general spot and quantitatively quantifying the extent to which a positive reaction has occurred in the general spot.

The negative/positive reading unit 140 calculates the number of white pixels by classifying foreign substances and reacted white pixels in the spot area, and determines the degree of positive reaction using the area of the white pixels occupied in the spot area. It can be quantitatively quantified to determine negative or positive.

When a positive reaction occurs in the spot, it changes to white. At this time, the percentage of white in the spot area is calculated to quantitatively quantify the degree of reaction.

More specifically, as shown in FIG. 5 , the number of white pixels having a pixel value greater than or equal to a specified threshold is obtained in the spot area, and the area of the spot area is divided thereto to quantitatively quantify the degree of a positive reaction.

In this way, the negative/positive reading unit 140 finally determines negative/positive according to whether the quantified positive reaction value exceeds the positive determination reference value, and the reading result output unit 150 outputs the determined result value. indicate

Hereinafter, with reference to FIG. 6, a diagnostic test method by excluding atypical foreign substances from an AI-based diagnostic kit test result image according to an embodiment of the present invention will be described.

First, the spot detection unit extracts a region of interest from the result image of the diagnostic kit obtained from the screening equipment, and detects the spots on the extracted region of interest (S210).

In more detail, the spot detection unit acquires a diagnostic kit test result image through an image scanner of a screening device,

After detecting a reference point on the obtained diagnostic kit test result image, aligning the diagnostic kit test result image based on the detected reference point to extract a region of interest (ROI), the image Spots may be detected on the extracted ROI using an image denoising technique and a blob detection algorithm.

Thereafter, the spot classifier classifies the spots into ring spots and general spots using the deep learning model (S220).

In more detail, the spot classifier may detect foreign substances inside or outside the general spot using a pre-trained convolutional neural network (CNN).

Thereafter, the foreign material detection unit detects a foreign material inside or outside the general spot by using a foreign material detection algorithm (S230).

In this case, the foreign material detection unit may detect a foreign material inside or outside the general spot by using a Maximally Stable External Region (MSER) algorithm.

Thereafter, the negative/positive reading unit excludes foreign substances detected in the general spot, and quantitatively quantifies the degree of positive reaction in the general spot to determine negative or positive (S240).

More specifically, the negative/positive reading unit calculates the number of white pixels by distinguishing between foreign substances and reacted white pixels in the spot area, and uses the area of the white pixels in the spot area to determine the positive reaction. Negative or positive can be determined by quantitatively quantifying the degree.

Accordingly, the read result output unit may display the result value determined in this way (S250).

As described above, according to the present invention, by using image processing and deep learning to automatically detect and exclude foreign substances present on the diagnostic kit test result image obtained through the image scanner of the screening equipment, the negative or positive diagnosis of the disease can be more quickly and can be judged accurately.

In the detailed description of the present invention as described above, specific embodiments have been described. However, various modifications are possible without departing from the scope of the present invention. The technical spirit of the present invention should not be limited to the above-described embodiments of the present invention, and should be defined by the claims as well as the claims and equivalents.

Claims

a spot detector for extracting a region of interest from the diagnostic kit galsa result image obtained from the screening equipment, and detecting spots on the extracted region of interest;

a spot classifier for classifying the spots into ring spots and normal spots using a deep learning model;

a foreign material detection unit that detects a foreign material inside or outside the general spot using a foreign material detection algorithm;

a negative/positive reading unit that excludes foreign substances detected in the general spot and quantitatively quantifies the degree of a positive reaction in the general spot to determine negative or positive; and

a reading result output unit for displaying the determined result value;

A diagnostic test device through the exclusion of atypical foreign substances from the test result image of an AI-based diagnostic kit comprising a.
The method according to claim 1,

The spot classification unit,

A diagnostic test device through the exclusion of atypical foreign substances from an artificial intelligence-based diagnostic kit test result image that detects foreign substances inside or outside the general spot using a pre-learned convolutional neural network (CNN).
The method according to claim 1,

The foreign material detection unit,

A diagnostic test device by excluding atypical foreign substances from an artificial intelligence-based diagnostic kit test result image that detects foreign substances inside or outside the general spot by using the MSER (Maximally Stable External Region) algorithm.
The method according to claim 1,

The negative/positive reading unit,

In the spot area, the number of white pixels is calculated by dividing the foreign substances and the white pixels that have reacted, and the degree of positive reaction is quantitatively quantified using the area of the white pixel occupied in the spot area to determine negative or positive. , a diagnostic test device that excludes atypical foreign substances from the test result image of an AI-based diagnostic kit.
The method according to claim 1,

The spot detection unit,

Acquire the diagnostic kit test result image through the image scanner of the screening equipment,

Detecting a reference point on the obtained diagnostic kit test result image,

Extracting a region of interest (ROI) by aligning the diagnostic kit test result image based on the detected reference point,

A diagnostic test apparatus by excluding atypical foreign substances from an AI-based diagnostic kit test result image, which detects spots on the extracted region of interest using an image denoising technique and a blob detection algorithm.
a spot detection step of extracting, by a spot detection unit, a region of interest from a diagnostic kit galsa result image obtained from a screening device, and detecting spots on the extracted region of interest;

a spot classification step in which a spot classifier classifies the spots into ring spots and normal spots using a deep learning model;

a foreign material detection step in which a foreign material detection unit detects a foreign material inside or outside the general spot using a foreign material detection algorithm;

a negative/positive reading step of excluding foreign substances detected in the general spot by a negative/positive reading unit and quantitatively quantifying the degree of positive reaction in the general spot to determine negative or positive; and

a read result output step of displaying the determined result value by a read result output unit;

A diagnostic test method by excluding atypical foreign substances from the test result image of an AI-based diagnostic kit comprising a.
7. The method of claim 6,

The spot classification step is

The spot classification unit uses a pre-trained convolutional neural network (CNN) to detect foreign substances inside or outside the general spot. Way.
7. The method of claim 6,

The foreign matter detection step,

A diagnostic test method by excluding atypical foreign substances from an artificial intelligence-based diagnostic kit test result image, in which the foreign material detection unit detects a foreign material inside or outside the general spot using an MSER (Maximally Stable External Region) algorithm.
7. The method of claim 6,

The negative/positive reading step includes:

The negative/positive reading unit calculates the number of white pixels by distinguishing between foreign substances and reactive white pixels in the area of the spot, and quantitatively quantifies the degree of positive reaction using the area of the white pixel occupied in the area of the spot. A diagnostic test method by excluding atypical foreign substances from the test result image of an AI-based diagnostic kit that judges negative or positive.
7. The method of claim 6,

The spot detection step is

The spot detection unit,

Acquire the diagnostic kit test result image through the image scanner of the screening equipment,

Detecting a reference point on the obtained diagnostic kit test result image,

Extracting a region of interest (ROI) by aligning the diagnostic kit test result image based on the detected reference point,

A diagnostic test method by excluding atypical foreign substances from an AI-based diagnostic kit test result image, which detects spots on the extracted region of interest using an image denoising technique and a blob detection algorithm.