Methods and systems for separating biochemical molecules

Abstract

The present invention relates to a biochemical molecule separation method and system, the biochemical molecule separation method which may include the steps of: photographing a tissue sample and acquiring a tissue image in either a multi-channel fluorescence image or a visible light image; acquiring and analyzing channel-specific object detection results from the tissue image to extract target spatial information, or analyzing the tissue image using a trained neural network model to predict target spatial information; and varying the size and position of the separation area of ​​the biochemical molecule separation apparatus based on the target spatial information, and then physically separating and recovering the target from the tissue sample.

Description

Technical Field

[0001] The present invention relates to a biochemical molecule separation method and system that can automatically detect cells for spatial omics and then more quickly separate and recover them.

Background Art

[0002] Spatial omics profiling technology has recently made it possible to decipher gene molecules that are structurally related to pathology.

[0003] In tumor biology, a tumor is not a disease caused only by tumor cells, but a complex disease in which various cells such as immune cells, vascular structures, and epidermal cells form a community and develop the disease.

[0004] Techniques for analyzing tumors have evolved into spatial omics profiling technology in large-scale sequencing analysis and single-cell sequencing analysis, which has had a great impact on deciphering cancer mechanisms by raising questions about tumor heterogeneity, the tumor microenvironment, and spatial biomarkers.

[0005] Many spatial omics techniques focus on mapping the spatial omics landscape on a large scale, introducing spatially barcoded capture probes or fluorescently labeled target probes to determine the positions of gene molecules.

[0006] Although such techniques vary in the depth and scalability of information depending on the purpose of the spatial analysis technique, even at a relatively low depth, it is possible to discover the spatial heterogeneity and spatial landscape of the constituent cell types, and deeper omics information is required to effectively handle target molecules for treatment and diagnosis.

[0007] To meet the requirements, region of interest (ROI)-based spatial technology isolates the target region and applies chemistry for higher-scope omics data.

[0008] Conventional cell classifiers include FACS (Fluorescence-activated cell sorter), MACS (Magnetic-activated cell sorting), and LCM (laser capture microdissection).

[0009] FACS and MACS, respectively, designate cells of interest in a dissociated cell solution using fluorescent or magnetic particles, but have the disadvantage of losing spatial context before the cells are classified.

[0010] While LCM can classify ROIs while maintaining spatial context, it has the disadvantage of requiring additional preparation time for isolation, such as focusing a UV beam, to isolate the target, and can damage cells by dissecting them with a UV laser or by dissolving IR-activated polymers to make them protrude. [Overview of the project] [Problems that the invention aims to solve]

[0011] Therefore, in order to solve the aforementioned problems, the present invention aims to provide a biochemical molecule separation method and system that can automatically detect targets while preserving the spatial context of cells, and then separate and recover them more quickly.

[0012] The objects of the present invention are not limited to those mentioned above, and any further objects not mentioned can be clearly understood by a person with ordinary skill to the present invention from the following description. [Means for solving the problem]

[0013] As a means to solve the above problems, the biochemical molecule separation method of the biochemical molecule separation system according to the first embodiment of the present invention includes the steps of: photographing a tissue sample and acquiring a tissue image in one of two image formats, a multi-channel fluorescence image or a visible light image; acquiring and analyzing channel-specific object detection results from the tissue image to extract target spatial information, or analyzing the tissue image via a trained neural network model to predict target spatial information; and varying the separation area size and position of the biochemical molecule separation device based on the target spatial information, and then physically separating and recovering the target from the tissue sample.

[0014] Furthermore, the target is characterized by being one of the following: a cell, a group of two or more cells, a microregion containing a cell, ribonucleic acid (RNA), deoxyribonucleic acid (DNA), epigenome, epitranscriptome, metabolome, organelle, B cell receptor (BCR), T cell receptor (TCR), extracellular membrane protein, extracellular vesicle, carrier containing biochemical molecules, or hydrogel containing biochemical molecules.

[0015] The target spatial information may include at least one piece of information from among position, size, shape, density, and staining pattern.

[0016] The step of extracting target spatial information is characterized by including the steps of: acquiring a multi-channel fluorescence image; separating the fluorescence channels and then performing an object detection operation; comparing and analyzing the object detection results for each fluorescence channel and filtering the object detection results; and selecting at least one of the fluorescence channels considering the type of target, and extracting and providing target spatial information from the object detection results of the selected fluorescence channels.

[0017] The filtering step, when the target is a cell, is characterized by obtaining object detection results for each fluorescence channel based on at least one of location, morphology, size, density, and fluorescence intensity, and then filtering out duplicate and noise objects from the object detection results for each fluorescence channel to purify and extract only the target spatial information.

[0018] The neural network model is characterized by repeatedly learning a large amount of training data, each having sample images as input conditions and target detection probabilities as output conditions, in order to pre-learn the correlation between tissue images and target detection results.

[0019] The step of physically separating and recovering the target is characterized by physically separating the target by one of the following methods: separation using a laser, mechanical separation by micro-excision, separation using high-frequency ultrasound, chemical separation by adding specific chemical substances or enzymes, biological separation using antibodies specific to specific cell surface antigens, optical separation using light, or thermal ablation using high temperature.

[0020] As a means to solve the aforementioned problems, the biochemical molecular space designation method for a biochemical molecular separation system according to the second embodiment of the present invention is characterized by including the steps of: photographing a tissue sample and acquiring a tissue image in one of two image formats: a multi-channel fluorescence image or a visible light image; and acquiring and analyzing channel-specific object detection results from the tissue image to extract target spatial information, or analyzing the tissue image via a trained neural network model to acquire target spatial information and then providing it to an external party.

[0021] The aforementioned target is characterized by being one of the following: a cell, a group of two or more cells, a microregion containing a cell, ribonucleic acid (RNA), deoxyribonucleic acid (DNA), epigenome, epitranscriptome, metabolome, organelle, B cell receptor (BCR), T cell receptor (TCR), extracellular membrane protein, extracellular vesicle, carrier containing biochemical molecules, or hydrogel containing biochemical molecules.

[0022] The target spatial information may include at least one piece of information from among position, size, shape, density, and staining pattern.

[0023] The steps for extracting target spatial information include: acquiring a multi-channel fluorescence image; separating the fluorescence channels and then performing an object detection operation; comparing and analyzing the object detection results for each fluorescence channel and filtering the object detection results; and selecting at least one of the fluorescence channels considering the type of target, and extracting and providing target spatial information from the object detection results of the selected fluorescence channels.

[0024] The neural network model is characterized by repeatedly learning a large amount of training data, each having sample images as input conditions and target detection probabilities as output conditions, in order to pre-learn the correlation between tissue images and target detection results.

[0025] As a means for solving the above problems, a method for providing target space information for a biochemical molecule automatic separation system according to a third embodiment of the present invention includes: a step of acquiring a multi-channel fluorescence image; a step of performing an object detection operation after separating fluorescence channels; a step of comparing and analyzing object detection results for each fluorescence channel to filter out overlapping objects and noise objects; and a step of selecting any one of the fluorescence channels in consideration of the type of target, extracting target space information from the object detection results of the selected fluorescence channel, and providing it externally.

[0026] As a means for solving the above problems, a method for providing target space information for a biochemical molecule automatic separation system according to a fourth embodiment of the present invention includes: a step of pre-training a neural network model using a large number of learning data having a sample image as an input condition and a target matching probability as an output condition; a step of photographing a tissue sample and acquiring a tissue image in any one of the forms of a multi-channel fluorescence image and a visible light image; and a step of analyzing the tissue image through the pre-trained neural network model to extract target space information and providing it externally.

[0027] As a means for solving the above problems, a biochemical molecule automatic separation system according to a fifth embodiment of the present invention includes: a target automatic detection device that receives a tissue image acquired in any one of the forms of a multi-channel fluorescence image and a visible light image, acquires and analyzes object detection results for each channel from the tissue image to extract target space information, or analyzes the tissue image through a pre-trained neural network model to acquire target space information; and a biochemical molecule separation device that acquires the tissue image and provides it to the target automatic detection device, and then physically separates and recovers the target from the tissue sample after varying the size and position of the separation region based on the target space information fed back from the target automatic detection device.

Advantages of the Invention

[0028] The present invention can automatically detect target space information using image processing technology and perform a target separation operation based on this. As a result, while preserving the spatial context of the target, the time required for target detection and separation can be significantly reduced.

Brief Description of Drawings

[0029] [Figure 1] It is a diagram showing a biochemical molecule automatic separation system for spatial omics according to an embodiment of the present invention. [Figure 2] It is a diagram for explaining a biochemical molecule separation device according to an embodiment of the present invention. [Figure 3] It is a diagram for explaining a fluorescence image-based target detection method according to an embodiment of the present invention. [Figure 4] It is a diagram for explaining an object detection stage for each channel according to an embodiment of the present invention. <​​​​​​​​​​​​​​​​​​​​​​​​​​Hereinafter, preferred embodiments will be described in detail with reference to the attached drawings so that a person with ordinary skill in the art to which the present invention pertains can easily implement the present invention. However, in describing preferred embodiments of the present invention in detail, if it is determined that a specific description of a related known function or configuration would obscure the gist of the present invention, such detailed description will be omitted. Furthermore, the same reference numerals will be used throughout the drawings for parts that have similar functions and operations.

[0031] In the specification as a whole, when any part is said to be "connected" to other parts, this includes not only cases where they are "directly connected" but also cases where they are "indirectly connected" with other elements in between. Furthermore, when any component is said to "include," this means that, unless otherwise stated, other components may be included rather than excluded.

[0032] Figures 1 and 2 show an automated biochemical molecule separation system for spatial omics according to one embodiment of the present invention.

[0033] Referring to Figures 1 and 2, the apparatus of the present invention consists of a target automatic detection device 100 that receives a tissue image acquired in either a multi-channel fluorescence image or a visible light image, and extracts spatial information of the target by acquiring and analyzing channel-specific object detection results from the tissue image, or by analyzing the tissue image using a trained neural network model to acquire spatial information of the target; and a biochemical molecule separation device 200 that acquires a tissue image and provides it to the target automatic detection device 100, then varies the size and position of the separation region based on the spatial information of the target fed back from the target automatic detection device 100, and then physically separates and recovers the target from the tissue sample.

[0034] The target automatic detection device 100 includes a target detection unit 110 based on a fluorescence image and a target detection unit 120 based on a neural network, and can automatically extract spatial information of the target in two ways depending on whether or not the tissue sample is stained.

[0035] The target detection unit 110 of the fluorescence image substrate extracts spatial information based on a stained tissue sample. This unit acquires a multi-channel fluorescence image of the tissue sample, obtains object detection results for each channel, and then compares and analyzes the object detection results for each channel to filter the detected objects. Furthermore, it selects and uses one of the object detection results for each channel depending on the type of target to extract and provide spatial information of the target.

[0036] The neural network-based target detection unit 120 extracts spatial information based on unstained tissue samples, utilizing a neural network model that has learned the correlation between tissue images and target detection results. That is, after acquiring tissue images, the neural network model is used to analyze the images and predict and provide spatial information of the target.

[0037] In this context, the target may be any one of the following: cells, ribonucleic acid (RNA), deoxyribonucleic acid (DNA), epigenome, epitranscriptome, metabolome, organelles, B cell receptors (BCRs), T cell receptors (TCRs), extracellular membrane proteins, or extracellular vesicles, but is not limited to these.

[0038] Furthermore, the spatial information of the target may include, but is not limited to, location, size, shape, density, and staining pattern.

[0039] The biochemical molecule separation apparatus 200 consists of an optical module 210, a mechanical module 220, and a processor 230.

[0040] The optical module 210 includes a CCD camera 211 for imaging tissue samples, a fluorescence light source 212 for irradiating the laser light path with a fluorescence light source, a filter cube array 213 containing multiple fluorescence filters to broaden the range of the stained sample, a target light source 214 for illuminating the upper side of the target area, a light source 215 for illuminating the lower side of the target area, an Nd:YAG nanosecond pulsed laser source 216 for irradiating an NIR laser (λ=1064nm, pulse=6 nanoseconds) to excise the target via a fixed laser light path, a spot size modulator 217 for controlling the laser spot size by adjusting the slit size and objective lens magnification by the x and y axes, and an objective lens 218 having a long working distance to focus the laser spot and provide sufficient working space to classify the tissue sample.

[0041] The machine module 220 consists of two motorized stages with sub-micrometer precision: a sample stage 221 for positioning and fixing the ITO glass on which the tissue sample is placed, and a recovery stage 222 for arranging a large number of PCR tubes used as wells and recovering cells separated from the tissue sample.

[0042] In this scenario, under the assumption that the laser path is fixed, the sample stage 221 focuses the image in real time through z-axis control and aims at the target through x-axis and y-axis control.

[0043] If the tissue sample is stained, the processor 230 acquires a multi-channel fluorescence image through the filter cube array 213 and obtains spatial information of the target via the target detection unit 110 of the fluorescence image substrate. If the tissue sample is not stained, the processor acquires a visible light image without using the filter cube array 213 and obtains spatial information of the target via the target detection unit 120 of the neural network substrate.

[0044] Furthermore, once spatial information of the target is acquired, the laser spot size is controlled through the slit size of the spot size modulator 217 and the magnification of the objective lens 218 according to the size of the target, and the position of the sample stage 221 is controlled according to the position of the target so that the laser can be irradiated only onto the target.

[0045] Furthermore, the recovery stage 222 is positioned to allow the targets separated from the tissue sample to be separated and recovered through each well of the recovery stage 222.

[0046] In other words, the present invention minimizes the time and effort required for target separation by automatically performing both the operation of extracting spatial information of the target and the operation of separating the target by irradiating it with a laser based on the spatial information.

[0047] In addition to the above, although the explanation above focused only on the physical separation of the target from the tissue sample using a laser, it goes without saying that a variety of other methods can be applied as needed, such as mechanical separation by micro-excision, separation using high-frequency ultrasound, chemical separation by adding specific chemical substances or enzymes, biological separation using antibodies specific to specific cell surface antigens, optical separation using light, and thermal ablation using high temperatures.

[0048] The following describes in more detail how the automated biochemical molecule separation system of the present invention detects a target and acquires and provides its spatial information.

[0049] Figures 3 to 7 illustrate a method for detecting a target on a fluorescent image substrate according to one embodiment of the present invention.

[0050] As shown in Figure 3, the method for acquiring spatial information on a fluorescence image substrate according to the present invention consists of a multi-channel fluorescence image acquisition step S11, a channel-specific object detection step S12, an object detection result filtering step S13, and a target extraction and spatial information provision step S14.

[0051] For the sake of explanation, the following example will describe a multi-channel fluorescence image composed of DAPI, FITC, and Cy5 fluorescence channels, where cancer cells are targeted for detection among nuclei, leukocytes, and cancer cells.

[0052] Multi-channel fluorescence image acquisition stage S11

[0053] Tissue samples are stained using known staining methods such as H&E staining and immunohistochemistry, and then multichannel fluorescence images with DAPI, FITC, and Cy5 fluorescence channels are obtained through a filter cube array 213.

[0054] Channel-specific object detection stage S12

[0055] After separating the multi-channel fluorescence image into images for each of the DAPI, FITC, and Cy5 fluorescence channels as shown in Figure 4, each fluorescence channel image is converted to a grayscale image (Step 1).

[0056] Furthermore, using the open-source tool CellProfiler, nuclei are detected in DAPI channel images, and potential primary objects based on size, morphology, and hue are detected in FITC channel images and Cy5 channel images, respectively (Step 2).

[0057] Furthermore, considering that cell lines such as leukocytes and cancer cells have the characteristic of having a nucleus, the nucleus detection results of the DAPI channel image and the primary object detection results of the FITC channel image are compared to detect only primary objects located in the same position as the nucleus as leukocytes. Similarly, using the same principle, the nucleus detection results of the DAPI channel image and the primary object detection results of the Cy5 channel image are compared to detect only primary objects located in the same position as the nucleus as cancer cells (Step 3).

[0058] Object detection result filtering stage S13

[0059] However, imaging certain fluorescence channels presents a problem: different cell lines may overlap in the images. For example, unlike the Cy5 channel, which only contains fluorescent phosphodes attached to cancer cells, the FITC channel has the problem of containing fluorescent phosphodes attached not only to leukocytes (especially anti-CD45) but also to cancer cells. The FITC channel with an emission wavelength of 530 nm is known to overlap with the spontaneous fluorescence emitted from cells.

[0060] Therefore, in this invention, as shown in Figure 5, the cancer cell detection results of the Cy5 channel are excluded from the leukocyte detection results of the FITC channel image, so that only actual leukocytes remain in the FITC channel image.

[0061] Target extraction and spatial information provision stage S14

[0062] Depending on the type of target, a fluorescence channel image to be used to acquire target spatial information is selected, and the position, size, and morphology of each object present in the fluorescence channel are converted into spatial information for each target and provided to the biochemical molecule separation device 200.

[0063] Specifically, if the target is a white blood cell, spatial information of the target is obtained based on the white blood cell detection result of the FITC channel image and provided to the biochemical molecular separation device 200. If the target is a cancer cell, spatial information of the target is obtained based on the cancer cell detection result of the Cy5 channel image and provided to the biochemical molecular separation device 200.

[0064] Next, the biochemical molecule separation device 200 adjusts the laser irradiation size and position based on the spatial information of the target to separate the target present in the tissue sample, and then separates and recovers it through a number of wells.

[0065] Figure 6 shows the target separation results obtained using a biochemical molecular separation device. Referring to this figure, it can be confirmed that only the target (e.g., cancer cells) was selectively separated and recovered from the tissue sample.

[0066] In addition, a key feature is that the size of the fluorescence image obtained through a single image acquisition operation is extremely small compared to the overall size of the tissue sample.

[0067] In this invention, as shown in Figure 7, a large number of fluorescence images are acquired by continuously scanning the entire tissue sample, and then stitching these images together into a single image. This allows for the acquisition of fluorescence images of a wide area corresponding to the entire tissue sample. However, considering that the illumination may change during image acquisition, illumination uniformity is performed along with the image stitching.

[0068] Furthermore, by using fluorescence images covering a wide area for target detection, a wider area can be analyzed more quickly.

[0069] Figures 8 to 11 illustrate a neural network-based target detection method according to another embodiment of the present invention.

[0070] As shown in Figure 8, the neural network-based spatial information acquisition method of the present invention consists of a training data acquisition step S21, a neural network model training step S22, and a tissue image analysis step S23 of the neural network model.

[0071] Training data acquisition stage S21

[0072] First, a large number of microparticles corresponding to the target are generated. In this process, the microparticles can be obtained through the following steps, as shown in Figure 9: (step 1) generating a liquid mixture of prepolymer and photoinitator; (step 2) filling a tube array with the liquid mixture of prepolymer and photoinitator, covering it with a photomask having homogeneous codes, and irradiating it with ultraviolet light to generate homogeneous microparticles; (step 3) curing the microparticles and removing them from the tube array; (step 4) washing the microparticles; and (step 5) recovering the washed microparticles. However, the process is not limited to these steps.

[0073] Furthermore, the generated microparticles are irregularly dispersed on a glass slide and then photographed to obtain a large number of sample images. Each of these sample images is flipped and rotated as shown in Figure 10 to increase the number of sample images used to generate training data.

[0074] Based on a large number of augmented sample images, a large volume of training data is generated, with sample images as input conditions and target match probability as output conditions. In this process, 70% of the training data consists of tissue images as the training set, 15% as the validation set, and 15% as the test set, but it is not limited to these.

[0075] Neural network model training stage S22

[0076] The neural network model of the present invention may be implemented as a DNN (Deep Neural Network), CNN (Convolutional Neural Network), RNN (Recurrent Neural Network), etc., and as shown in Figure 11, it comprises an input layer, a hidden layer, and an output layer, and learns the target detection probability corresponding to the input image using a large amount of training data.

[0077] Furthermore, the classification results and prediction results using the neural network model are compared to calculate the prediction accuracy. If the prediction accuracy exceeds the target value, the neural network learning process is terminated.

[0078] Target extraction and spatial information provision stage S23 via neural network model

[0079] Once the neural network model has completed its training, the tissue sample is mounted on an ITO glass slide and then placed on the sample stage 221 of the biochemical molecule separation device 200. At this point, the tissue sample is illuminated via the target light source 214 and the light source 215, while a general image of the tissue sample is acquired via the CCD camera 211.

[0080] Furthermore, using a pre-trained neural network model, tissue images are detected to find areas where the probability of matching with the target is greater than or equal to a pre-set value. The position, size, and morphology of each of these image areas are then acquired and provided as spatial information of the target.

[0081] In the explanation above, training data was generated by directly manufacturing microparticles corresponding to the target. However, if necessary, target images can also be obtained from medical images captured via various medical imaging devices, and training data can be generated from these images.

[0082] Furthermore, target images can be acquired through multi-channel fluorescence imaging, and training data can be generated from these images.

[0083] However, in such cases, it is preferable that, without a process to confirm whether or not the tissue sample is stained, the system can respond to the operator's request by assuming that the tissue sample is stained by performing a target detection operation based on a neural network.

[0084] Although preferred embodiments of the present invention have been illustrated and described above, the present invention is not limited to the specific embodiments described above, and various modifications can be made by a person with ordinary skill in the art to which the invention belongs without departing from the gist of the invention claimed in the claims. Such modifications should not be understood individually from the technical idea or prospects of the present invention.

Claims

1. In a biochemical molecular separation method for a biochemical molecular separation system, The process involves photographing a tissue sample and acquiring a tissue image in either a multi-channel fluorescence image or a visible light image, The steps include: obtaining and analyzing channel-specific object detection results from the tissue image to extract target spatial information, or analyzing the tissue image using a trained neural network model to predict target spatial information; A biochemical molecule separation method comprising the steps of: varying the size and position of the separation area of ​​a biochemical molecule separation apparatus based on the target spatial information, and then physically separating and recovering the target from the tissue sample.

2. The aforementioned target is The biochemical molecule separation method according to claim 1, characterized in that the material is one of the following: a cell, a group of two or more cells, a microregion containing a cell, ribonucleic acid (RNA), deoxyribonucleic acid (DNA), epigenome, epitranscriptome, metabolome, organelle, B cell receptor (BCR), T cell receptor (TCR), extracellular membrane protein, extracellular vesicle, carrier containing a biochemical molecule, or hydrogel containing a biochemical molecule.

3. The aforementioned target spatial information is, The biochemical molecule separation method according to claim 1, characterized in that it can include at least one piece of information among position, size, morphology, density, and staining pattern.

4. The step of extracting the target spatial information is as follows: The step of acquiring a multi-channel fluorescence image, After separating the fluorescence channels, the object detection operation is performed. The process involves comparing and analyzing object detection results for each fluorescence channel and filtering the object detection results. The biochemical molecule separation method according to claim 1, comprising the steps of selecting at least one of the fluorescent channels considering the type of target, and extracting and providing target spatial information from the object detection results of the selected fluorescent channels.

5. The filtering step described above is The automated cell separation method according to claim 4, characterized in that, when the target is a cell, object detection results for each fluorescence channel are obtained based on at least one of location, morphology, size, density, and fluorescence intensity, and then duplicate objects and noise objects are filtered from the object detection results for each fluorescence channel to purify and extract only the target spatial information.

6. The aforementioned neural network model is The biochemical molecule separation method according to claim 1, characterized in that a large number of training data having sample images as input conditions and target detection probability as output conditions are repeatedly trained to pre-learn the correlation between tissue images and target detection results.

7. The step of physically separating and recovering the target is: The automated cell separation method according to claim 1, characterized in that the target is physically separated by one of the following methods: separation using a laser, mechanical separation by micro-excision, separation using high-frequency ultrasound, chemical separation by adding a specific chemical substance or enzyme, biological separation using an antibody specific to a specific cell surface antigen, optical separation using light, or thermal ablation using high temperature.

8. In a method for specifying the biochemical molecular space in a biochemical molecular separation system, The process involves photographing a tissue sample and acquiring a tissue image in either a multi-channel fluorescence image or a visible light image, A method for specifying a biochemical molecular space, characterized by comprising the steps of: obtaining and analyzing channel-specific object detection results from the tissue image to extract target spatial information; or analyzing the tissue image using a trained neural network model to obtain target spatial information and then providing it to an external source.

9. The aforementioned target is The biochemical molecule separation method according to claim 8, characterized in that the material is one of the following: a cell, a group of two or more cells, a microregion containing a cell, ribonucleic acid (RNA), deoxyribonucleic acid (DNA), epigenome, epitranscriptome, metabolome, organelle, B cell receptor (BCR), T cell receptor (TCR), extracellular membrane protein, extracellular vesicle, carrier containing a biochemical molecule, or hydrogel containing a biochemical molecule.

10. The aforementioned target spatial information is, The biochemical molecule separation method according to claim 8, characterized in that it includes at least one piece of information among location, size, morphology, density, and staining pattern.

11. The step of extracting the target spatial information is as follows: The step of acquiring a multi-channel fluorescence image, After separating the fluorescence channels, the object detection operation is performed. The process involves comparing and analyzing object detection results for each fluorescence channel and filtering the object detection results. The biochemical molecule separation method according to claim 8, comprising the steps of selecting at least one of the fluorescent channels considering the type of target, and extracting and providing target spatial information from the object detection results of the selected fluorescent channels.

12. The aforementioned neural network model is The biochemical molecule separation method according to claim 8, characterized in that a large number of training data having sample images as input conditions and target detection probability as output conditions are repeatedly trained to pre-learn the correlation between tissue images and target detection results.

13. In a method for providing target spatial information for an automated biochemical molecule separation system, The step of acquiring a multi-channel fluorescence image, After separating the fluorescence channels, the object detection operation is performed. The process involves comparing and analyzing object detection results for each fluorescence channel to filter out duplicate and noise objects, and A method for providing target spatial information, which includes the steps of: selecting one of the fluorescence channels considering the type of target; extracting target spatial information from the object detection result of the selected fluorescence channel; and providing it to an external party.

14. In a method for providing target spatial information for an automated biochemical molecule separation system, The process involves pre-training a neural network model using a large number of training data sets, each having sample images as input and target matching probability as output. The process involves photographing a tissue sample and acquiring a tissue image in either a multi-channel fluorescence image or a visible light image, A method for providing target spatial information, comprising the steps of: analyzing the aforementioned organizational image using the trained neural network model to extract target spatial information and providing it to an external party.

15. An automated target detection device that receives a tissue image acquired in either a multi-channel fluorescence image or a visible light image as input, and extracts target spatial information by capturing and analyzing channel-specific object detection results from the tissue image, or by analyzing the tissue image using a trained neural network model to acquire target spatial information. A biochemical molecule automated separation system comprising: a biochemical molecule separation device that acquires the tissue image and provides it to the target automated detection device, then varies the size and position of the separation region based on target spatial information fed back from the target automated detection device, and then physically separates and recovers the target from the tissue sample.

16. A computer-readable recording medium that stores a program for performing the automated cell separation method described in any one of claims 1 to 9.

Images