Method for determining normal retinal tissue region in cell aggregate, and method for producing retinal tissue implantable in animals

Abstract

Provided is a method for quickly, objectively and properly evaluating the quality of a cell aggregate by determining the quality of the cell aggregate by determining an evaluation score on the basis of a determination criterion from an appropriate point of view with respect to an image of the cell aggregate.　With respect to a cell aggregate which can include a retinal tissue derived from a pluripotent stem cell, in an image region of the cell aggregate which is each of an inverted observation image of the cell aggregate and an elected observation image of the cell aggregate, a region that is determined as satisfying a color-related candidate region determination criterion for each of the inverted observation image and the elected observation image on the basis of the candidate region determination criterion is identified as a candidate region included in a region that is a candidate for a normal retinal tissue region. Subsequently, in the candidate region of each of the inverted observation image and the elected observation image, a candidate region that is determined as satisfying a brightness- and / or shape-related normal region determination criterion for each of the inverted observation image and the elected observation image is determined as the normal retinal tissue region.

Description

Method for determining normal retinal tissue area in cell aggregates and method for producing retinal tissue that can be transplanted into animals

[0001] The present disclosure relates to methods for determining normal retinal tissue regions in cell aggregates that can be transplanted into animals, and methods for producing retinal tissue that can be transplanted into animals. The present disclosure also relates to methods for determining tissue regions in animals other than retinal tissue.

[0002] Various studies have been conducted with the aim of neural retinal tissue transplantation therapy in humans. As a result, methods for producing cell aggregates containing neural retinal tissue (three-dimensional tissue) have been improved, making it possible to produce high-quality neural retinal tissue with high efficiency. However, neural retinal tissue is a highly complex tissue, and due to the nature of artificial cell differentiation in vitro, some non-target cells may be induced. Therefore, removing cell aggregates containing non-target cells and excising normal regions of cell aggregates free of non-target cells for transplantation are very important processes for quality control.

[0003] In recent years, methods for evaluating the quality of neural retina for transplantation have been developed. For example, one method for evaluating the quality of neural retina for transplantation involves extracting part or all of a cell aggregate containing a neural retina with an epithelial structure derived from pluripotent stem cells as a quality assessment sample, detecting the expression of neural retinal cell-related genes and non-neural retinal cell-related genes in the quality assessment sample, and determining that the neural retina (neural retina for transplantation) in a cell aggregate from the same or the same lot as the cell aggregate containing the quality assessment sample is suitable for use as a neural retina for transplantation if expression of neural retinal cell-related genes is detected but expression of non-neural retinal cell-related genes, including one or more genes selected from the group consisting of cerebrospinal tissue marker genes and ocular tissue marker genes, is not detected (Patent Document 1). This method allows the quality of the neural retina for transplantation to be evaluated by measuring gene expression levels. However, this method requires the use of a PCR device and is not suitable for assessing the quality of large quantities of cell aggregates. Therefore, a method that utilizes a microscope or the like to appropriately assess the quality of large quantities of cell aggregates has been desired.

[0004] International Publication No. 2020 / 184720

[0005] The three-dimensional tissues (organoids) contained in cell aggregates generally have irregular shapes, so appropriate techniques must be used when observing them under a microscope. Typically, multiple small epithelial tissues exist within a single three-dimensional tissue. Each three-dimensional tissue contains several pieces of epithelial tissue, and the size of each piece, the angle at which the epithelial tissue is attached, and other aspects of the structure are completely different for each individual three-dimensional tissue.

[0006] The three-dimensional tissues contained in cell aggregates vary in shape from individual to individual. Typically, neural cells, including the neural retina, form tightly bound, aligned, and stratified neuroepithelial structures. Neuroepithelium is composed of closely related cells. This continuous mass of neuroepithelium is called "epithelial tissue" or "lobe." Cell aggregates often contain multiple lobes. Figure 1 shows an upright image of a cell aggregate containing three lobes. For example, the cell aggregate shown in Figure 1 has three epithelial tissues, or lobes, but they vary in size. In other words, Figure 1 shows three lobes of different sizes: epithelial tissue ET1, epithelial tissue ET2, and epithelial tissue ET3. Due to the effect of gravity on the cell aggregate, the cell aggregate as a whole is roughly elliptical when viewed vertically and has a convex lens-like shape with a thick center when viewed horizontally, with each lobe essentially spreading horizontally. Because each lobe has a different thickness, the focal point of the microscope within each lobe also differs.

[0007] However, epithelial tissue has characteristics that make it difficult to observe under a microscope. First, when creating a three-dimensional retina from iPS cells, non-target cells (non-target tissues) may also be produced in addition to the retinal tissue. Furthermore, in commercial production, it is common to simultaneously create, for example, 30 to 30,000 three-dimensional tissues in a single differentiation culture. Furthermore, the size and state of the epithelial tissue differ spatially and three-dimensionally for each three-dimensional tissue created. Therefore, even if three-dimensional retinas are created by differentiating iPS cells under the same differentiation conditions, subtle differences in quality will occur between differentiation lots (batches).

[0008] As a method for observing and evaluating the quality of a stereoscopic retina, inspection using an optical microscope, which is suitable for large-scale observation, is often performed. Microscopic inspection techniques include upright observation using an upright microscope and inverted observation using an inverted microscope. Observation of a stereoscopic retina is often performed using only inverted observation. However, when using either upright or inverted observation, there is a possibility that undetectable cells exist behind the stereoscopic retina, which can lead to overlooking the undetectable cells and resulting in an incorrect quality evaluation. Thus, there are problems with quality evaluation using either upright observation images captured using upright observation with an upright microscope or inverted observation images captured using inverted observation with an inverted microscope.

[0009] Furthermore, quality assessment is based on observation using an optical microscope, and because the cell aggregates being observed have a three-dimensional shape, focusing on one area means that other areas will be out of focus. Furthermore, if there is a subtle shadow in the image, it may not be possible to clearly determine whether it is an unintended cell or a shadow caused by factors such as the way the light hits it.

[0010] In order to reliably observe cell aggregates, it is necessary to take measures to address the above-mentioned undesirable characteristics, but there are technical limitations to doing so. For example, common sense would suggest the following countermeasures, but all of them are extremely difficult to implement in practice due to technical limitations.

[0011] First, a method of removing cell aggregates from the culture medium and observing and evaluating them in air is considered. This method would allow for more flexible observation from any direction because the observation is performed in air. However, removing cell aggregates from the culture medium would dry out and cause damage, which would have a negative impact on quality, so this method cannot be adopted.

[0012] Another possible approach is to move cell aggregates in culture under a microscope and observe them spatially. However, because cell aggregates are soft and vulnerable to physical damage, moving them can damage them, and the manipulation required for observation is extremely time-consuming, making this approach unrealistic.

[0013] Furthermore, to allow for more flexible observation from different directions, one method of observing and evaluating cell aggregates by rotating the entire container is conceivable. However, in reality, rotating the entire container does not allow for proper observation. This is because ordinary culture plates are not sealed, and rotating the container causes the culture medium to leak out. Furthermore, while multi-well plates allow for observation from above and below, observation from the side is not possible because other wells get in the way.

[0014] Furthermore, to observe the cell aggregates from the back, it is conceivable to rotate the cell aggregates. However, because cell aggregates are elastic and their centers of gravity are offset due to individual differences, it is difficult to physically rotate the cell aggregates, even if the container is rotated. While it is conceivable to rotate the cell aggregates by gripping them with tweezers or other tools, this is not practical due to the risk of damaging the cells and tissues. In particular, the surface of the three-dimensional retina is easily damaged, making it virtually impossible to rotate it using tweezers or other tools.

[0015] As described above, the commonly accepted methods described above do not allow sufficient observation of three-dimensional tissues under a microscope, and the quality of large quantities of three-dimensional tissues cannot be efficiently observed, making automation and speeding up difficult. In fact, no research institution has yet established a method for evaluating the quality of neural retinal tissues at a clinical level. Furthermore, there is no technology available for evaluating large quantities of neural retinal tissues using both upright and inverted microscopes.

[0016] Furthermore, there has been no technology to evaluate the quality of images of cell aggregates taken with upright and inverted microscopes through information processing. There is also no method to objectively assess quality by calculating an evaluation score based on appropriate criteria for the image. Furthermore, there is no method to set criteria through machine learning and objectively assess quality by integrating various image features.

[0017] The present invention has been made in consideration of the above-mentioned problems, and provides a method for quickly, objectively, and appropriately evaluating the quality of cell aggregates by calculating an evaluation score based on judgment criteria from an appropriate perspective for images of cell aggregates taken with an upright microscope and an inverted microscope.

[0018] The present invention has been made in view of the above-mentioned problems and solves the problems by adopting the following characteristic configuration: That is, the present invention provides a method for determining a normal retinal tissue region transplantable to an animal in a cell aggregate such as a continuous epithelial tissue that may include retinal tissue derived from pluripotent stem cells, comprising: an image data acquisition step of acquiring image data of an inverted observation image and an upright observation image of the cell aggregate by photographing the cell aggregate using an inverted microscope and an upright microscope; a candidate region identification step of identifying, in each image region of the cell aggregate in the inverted observation image and the upright observation image, a region that is determined to satisfy the corresponding candidate region criteria related to color, based on the corresponding candidate region criteria for the inverted observation image and the upright observation image, as a candidate region included in a region that is a candidate for a normal retinal tissue region; and a normal region determination step of determining, in each candidate region in the inverted observation image and the upright observation image, a candidate region that is determined to satisfy the corresponding normal region criteria, based on the corresponding normal region criteria related to brightness and / or morphology, based on the corresponding normal region determination criteria for the inverted observation image and the upright observation image, as a normal retinal tissue region.

[0019] In the present invention, the inverted observation image and the upright observation image are positionally correlated so that the positions in each image correspond to the same positions of the cell aggregate when viewed from the vertical direction, and the normal region determination step can be such that, in each candidate region of the inverted observation image and the upright observation image, candidate regions that are determined to simultaneously satisfy the corresponding normal region determination criteria for brightness and shape corresponding to the inverted observation image and the upright observation image are determined to be normal retinal tissue regions based on the normal region determination criteria for brightness and shape corresponding to the inverted observation image and the upright observation image, respectively.

[0020] In the present invention, the normal region determination step can include: a region portion evaluation step for determining, for a candidate region identified in an image region of a positionally associated cell aggregate in each of the inverted observation image and the upright observation image, an evaluation score for each of the region portions present in a range obtained by dividing the candidate region into a plurality of regions along a curve representing the extension direction of the candidate region, based on evaluation criteria related to color and / or shape corresponding to each of the inverted observation image and the upright observation image; and a continuous length-based normal region determination step for determining, as a normal retinal tissue region, a region portion of a candidate region for which the evaluation scores in both the inverted observation image and the upright observation image are equal to or greater than the corresponding normal region determination criterion value, if the region portion is determined to exist continuously along the curve representing the extension direction of the candidate region for a length equal to or greater than a predetermined normal region determination criterion length.

[0021] In the present invention, the candidate region determination criteria in the candidate region identification step may further include that the brightness of the region to be identified is higher than the maximum or average brightness of the central region of the image region of the cell aggregate by a predetermined degree or more.

[0022] In the present invention, the candidate region determination criteria in the candidate region identification step can further include that, in the region to be identified, the outermost contour of the region portion contained therein as viewed from the center of the image region of the cell aggregate substantially matches the contour of the image region of the cell aggregate.

[0023] In the present invention, the candidate region determination criteria in the candidate region identification step may further include that the change in the angle of the tangent direction of the outermost contour of the region to be identified is continuous.

[0024] In the present invention, the candidate region determination criteria in the candidate region identification stage may further include that the width between the outermost and innermost parts of the region to be identified as viewed from the center of the image region of the cell aggregate is greater than or equal to a predetermined standard width.

[0025] In the present invention, the candidate region identification step may further include a non-target cell identification step of identifying whether or not non-target cells are present in the cell aggregate image region of each of the inverted observation image and the upright observation image based on predetermined non-target cell determination criteria based on color and morphology, and the candidate region determination criteria may further include the absence of non-target cells in the region to be identified.

[0026] In the present invention, the region portion evaluation step determines an evaluation score for each of the region portions present in a plurality of divided ranges within the candidate region present in a range of a predetermined angle from the center of the image region of the cell aggregate, based on evaluation criteria related to the morphology corresponding to each of the inverted observation image and the upright observation image, for the candidate region identified in the image region of the cell aggregate that is positionally associated with each of the inverted observation image and the upright observation image; and the continuous length-based normal region determination step can be configured to determine, as having a normal retinal tissue region, a cell aggregate in which region portions determined to have evaluation scores in both the inverted observation image and the upright observation image that are equal to or greater than the corresponding normal region determination criterion value exist consecutively for a predetermined normal region determination criterion angle or more.

[0027] In the present invention, the reference width for the candidate region identification step can be 0.03 mm for an inverted observation image and 0.15 mm for an upright observation image.

[0028] In the present invention, the normal region determination reference length in the continuous length-based normal region determination step can be set to 0.4 mm.

[0029] In the present invention, the normal region determination reference angle in the continuous length-based normal region determination step can be set to 45 degrees.

[0030] In the present invention, image data of inverted and upright observation images of a cell aggregate that may include retinal tissue can be obtained by photographing the cell aggregate suspended in a liquid.

[0031] In the present invention, image data of inverted and upright observation images of a cell aggregate that may contain retinal tissue can be obtained by photographing the cell aggregate in a plastic container.

[0032] In the present invention, the plastic container may be a container having one or more isolated culture areas.

[0033] In the present invention, the plastic container may be a spherical or cone-shaped container having a downward convex shape or a depression.

[0034] The present invention can be a method for producing retinal tissue that can be transplanted into an animal by excising a normal retinal tissue area determined by any of the above-mentioned methods.

[0035] The present invention can be a method for determining a normal retinal tissue area that can be transplanted into an animal, which further comprises the steps of: detecting the expression of neural retinal cell-related genes and non-neural retinal cell-related genes in a normal retinal tissue area determined by any of the above-mentioned methods; and determining that the normal retinal tissue area is transplantable into an animal when expression of neural retinal cell-related genes is observed and expression of non-neural retinal cell-related genes is not observed.

[0036] The present invention can be a method for producing retinal tissue that can be transplanted into an animal, by further extracting a part or all of the normal area as a quality assessment sample from a normal retinal tissue area, which is a part of a candidate area that has been determined by any of the above-mentioned methods to satisfy the normal area determination criteria; detecting the expression of neural retinal cell-related genes and non-neural retinal cell-related genes in the quality assessment sample; and, when expression of neural retinal cell-related genes is observed but expression of non-neural retinal cell-related genes is not observed, determining that (1) the neural retina (neural retina for transplantation) in the same cell aggregate as the cell aggregate that includes the quality assessment sample, which is a part of the cell aggregate; (2) the neural retina (neural retina for transplantation) in a cell aggregate that is the same lot as the cell aggregate that includes the quality assessment sample, which is a part of the cell aggregate; or (3) the neural retina (neural retina for transplantation) in a cell aggregate that is the same lot as the cell aggregate that includes the quality assessment sample, which is the whole of the cell aggregate, is usable as a neural retina for transplantation; and excising the normal retinal tissue area that has been determined to be usable as a neural retina for transplantation.

[0037] The present invention distinguishes, in an image region of each of the inverted observation image and the upright observation image of a cell aggregate that may contain retinal tissue derived from pluripotent stem cells, regions that are determined to satisfy the corresponding candidate region criteria for color, based on the corresponding candidate region criteria for the inverted observation image and the upright observation image, as candidate regions included in regions that are candidates for normal retinal tissue regions; and determines, in the candidate region of each of the inverted observation image and the upright observation image, cell aggregates that have candidate regions that are determined to satisfy the corresponding normal region criteria for brightness and / or morphology, based on the corresponding normal region criteria for the inverted observation image and the upright observation image, as containing normal retinal tissue regions. This makes it possible to analyze three-dimensional tissues in a culture medium with high accuracy in both space and three-dimensional terms using image data observed from two directions, top and bottom, using an upright microscope and an inverted microscope. Furthermore, by determining a pass / fail evaluation score from an appropriate perspective, detecting regions that satisfy the evaluation score when observed from two directions, top and bottom, and integrating these evaluation scores for evaluation, it becomes possible to automatically extract cell aggregates that contain three-dimensionally good normal regions by information processing.

[0038] 1 is an image of a cell aggregate containing retinal tissue observed by upright observation. 2 is an image of a region of one lobe of a cell aggregate photographed by inverted observation. 3 is an image of inverted and upright observation. 4 is an upright observation image and an inverted observation image of a cell aggregate of sample SPL1. 5 is an upright observation image and an inverted observation image of a cell aggregate of sample SPL2. 6 is an upright observation image and an inverted observation image of a cell aggregate of sample SPL3. 7 is an upright observation image and an inverted observation image of a cell aggregate of sample SPL4. 8 is an upright observation image and an inverted observation image of a cell aggregate of sample SPL5. 9 is an upright observation image and an inverted observation image of a cell aggregate of sample SPL6. 10 is an upright observation image and an inverted observation image of a cell aggregate of sample SPL7. 11 is an upright observation image and an inverted observation image of a cell aggregate of sample SPL8 held in a 96-well plate (U-bottom). 1 shows upright and inverted observation images of a cell aggregate of sample SPL8 held in a 96-well plate (V-bottom). 2 shows upright and inverted observation images of a cell aggregate of sample SPL8 held in a 24-well plate. 3 shows upright and inverted observation images of a cell aggregate of sample SPL8 held in a 6-well plate. 4 shows upright and inverted observation images of a cell aggregate of sample SPL8 held in a 60 mm dish. 5 shows upright and inverted observation images of a cell aggregate of sample SPL8 held in a T25 flask. 6 shows upright and inverted observation images of retinal tissue. 7 shows upright and inverted observation images of retinal pigment epithelium (RPE). 8 shows upright and inverted observation images of ciliary body tissue. 9 shows upright and inverted observation images of the optic stalk. 10 shows upright and inverted observation images of telencephalic tissue. 1. Upright and inverted observation images of spinal cord tissue. 2. Upright and inverted observation images of non-neural retinal tissue of sample SPL15. 3. Upright and inverted observation images of non-neural retinal tissue of sample SPL16. 4. Upright and inverted observation images of non-neural retinal tissue of sample SPL17. 5. Upright observation image of cell aggregate of sample SPL18 at 12x magnification. 6. Upright observation image of cell aggregate of sample SPL18 at 20x magnification. 7. Upright observation image of cell aggregate of sample SPL18 at 32x magnification. 8. Upright observation image of cell aggregate of sample SPL18 at 50x magnification.1 is an upright observation image of a cell aggregate of sample SPL18 at 80x magnification. 2 is an upright observation image of a cell aggregate of sample SPL18 at 100x magnification. 3 is an inverted observation image of a cell aggregate of sample SPL18 at 4x magnification, observed using phase contrast. 4 is an inverted observation image of a cell aggregate of sample SPL18 at 10x magnification, observed using phase contrast. 5 is an inverted observation image of a cell aggregate of sample SPL18 at 10x magnification, observed using Hoffman interference. 6 is an upright observation image of a cell aggregate of sample SPL19 at 25x magnification. 7 is an upright observation image of a cell aggregate of sample SPL19 at 50x magnification. 8 is an upright observation image of a cell aggregate of sample SPL19 at 80x magnification. 9 is an inverted observation image of a cell aggregate of sample SPL19 at 4x magnification, observed using phase contrast. 10 is an inverted observation image of a cell aggregate of sample SPL19 at 10x magnification. 1 is an inverted observation image of a cell aggregate of sample SPL19 by Hoffman interference at 10x magnification. FIG. 2 is a schematic diagram showing the configuration of a normal cell aggregate determination system 100. FIG. 3 is a block diagram of a normal cell aggregate determination server 110. FIG. 4 is an overall operational flow diagram of Example 1. FIG. 5 is a detailed operational flow diagram related to normal region determination in Example 1. FIG. 6 is an overall operational flow diagram of Example 2. FIG. 7 is an operational flow diagram of a preparation procedure in Example 2. FIG. 8 is a detailed operational flow diagram related to normal region determination in Example 2. FIG. 9 is an overall operational flow diagram of Example 3. FIG. 10 is an overall operational flow diagram of Example 4. FIG. 11 is a detailed operational flow diagram related to normal region determination in Example 4. FIG. 12 is an inverted observation image (before marking and after marking) of a cell aggregate to be determined in Example 1. FIG. 13 is an upright observation image (before marking and after learning marking) of a cell aggregate for learning in Example 2. FIG. 14 is an upright observation image (before marking and after marking) of a cell aggregate to be determined in Example 2. FIG. 15 is an inverted observation image and an upright observation image after marking of a cell aggregate to be determined in Example 4. Fig. 1 is a conceptual diagram for measuring the width of a candidate region for a cell aggregate. Fig. 2 is a diagram showing the width of a region portion of a candidate region, an evaluation score, and a determination result in an inverted observation image of Example 1. Fig. 3 is a diagram showing the width of a region portion of a candidate region, an evaluation score, and a determination result in an upright observation image of Example 2. Fig. 4 is a diagram showing the width of a region portion of a candidate region, an evaluation score, and a determination result in an upright observation image and an inverted observation image of Example 3. Fig. 5 is a diagram showing the width of a region portion of a candidate region, an evaluation score, and a determination result in an upright observation image and an inverted observation image of Example 4.

[0039] Explanation of Terms "Stem cells" refer to undifferentiated cells that have the ability to differentiate and proliferate (especially the ability to self-renew). Stem cells include subpopulations such as pluripotent stem cells, multipotent stem cells, and unipotent stem cells, depending on their differentiation ability. Pluripotent stem cells refer to stem cells that can be cultured in vitro and have the ability (pluripotency) to differentiate into all cell lineages belonging to the three germ layers (ectoderm, mesoderm, and endoderm) and / or extraembryonic tissues. Multipotent stem cells refer to stem cells that have the ability to differentiate into multiple types of tissues and cells, but not all types. Unipotent stem cells refer to stem cells that have the ability to differentiate into specific tissues or cells.

[0040] "Pluripotent stem cells" can be induced from fertilized eggs, cloned embryos, germline stem cells, tissue stem cells, somatic cells, etc. Examples of pluripotent stem cells include embryonic stem cells (ES cells), embryonic germ cells (EG cells), and induced pluripotent stem cells (iPS cells). Muse cells (Multi-lineage Differentiating Stress Enduring Cells) obtained from mesenchymal stem cells (MSCs) and GS cells prepared from germ cells (e.g., testes) are also included in pluripotent stem cells.

[0041] "Retinal tissue" refers to tissue in which one or more types of retinal cells that make up each retinal layer in a living retina exist in a certain order, and "neural retina" refers to retinal tissue that includes the inner neural retinal layer, which does not include the retinal pigment epithelium layer, among the retinal layers described below.

[0042] In this specification, a region of retinal tissue that does not contain undesired cells and has a certain size suitable for transplantation is referred to as a "normal retinal tissue region."

[0043] The term "retinal cells" refers to cells that constitute each retinal layer in a living retina or their progenitor cells. Retinal cells include, but are not limited to, photoreceptors (rod photoreceptors, cone photoreceptors), horizontal cells, amacrine cells, interneurons, retinal ganglion cells (ganglion cells), bipolar cells (rod bipolar cells, cone bipolar cells), Müller glial cells, retinal pigment epithelial (RPE) cells, ciliary bodies, their progenitor cells (e.g., photoreceptor progenitor cells, bipolar cell progenitor cells, etc.), and retinal progenitor cells. Among retinal cells, cells that constitute the neural retinal layer (also referred to as neural retina cells or neural retina-related cells) specifically include photoreceptors (rod photoreceptors, cone photoreceptors), horizontal cells, amacrine cells, interneurons, retinal ganglion cells (ganglion cells), bipolar cells (rod bipolar cells, cone bipolar cells), Müller glial cells, and their precursor cells (e.g., photoreceptor precursor cells, bipolar cell precursor cells, etc.). In other words, neural retinal cells do not include RPE cells and ciliary body cells.

[0044] The term "photoreceptor precursor cells" refers to precursor cells that are committed to differentiating into photoreceptor cells.

[0045] The term "retinal progenitor cells" refers to precursor cells that can differentiate into any immature retinal cell, such as a photoreceptor precursor cell, horizontal cell precursor cell, bipolar cell precursor cell, amacrine cell precursor cell, retinal ganglion cell precursor cell, Müller glial cell, or RPE precursor cell, and ultimately can differentiate into any mature retinal cell, such as a photoreceptor cell, rod photoreceptor cell, cone photoreceptor cell, horizontal cell, bipolar cell, amacrine cell, retinal ganglion cell, or RPE cell.

[0046] "Neural retinal progenitor cells" refer to progenitor cells that can differentiate into any immature neural retinal cell, such as a photoreceptor progenitor cell, horizontal cell progenitor cell, bipolar cell progenitor cell, amacrine cell progenitor cell, retinal ganglion cell progenitor cell, or Müller glial cell, and ultimately, can differentiate into any mature neural retinal cell, such as a photoreceptor cell, rod photoreceptor cell, cone photoreceptor cell, horizontal cell, bipolar cell, amacrine cell, or retinal ganglion cell. In this specification, when the term "retinal progenitor cell" is used without any special explanation, it is intended to refer to a broader concept of cells that also includes "neural retinal progenitor cells."

[0047] "Photoreceptor cells" are present in the photoreceptor layer of the retina in living organisms and have the role of absorbing light stimuli and converting them into electrical signals. There are two types of photoreceptor cells: cones that function in bright light and rods that function in dark light (referred to as cone photoreceptors and rod photoreceptors, respectively). Examples of cone photoreceptors include S-cone photoreceptors that express S-opsin and receive blue light, L-cone photoreceptors that express L-opsin and receive red light, and M-cone photoreceptors that express M-opsin and receive green light. Photoreceptors differentiate and mature from photoreceptor precursor cells.

[0048] The term "retinal layer" refers to each layer that constitutes the retina, and specific examples include the retinal pigment epithelium layer, photoreceptor layer, outer limiting membrane, outer nuclear layer, outer plexiform layer, inner nuclear layer, inner plexiform layer, ganglion cell layer, nerve fiber layer, and inner limiting membrane.

[0049] "Neural retinal layer" refers to each layer that constitutes the neural retina, and specific examples include the photoreceptor layer, outer limiting membrane, outer nuclear layer, outer plexiform layer, inner nuclear layer, inner plexiform layer, ganglion cell layer, nerve fiber layer, and internal limiting membrane. "Photoreceptor layer" refers to the retinal layer that is formed on the outermost side of the neural retina and contains many photoreceptors (rod photoreceptors, cone photoreceptors), photoreceptor precursor cells, and retinal progenitor cells. Considering the continuity of layer structure formation, the photoreceptor layer may include a retinal progenitor cell layer in addition to the photoreceptor layer. Layers other than the photoreceptor layer are referred to as inner layers. Which retinal layer each cell constitutes can be confirmed by known methods, for example, the presence or absence of expression or the degree of expression of a cell marker.

[0050] The outer neuroblastic layer (ONbL) is a layer including a photoreceptor progenitor cell layer and a neural retinal progenitor cell layer. Most of the photoreceptor progenitor cells included in the photoreceptor progenitor cell layer differentiate into photoreceptors (rod photoreceptors and cone photoreceptors) to form part of the photoreceptor layer. The photoreceptor progenitor cells continue to exist as a layer with several rows of densely packed spindle-shaped nuclei and differentiate into cone cells and rod cells. The neural retinal progenitor cell layer differentiates into neural retinal cells to form part of the neural retinal layer. The photoreceptor progenitor cells continue to exist as a layer with several rows of densely packed spindle-shaped nuclei and differentiate into cone cells and rod cells.

[0051] The inner neuroblastic layer (INbL) is a layer containing retinal nerve cells, which corresponds to the inner layer of retinal nerve cells.

[0052] The term "ciliary body" includes the "ciliary body" and "ciliary marginal zone (CMZ)" during development and in adults. The "ciliary marginal zone" refers to, for example, tissue present at the boundary between the neural retina and the RPE in a living retina, and is a region containing retinal tissue stem cells (retinal stem cells). The ciliary marginal zone is also called the ciliary margin or retinal margin, and the ciliary marginal zone, ciliary marginal zone, and retinal marginal zone are equivalent tissues. The ciliary marginal zone is known to play an important role in supplying retinal progenitor cells and differentiated cells to retinal tissue, maintaining retinal tissue structure, and the like. A "ciliary marginal zone-like structure" refers to a structure similar to the ciliary marginal zone. The ciliary body is considered to be a non-target cell in relation to retinal tissue.

[0053] A "cell aggregate" is a three-dimensional structure formed by the adhesion of multiple cells. There are no particular limitations on the structure, and it refers to, for example, a mass formed by the aggregation of cells dispersed in a medium such as a culture medium, or a mass of cells formed through cell division. Cell aggregates also include those that form specific tissues. In this specification, cell aggregates that form specific tissues may be referred to as "three-dimensional tissues," and three-dimensional tissues that form retinal tissues may be referred to as "three-dimensional retina."

[0054] "Epithelial tissue" (epithelium) is a tissue formed by cells tightly covering the surface of the body, a lumen (such as the digestive tract), a body cavity (such as the pericardial cavity), etc. Cells forming epithelial tissue are called epithelial cells. Epithelial cells have an apical-basal polarity. Epithelial cells form strong bonds with each other through adherens junctions and / or tight junctions, and can form cell layers. Tissues formed by one to a dozen of these cell layers overlapping are epithelial tissue. Tissues that can form epithelial tissue include fetal and / or adult retinal tissue, cerebrospinal tissue, ocular tissue, nervous tissue, etc. The neural retina in this specification is also an epithelial tissue. "Epithelial structure" refers to a structure characteristic of epithelial tissue, such as an apical surface or a basement membrane. "Epithelial tissue" is sometimes referred to as a "leaf."

[0055] "Continuous epithelial tissue" refers to tissue with a continuous epithelial structure. A continuous epithelial structure is a structure in which epithelial tissue is continuous. A continuous epithelial tissue is, for example, a structure in which 10 cells to 10 cells are connected in a tangential direction to the epithelial tissue. 7 cells, preferably 30 cells to 10 7 cells, more preferably up to 10 2 cells ~10 7 It refers to the state of cells lined up.

[0056] For example, a continuous epithelial structure formed in retinal tissue has an apical surface characteristic of epithelial tissue, and the apical surface is formed on the surface of the retinal tissue generally parallel to and continuous with at least the photoreceptor layer (outer nuclear layer) among the layers forming the neural retina. For example, in the case of a cell aggregate containing retinal tissue prepared from pluripotent stem cells, the apical surface is formed on the surface of the aggregate, and 10 or more, preferably 30 or more, more preferably 100 or more, and even more preferably 400 or more photoreceptor cells or photoreceptor precursor cells are regularly and continuously arranged in a tangential direction to the surface.

[0057] "Pseudostratified epithelium" is a type of epithelial tissue composed of superficial epithelial cells, interstitial epithelial cells, and basal epithelial cells. Although it appears to be stratified epithelium with two or more cells arranged in an apical-basal direction, the bottoms of all cells are in contact with the basement membrane, and it is actually a single-layer epithelial tissue.

[0058] The "optic stalk" is a tissue that forms at the base of the optic cup, connecting the eye to the forebrain, and later becomes the optic nerve. The optic stalk is a non-target cell in relation to the retinal tissue.

[0059] The "retinal pigment epithelium" (RPE) is a sheet-like single layer of cells located on the outer side of the retina, and is responsible for supporting the photoreceptor cells of the retina. The RPE is a non-target cell in relation to the retinal tissue.

[0060] The "telencephalic tissue" is the anterior part of the brain vesicles that form at the cranial end of the neural tube, and is the tissue that ultimately becomes the left and right cerebral hemispheres during the developmental process of the brain. Telencephalic tissue is a non-target cell in relation to retinal tissue.

[0061] "Spinal cord tissue" refers to the tissue of the spinal cord, which is a cylindrical extension of the central nervous system extending from the medulla oblongata inside the spinal canal. Spinal cord tissue is a non-target cell in relation to retinal tissue.

[0062] Morphological and Color Characteristics of Retinal Tissue Cell aggregates can be characterized by their morphology and color. Below, we will explain the morphological and color characteristics of retinal tissue used to determine the area of ​​normal retinal tissue contained in an image of a cell aggregate. Note that the morphological characteristics of retinal tissue used to determine the area of ​​normal retinal tissue are visually recognizable morphological characteristics, which may vary depending on the observation method. Such morphological characteristics of retinal tissue were first discovered in the process of developing the present invention by confirming images of cell aggregates that may contain large amounts of retinal tissue using the observation method described below. Color refers to a color stimulus that causes color perception and has three attributes: brightness, hue, and saturation. Note that the attribute of color that is judged sensorily, focusing particularly on hue and taking brightness and saturation into account, is sometimes referred to as color tone. Note that color and tone also include cases where only achromatic colors are the subject of judgment. In this case, the only difference in color and tone is brightness, and the image with only this difference is grayscale. Figure 2 shows an image of one lobe region of a cell aggregate photographed by inverted observation, and the morphological characteristics of typical retinal tissue can be confirmed.

[0063] First, when observing the entire cell aggregate macroscopically, retinal tissue has the following characteristics. That is, the surface of the cell aggregate in the region containing retinal tissue is smooth, and the shape of the cell aggregate is a convex arc that faces upward from the center to the periphery, and the curvature of the arc is also gentle. In Figure 2, the shape is an arc that faces upward. Furthermore, the interior of the cell aggregate may contain black material (collagen and Pax6-positive cells).

[0064] Furthermore, good retinal tissue has a continuous epithelial structure, which is observed as being clearly separated into two regions: an outer bright ring-shaped or arc-shaped region ONbL with a certain width, and an inner dark region INbL. Here, the outer ONbL region is an ONbL region composed of retinal progenitor cells and photoreceptor progenitor cells, and the inner INbL region is an INbL region composed of inner layer retinal neurons.

[0065] The ONbL region is characterized by a bright color tone and a high degree of graininess (grainy texture) in the cells contained therein, meaning that each cell can be clearly distinguished compared to other tissues. Furthermore, in normal retinal tissue, the ONbL region has a width greater than a certain value.

[0066] In a continuous epithelial structure with an ONbL region, the same type of epithelial cells form the epithelial structure and exist continuously in the tangential direction. Therefore, in the ONbL region, the tissue cells are arranged continuously and regularly in the tangential direction. Furthermore, the periphery of the ONbL region is smooth, and the angle of the tangential line does not change suddenly.

[0067] Furthermore, retinal tissue contains a dark INbL region inside the ONbL region, which corresponds to the inner layer of retinal nerve cells. Its color is dark brown, and the cells contained therein have a low degree of granularity (making it difficult to clearly distinguish individual cells). Furthermore, this INbL region does not reach the outer periphery of the outer ONbL region, and the outer ONbL region and the inner INbL region typically each have a width of at least a certain value. Conversely, a cell aggregate portion having such characteristics can be determined to be normal retinal tissue. Typically, if the outer ONbL region has a width of at least a certain value, the inner INbL region also has a width of at least a certain value. Therefore, if it is confirmed that the outer ONbL region has a width of at least a certain value, the ONbL region and the adjacent inner INbL region are determined to constitute normal retinal tissue.

[0068] Furthermore, there is a clear boundary between the outer ONbL region and the inner INbL region. First, the arrangement of cells differs between the ONbL region and the INbL region. In the ONbL region, cells are densely arranged, and cell bodies (Soma) are in close contact with each other. Furthermore, in the ONbL region, the orientation of each cell body is consistent. That is, the ONbL region has a structure in which granular cells are arranged in a continuous, regular pattern in the tangent direction of the tissue contour. On the other hand, in the inner INbL region, cells are sparsely arranged, and cell bodies are not in close contact with each other. Furthermore, in the INbL region, the orientation of each cell body is slightly disordered. Note that the orientation of non-target cells is significantly disordered and disordered.

[0069] Here, granularity refers to the visibility of the structural characteristics of individual, independent cells, and the degree of granularity is particularly noticeable when the focus position is changed. By shifting the focus position when observing a cell aggregate, the degree of granularity at the focused position can be confirmed. In Figure 2, the cell aggregate is shown as an arc-shaped shape, but the ONbL region has a high degree of granularity of cells. Thus, the ONbL region has a structure in which granular cells are arranged continuously and regularly in the tangential direction of the tissue outline.

[0070] The granularity of retinal tissue differs from that of the eyestalk, for example. The eyestalk has an epithelial structure similar to retinal tissue and a bright color tone, but the tissue appears to have vertical lines running through it, making it difficult to see individual cells as separate particles. This is thought to be because the cells of the eyestalk are more elongated than cells in retinal tissue. On the other hand, the cells in retinal tissue are neural cells and therefore somewhat elongated, but their elongation is less pronounced than in other tissues such as the eyestalk, making them more easily visible as separate particles. In other words, the granularity of retinal tissue is higher. In fact, cells in the ONbL region (retinal progenitor cells and photoreceptor progenitor cells) are less elongated and have elliptical nuclei. This structure is called pseudostratified epithelium. The nuclei of mesenchymal cells and inner layer retinal cells are nearly circular.

[0071] Explanation of Observation Method In order to identify high-quality retinal tissue in cell aggregates, it is necessary to accurately distinguish between retinal tissue and non-retinal tissue. As explained below, it was confirmed that retinal tissue can be identified with high accuracy by using images acquired using both an upright microscope and an inverted microscope. Observing with both an upright microscope and an inverted microscope has the following advantages:

[0072] First, observing with both an upright microscope and an inverted microscope has the advantage of being able to detect unwanted cells not only on the front side of the three-dimensional retina but also on the back side. In other words, observation from only one side only allows for a view from one side of the object, and it may not be possible to fully focus the entire area of ​​the object, making it difficult to accurately assess quality. This is particularly noticeable when observing from an inverted microscope. As a result, there is a risk that quality issues such as collapsed retinal structure or the presence of dead cells may not be detected. However, observing from both sides reduces the likelihood of such issues occurring.

[0073] In addition, since images can be obtained that are suited to the characteristics of both the upright and inverted microscopes, there is the advantage that quality can be judged more accurately by combining images observed from both the upright and inverted microscopes.

[0074] Because inverted microscopes have higher resolution than upright microscopes, inverted observation allows for highly accurate observation of intracellular structures, making it easier to determine whether an object is normal or not based on its structural characteristics. Therefore, inverted observation allows for clear identification of extraneous tissues, such as the eyestalk and telencephalic tissue, which have epithelial structures similar to the retina. Here, the intracellular structure typically consists of cells with a granular texture arranged in a regular, continuous pattern tangential to the tissue's contour. Figure 2 shows an image of a cell aggregate containing retinal tissue observed using inverted observation. While a portion of the cell aggregate is shown as an arc in Figure 2 , a bright ring-shaped region near its circumference exhibits a structure resembling a collection of granular particles. This collection of particles is part of the ONbL region contained in the granular retinal tissue, and such characteristics can be clearly observed using inverted observation. Thus, inverted observation allows for clear distinction between retinal tissue and extraneous tissue within the cell aggregate.

[0075] On the other hand, upright microscopes have a wider focal range than inverted microscopes, so upright observation allows for the acquisition of an image in which the entire cell aggregate is in focus, enabling the overall characteristics of the cell aggregate to be understood and assessed. However, the resolution is somewhat inferior to that of inverted observation. Even with an inverted microscope, acquiring multiple images can provide information about out-of-focus areas, but this is not practical when observing large numbers of cell aggregates. Furthermore, with upright microscopes, oblique illumination allows for longer exposure times, making it easier to detect cells with low light transmittance, such as RPE and dead cells, and shadows in the shape of the cell aggregate. Figure 1 shows an image of a cell aggregate containing retinal tissue observed using upright observation. While Figure 1 shows the entire cell aggregate, a darker region can be observed slightly below the center. Thus, upright observation allows for easier detection of non-target cells.

[0076] Combining the two observation methods, each with its own unique characteristics, as described above, can compensate for the shortcomings of each method, enabling synergistically more effective observations. In other words, both the overall analysis of cell aggregates and the detailed structural analysis enable reliable detection of good retinal tissue in a shorter time, and also more reliable detection of unintended cells, which are easily detected by each observation method. The features of synergistic analysis using upright and inverted observation are described below.

[0077] Inverted observation generally has a shallow depth of focus and a narrow range of focus, but high resolution, making it suitable for extracting detailed information about a cross section of three-dimensional tissue. However, it is difficult to obtain information about out-of-focus areas. On the other hand, it is possible to evaluate in detail structures such as cell arrangement for cross sections within a narrow range of focal depth. In this way, inverted observation makes it possible to obtain excellent "structural information about the cross section."

[0078] Upright observation generally has a deep depth of focus and a wide range of focus, but relatively low resolution, making it suitable for extracting overall surface texture information of three-dimensional tissue within a wide range of depth of focus. While upright observation can acquire a large amount of overall surface texture information, it is difficult to acquire more detailed structural information, such as the arrangement of cells within the tissue, than in inverted observation. Thus, upright observation makes it possible to acquire "overall surface texture information" well.

[0079] When observing cell aggregates and determining whether a region of the aggregate is retinal tissue, the presence of an epithelial structure is typically used as an indicator. However, some non-target cells also contain epithelial structures similar to those of the retina. Therefore, while the surface texture information obtained from upright observation images can detect the presence of epithelial structures, it is difficult to determine the cell arrangement and detailed structure within the epithelial structures. Thus, it is difficult to distinguish between retinal tissue with epithelial structures and non-target cells with epithelial structures using upright observation alone. On the other hand, cross-sectional structural information obtained from inverted observation images allows for the distinction between retinal tissue and non-target cells by observing the cell arrangement and detailed epithelial structure. Furthermore, some non-target cells (mainly RPE cells) are close to black in color. Therefore, using cross-sectional structural information obtained from inverted observation images alone may not be able to determine whether they are black non-target cells due to the narrow focus range, poor light penetration within the cell aggregate, and the overall dark appearance of inverted observation images. Even in such cases, the upright observation image provides surface texture information from a wide, in-focus area, making it easy to detect the black non-target cells. Therefore, by comparing the inverted and upright observation results in a complementary manner, normal retinal tissue can be detected with high accuracy.

[0080] Whether upright or inverted, images captured by a microscope can be discriminated through information processing to rapidly obtain objective evaluation results. Information processing can involve establishing criteria based on various perspectives, calculating scores for evaluation parameters of images of epithelial tissue captured by a microscope based on the criteria, and then performing such evaluations based on the scores. The criteria can be based on characteristics such as color and morphology in images of cellular aggregates containing retinal tissue. The criteria can also be set by setting thresholds or other criteria through information processing using images of cellular aggregates containing retinal tissue and the evaluation results as training data. Another example of information processing involves training a machine learning model using deep learning, using a combination of images of cellular aggregates containing retinal tissue captured by a microscope and the labels of the evaluation results as training data, and inputting the images to be evaluated into the machine learning model based on the training data to output an estimated evaluation result.

[0081] Below, we describe two examples in which combining the results of inverted and upright observations allows for highly accurate detection of unwanted cells. In both cases, small shadows are observed when observing a 3D retina using inverted observation, and it is necessary to confirm whether these shadows are unwanted cells. Generally, when culturing 3D tissues, dead cells accumulate at the bottom of the wells, and these are usually observed with a certain frequency when observing them from below using inverted observation. Furthermore, because the color (darkness) of unwanted cells (RPE cells) varies depending on the maturity of the cells and the imaging conditions (exposure conditions and medium volume), it is difficult to accurately identify them when they are small. In this case, combining the shadows observed using inverted observation with upright observation, which makes it easier to capture the surface texture, makes it easier to determine whether the shadows are dead cells or unwanted cells.

[0082] Figure 3 shows conceptual diagrams of inverted observation and upright observation. Cell aggregate 10 is held in culture solution 24 in container 23. Upright observation 31 is when cell aggregate 10 is observed from above with upright microscope 21 through the surface of culture solution 24 (the observation form represented by the symbol with a downward-pointing eye and arrow in Figure 3), and inverted observation 32 is when cell aggregate 10 is observed from below with inverted microscope 22 through the bottom of the container (the observation form represented by the symbol with a upward-pointing eye and arrow in Figure 3). Upright microscope 21 and inverted microscope 22 can output the observation images obtained by these observations, that is, upright observation images and inverted observation images, respectively, as image data.

[0083] The upper image in Figure 4 is an upright observation image of cell aggregate sample SPL1, and the lower image in Figure 4 is an inverted observation image of sample SPL1. In the upright and inverted observation images in Figure 4, one of the images is mirror-inverted so that the upright and inverted observation images of the cell aggregate have similar contours, and the positions in each image correspond to the same positions on the cell aggregate as viewed from a vertical direction. The same applies to the following figures comparing upright and inverted observation images of the same cell aggregate. In the inverted observation image in Figure 4, a small shadow is observed in the area circled. However, because the inverted observation image only contains information about a flatter area than the actual image, it is difficult to determine whether this shadow represents a dead cell or an unintended cell. However, in the upright observation image in Figure 4, the smooth continuity of light and dark tones characteristic of a continuous epithelial structure is disrupted, and cell bulges can be confirmed on the surface, indicating the presence of cells with different textures. Therefore, this shadow can be determined to be a three-dimensional structure. Furthermore, returning to the inverted image in Figure 4, it can be observed that the shadow in the circled area has a similar color tone (blackness) to the black non-target cells to the right. Taking these into consideration, it can be inferred that the shadow is the RPE of a similar non-target cell.

[0084] The upper image in Figure 5 is an upright observation image of cell aggregate sample SPL2, and the lower image in Figure 5 is an inverted observation image of the same sample. A small shadow can be seen in the circled area in the inverted observation image in Figure 5. When examining the upright observation image in Figure 5 at the location corresponding to the circled area in the inverted observation image, a smooth continuity of light and dark tones is observed, and the surface texture is also confirmed to be undisturbed. This is characteristic of a smooth, continuous epithelial structure, which allows us to determine that the shadow is not a three-dimensional structure. Furthermore, returning to the inverted observation image in Figure 5, the shadow has the same color tone as the dead cells visible around the cell aggregate. Taking these factors into consideration, we can infer that the shadow is dead cells that have accumulated at the bottom of the well. Therefore, we can determine that the dead cells confirmed as shadows in the circled area are not attached to the cell aggregate, and that the cell aggregate is epithelial tissue that does not contain non-target cells.

[0085] As mentioned above, it is clear that combining inverted and upright observation images allows for a more accurate determination of whether a cell aggregate contains normal cells. The characteristics of various observation methods will now be examined in more detail. The following table lists the observation methods examined, along with the samples used and the corresponding drawings.

[0086]

[0087] <Observation Method 1 - Observation of Cell Aggregates Including Neural Retina Using Upright and Inverted Microscopes> As a new observation method for detecting unintended cells contained in three-dimensional tissues, the effectiveness of observation using upright and inverted microscopes was investigated using the following procedure. First, we investigated whether unintended cells on cell aggregates could be reliably detected by combining observation using upright and inverted microscopes.

[0088] Human iPS cells (DSP-SQ line, established by Sumitomo Pharma Co., Ltd.) were regenerated using commercially available Sendai virus vectors (four factors: Oct3 / 4, Sox2, KLF4, and c-Myc, CytoTune Kit, manufactured by IDPharma) according to the published protocol of Thermo Fisher Scientific (iPS2.0 Sendai Reprogramming Kit, Publication Number MAN0009378, Revision 0). 1.0) and the published protocol of Kyoto University (Establishment and maintenance of human iPS cells in the feeder-free environment, CiRA_Ff-iPSC_protocol_JP_v140310, http: / / www.cira.kyoto-u.ac.jp / j / research / protocol.html), using StemFit medium (AK03; manufactured by Ajinomoto Co., Inc.) and Laminin 511-E8 (manufactured by Nippi Co., Ltd.).

[0089] The human iPS cells (DSP-SQ strain) were cultured in a feeder-free environment according to the method described in Scientific Reports, 9, 18936 (2019). StemFit medium (AK03N, manufactured by Ajinomoto Co., Inc.) was used as the feeder-free medium, and Laminin 511-E8 (manufactured by Nippi Co., Ltd.) was used as the feeder-free scaffold.

[0090] Specifically, the maintenance culture procedure involved first washing subconfluent human iPS cells (DSP-SQ strain) with PBS and dispersing them into single cells using TrypLE Select (Life Technologies). The dispersed human iPS cells were then seeded onto a plastic culture dish coated with Laminin 511-E8 and cultured in a feeder-free manner in StemFit medium in the presence of Y27632 (a ROCK inhibitor, 10 μM). The aforementioned plastic culture dish was a 6-well plate (Iwaki, for cell culture, culture area 9.4 cm). 2 ), the seeding number of human iPS cells dispersed into single cells was 1.0 × 10 4One day after seeding, the medium was replaced with StemFit medium without Y27632. Thereafter, the medium was replaced with StemFit medium without Y27632 once every 1 to 2 days. Thereafter, the cells were cultured for 5 days after seeding.

[0091] For differentiation induction, human iPS cells (DSP-SQ strain) were cultured in StemFit medium in a feeder-free manner until they reached subconfluence (approximately 30% of the culture area was covered with cells) two days before the subconfluence. These human iPS cells were then cultured in the presence of SAG (300 nM) for two days in a feeder-free manner (preconditioning treatment).

[0092] The preconditioned human iPS cells were treated with TrypLE Select (Life Technologies) to prepare a cell dispersion solution, and then dispersed into single cells by pipetting. The dispersed human iPS cells were then cultured in a non-cell-adhesive 96-well culture plate (PrimeSurface 96V bottom plate, Sumitomo Bakelite Co., Ltd.) at 1 × 10 cells per well. 4 The cells were suspended in 100 μL of serum-free medium and cultured at 37 °C and 5% CO 2 . The serum-free medium (gfCDM + KSR) used was a 1:1 mixture of F-12 medium and IMDM medium supplemented with 10% KSR, 450 μM 1-monothioglycerol, and 1× chemically defined lipid concentrate. At the start of suspension culture (day 0 after the start of suspension culture), Y27632 (final concentration 20 μM) and SAG (final concentration 10 nM) were added to the serum-free medium. On the third day after the start of suspension culture, 50 μL of the above-mentioned serum-free medium was added to the medium containing human recombinant BMP4 (manufactured by R&D Co.) but not containing Y27632 or SAG, so that the final concentration of exogenous human recombinant BMP4 was 1.5 nM (55 ng / mL).

[0093] Three days later (six days after the start of suspension culture), the medium was replaced with the serum-free medium described above that did not contain Y27632, SAG, or human recombinant BMP4. The medium replacement procedure involved discarding 60 μL of the medium in the incubator and adding 90 μL of fresh serum-free medium described above, for a total medium volume of 180 μL. Thereafter, half of the medium was replaced every two to four days with the serum-free medium described above that did not contain Y27632, SAG, or human recombinant BMP4. The half-medium replacement procedure involved discarding half the volume of the medium in the incubator (i.e., 90 μL) and adding 90 μL of fresh serum-free medium described above, for a total medium volume of 180 μL.

[0094] The cell aggregates thus obtained on day 13 after the initiation of suspension culture were cultured in serum-free medium (DMEM / F12 medium supplemented with 1% N supplement) containing CHIR99021 (3 μM) and SU5402 (5 μM) for 3 days, i.e., until day 16 after the initiation of suspension culture.

[0095] The resulting cell aggregates on day 16 after the initiation of suspension culture were cultured under 5% CO2 conditions using the serum media shown in [1], [2], and [3] below until day 75 after the initiation of suspension culture. [1] From day 16 to day 40 after the initiation of suspension culture: DMEM / F12 medium supplemented with 10% fetal bovine serum, 1% N2 supplement, and 100 μM taurine (hereinafter referred to as medium A). [2] From day 40 to day 60 after the initiation of suspension culture: A medium was mixed at a 1:3 ratio with medium A and Neurobasal medium supplemented with 10% fetal bovine serum, 2% B27 supplement, 2 mM glutamine, 60 nM T3, and 100 μM taurine (hereinafter referred to as medium B). [3] From day 60 after the initiation of suspension culture: B medium.

[0096] Cell aggregates were transferred to a 96-well slit-well plate 76 days after the start of suspension culture. Two cell aggregates were selected from the resulting aggregates and designated as samples SPL3 and SPL4. Bright-field images (phase contrast images, 4x objective lens) of samples SPL3 and SPL4 were obtained using an inverted microscope (Olympus IX-83) to obtain inverted images. The same cell aggregates were also observed using an upright microscope (Carl Zeiss AxioZoom.V16) at a zoom magnification of 33.3x to obtain upright images. The obtained images of samples SPL3 and SPL4 are shown in Figures 6 and 7, respectively. In both figures, the upper image is the upright image, and the lower image is the inverted image.

[0097] The general procedure for capturing observation images is described below. First, the XY axis of the object to be captured on the microscope is adjusted. In the case of an upright microscope, the cell aggregate is positioned below the objective lens when capturing images. In the case of an inverted microscope, the cell aggregate is positioned above the objective lens when capturing images. It is preferable to store the type of culture vessel and the position of the cell aggregate in the microscope's control software, and then move the microscope stage on which the culture vessel is placed so that each cell aggregate is positioned in front of the objective lens. When a special multi-well plate with a U-bottom or V-bottom is used as the culture vessel, the XY coordinates of the cell aggregate are uniquely determined as the well position. It is also possible to store the coordinates of each well in the microscope's control software and control the movement of each well in front of the objective lens.

[0098] Next, the microscope's subject is focused in the Z direction. When using an upright microscope (stereo microscope and upright zoom microscope), the focus is adjusted to the vertical center of the cell aggregate. When using an inverted microscope, the focus is adjusted to the center of the large epithelial tissue contained in the cell aggregate. For more detailed analysis, it is preferable to use a Z-stack image reconstructed from images captured by gradually shifting the focal plane.

[0099] Next, microscopic images are captured using both the upright microscope and the inverted microscope. While the order of inverted and upright observations can be chosen, it is preferable to perform the upright observation first, since steam on the well plate lid makes upright observation difficult. After capturing the observation images, the image data is output from the microscope and transferred to an image analysis system such as the normal cell aggregate determination system 100 (described below) via a data transfer device (e.g., LAN, wireless LAN, Bluetooth (registered trademark), USB, USB memory, external SSD, external hard disk, etc.).

[0100] For cell aggregate sample SPL3, the presence of RPE cells with black pigment, which are non-target cells, can be confirmed in the area indicated by the arrow in the upright observation image of Figure 6. However, in the inverted observation image of Figure 6, only a portion of the same RPE cells could be confirmed. Furthermore, for cell aggregate sample SPL4, the presence of RPE cells, which are non-target cells, can be confirmed in the area indicated by the arrow in the inverted observation image of Figure 7. However, in the upright observation image of Figure 7, RPE cells could not be confirmed. These results confirmed that when observing three-dimensional tissues with spatially complex structures, there are tissues that cannot be confirmed by observation from a single direction, and that the entire tissue can be observed by observing from both the top and bottom vertical directions.

[0101] Observation Method 2: Observation of Neural Retina Sheets Using Upright and Inverted Microscopes The effectiveness of bidirectional observation using upright and inverted observation images for neural retinal sheets was investigated using the following procedure. Some of the cell aggregates prepared using the method described in Observation Method 1 were observed, and upright and inverted images were acquired. Based on the acquired upright and inverted images, a region in the cell aggregate that had a continuous epithelial structure and appeared to be divided into two layers, the ONbL (including the photoreceptor layer and neural retinal progenitor cell layer) and the INbL, when viewed from above and below was identified and determined to be neural retina. Cell aggregates determined to contain neural retina were then selected from the multiple observed cell aggregates. While observing the selected cell aggregates using a stereomicroscope, neural retina was excised from the cell aggregate using fine-tipped tweezers and scissors under observation using a stereomicroscope (Carl Zeiss, STEREO) to prepare neural retinal sheets consisting of a single lobe. Three neural retinal sheets were selected from these samples and were used for the study as samples SPL5 to SPL7.

[0102] The prepared neural retinal sheets were transferred to a 96-well U-bottom plate. Bright-field images (phase-contrast images, 4x objective lens) of the neural retinal sheets were observed using an inverted microscope (Olympus IX-83). The same neural retinal sheets were also observed using an upright microscope (Carl Zeiss AxioZoom zoom microscope) at a zoom magnification of 50x. Observation images of the resulting samples SPL5 to SPL7 are shown in Figures 8 to 10, respectively. In each figure, the upper image is an upright observation image, and the lower image is an inverted observation image.

[0103] In Figures 8 to 10, in both the upright observation image (top) and the inverted observation image (bottom), neural retinal characteristics (continuous epithelial structure and a two-layer structure consisting of ONbL and INbL) are observed throughout the entire circumference, while non-neural retinal characteristics are not observed at any position in the samples. Thus, it was confirmed that observing cell aggregates from both upright and inverted observation images allows for the selection of regions with high purity of target cells. Therefore, it was confirmed that observing cell aggregates from both upright and inverted observation images from both upright and inverted observation images is also an effective method for determining neural retinal sheets.

[0104] Observation Method 3: Observation Containers for Upright and Inverted Microscope Observation We investigated observation containers suitable for upright and inverted microscope observation. While flasks, petri dishes, and multiwell plates are considered to be possible culture vessel shapes, we found that multiwell plates are preferable, and that multiwell plates with a concave V- or U-bottom are even more preferable. This is because the use of multiwell plates with such shapes allows cell aggregates to remain stationary even in liquid.

[0105] Sample SPL8 was obtained by selecting one of the cell aggregates prepared by the method described in Observation Method 1. Sample SPL8 was transferred to container PLT1: a 96-well plate (U-bottom (slit well)), container PLT2: a 96-well plate (V-bottom), container PLT3: a 24-well plate, container PLT4: a 6-well plate, container PLT5: a 60 mm dish, and container PLT6: a T25 flask, and observed as a bright-field image (Hoffmann interference, objective lens: 10x magnification, tiling 2 × 2) using an inverted microscope (Olympus IX-83), and further observed using an upright microscope (Carl Zeiss) to obtain inverted and upright observation images. Observation images of the obtained sample SPL8 taken while it was held in container PLT1: a 96-well plate (U-bottom (slit well)), container PLT2: a 96-well plate (V-bottom), container PLT3: a 24-well plate, container PLT4: a 6-well plate, container PLT5: a 60 mm dish, and container PLT6: a T25 flask are shown in Figures 11 to 16. In each figure, the upper image is an upright observation image, and the lower image is an inverted observation image.

[0106] As a result, in all of the culture vessels, vessel PLT1: 96-well plate (U-bottom (slit well)), vessel PLT2: 96-well plate (V-bottom), vessel PLT3: 24-well plate, vessel PLT4: 6-well plate, vessel PLT5: 60 mm dish, and vessel PLT6: T25 flask, the characteristics of retinal tissue could be captured, and RPE cells could be confirmed in at least one of the upright observation images or the inverted observation images. This confirmed that the technique of combining upright and inverted observation can be used regardless of the shape of the vessel.

[0107] Furthermore, for vessel PLT1: 96-well plate (U-bottom) and vessel PLT2: 96-well plate (V-bottom), the observation angle (angle at which the cell aggregates are viewed) of the cell aggregates was consistent between the inverted and upright observation images, whereas for the other culture vessels, slight changes in the observation angle of the cell aggregates were observed between the two images. This is thought to be because, when the vessel was moved between the upright and inverted observation images, the cell aggregates were free to move freely in vessels other than vessel PLT1: 96-well plate (U-bottom), given the size of the cell aggregates used, causing the cell aggregates to rotate slightly during movement. Therefore, it was confirmed that selecting an observation vessel that combines the characteristics of a size appropriate for the three-dimensional tissue to be observed and the characteristics of a well plate with a downward convex bottom shape enables more accurate observation from both directions. Specifically, it was confirmed that vessels PLT1 and PLT2 were suitable for sample SPL8. Generally, the larger the number of wells in a vessel, the smaller the well size, making it possible to process more cell aggregates at one time. Therefore, by selecting wells of an appropriate size that are not too large for the size of the cell aggregate to be evaluated, it is possible to make the observation angles of the cell aggregate in the inverted observation image and the upright observation image as similar as possible, and to increase the number of wells and improve observation throughput.

[0108] The liquid volume of the culture vessel is, for example, 0.05 to 0.8 mL / cm 2 , preferably 0.15 to 0.6 mL / cm 2 For example, when 100 μL is placed in a 96-well plate, the volume is approximately 0.3125 mL / cm 2 For example, if 10 mL is placed in a 90 mm dish, the volume is approximately 0.17 mL / cm 2 This is the most suitable amount of liquid.

[0109] The culture vessel can be made of plastic or glass, but transparent plastic is preferable. This is because plastic is easy to process and can be coated to facilitate cell culture. The vessel must be colorless and transparent, and plastic culture dishes can generally be used.

[0110] Observation Method 4: Observation of various types of three-dimensional tissues using upright and inverted microscopes We investigated whether observation using upright and inverted microscopes is applicable to three-dimensional tissues other than retinal tissue obtained during culture. We also verified whether each three-dimensional tissue can be distinguished using upright and inverted microscopes.

[0111] Cell aggregates were obtained using the method described in Observation Method 1. Cell aggregates containing tissues with different morphological characteristics, namely, Tissue 1 (retinal tissue), Tissue 2 (RPE), Tissue 3 (ciliary body tissue), Tissue 4 (eye stalk), Tissue 5 (telencephalon tissue), and Tissue 6 (spinal cord tissue), were selected and obtained as Samples SPL9 to SPL14. These were observed using the method described in Observation Method 2, and inverted and upright observation images were obtained according to the procedure described above. The obtained observation images of Samples SPL9 to SPL14 were appropriately enlarged to show the respective tissues in Figures 17 to 22, respectively. In each figure, the upper image is the upright observation image, and the lower image is the inverted observation image.

[0112] In addition to the color characteristics of retinal tissue confirmed in both upright and inverted images, the inverted images also confirmed the presence of granular-shaped cells arranged in rows, with individual cells clearly distinguishable. Furthermore, a comparison of the morphology and quality of many obtained cell aggregates confirmed that, in order to obtain cell aggregates containing normal regions with no quality issues, it is desirable for cell aggregates with a diameter of 1 mm or more to have an ONbL with a width of 0.03 mm or more and a length that extends continuously at 45 degrees or more from the center of the cell aggregate. Cell aggregates typically have a shape resembling the fusion of multiple ellipses. The center of a cell aggregate is preferably the center of a shape approximated by a predetermined method. Specifically, the center of a cell aggregate can be the center of a reference ellipse, which is the smallest ellipse that allows the cell aggregate to touch its outline at the greatest number of points. A reference ellipse of this shape fits immediately inside the outline of the cell aggregate and touches the outline at as many points as possible (usually four or more), thereby providing a good approximation of the shape of the cell aggregate. In this way, if an outer neuroblast layer with a width of 0.03 mm or more is present continuously at an angle of 45 degrees or more from the center of the cell aggregate, it can be confirmed that the retinal tissue is of good quality, as long as it does not contain any undesired cells.

[0113] The RPE was observed as a tissue with a color tone close to black in both upright and inverted images. However, it was confirmed that the ability to distinguish the black pigment of the RPE from the black color exhibited by low light transmittance due to discontinuities in the tissue structure, such as inside aggregates, was better in upright images than in inverted images.

[0114] The ciliary body tissue was observed as a grape-like tissue with a wavy surface in both upright and inverted images. It was confirmed that the upright image allowed for better observation because the ciliary body tissue was in focus over a slightly wider range in the vertical direction.

[0115] The eyestalk exhibits similar color characteristics to retinal tissue in both upright and inverted images, while the inverted image shows an arrangement of elongated cells, confirming that it can be identified from the morphology of retinal tissue.

[0116] In telencephalic tissue, a bright layer can be seen at the outermost side in both upright and inverted images, but it was confirmed that the width of the bright layer was smaller than in retinal tissue.When comparing telencephalic tissue and retinal tissue, the difference in layer width was more pronounced in inverted images than in upright images, confirming high discrimination.

[0117] In both upright and inverted images, the spinal cord tissue was observed as having an orderly structure with an uneven surface. The spinal cord tissue had a moderate degree of optical transparency, and in upright images, the entire tissue appeared grayish, demonstrating greater distinguishability from telencephalic tissue than in inverted images.

[0118] These results confirmed that inverted observation images, by focusing vertically on epithelial tissue, can capture structural information about tissues and the shapes of the cells that make up the tissues with high resolution, while upright observation images, by being able to focus over a wide range vertically, are excellent for capturing three-dimensional tissues, and by making it easier to obtain brightness contrast, can capture tissues with low light transparency with greater accuracy. Therefore, it was confirmed that observation using both an inverted microscope and an upright microscope not only makes it possible to observe three-dimensional tissues as a whole from both directions, but also improves the accuracy of detecting multiple types of tissue by taking advantage of the characteristics of the images obtained with each observation method.

[0119] <Observation Method 5: Observation of Non-Neural Retinal Tissues Using Upright and Inverted Microscopes> We investigated whether upright and inverted microscope observations are applicable to three-dimensional non-neural retinal tissues obtained using a culture method different from the method for producing neural retina.

[0120] Cell aggregates were prepared by the method described in Observation Method 1, except that the medium added on day 3 after the start of suspension culture was changed to a serum-free medium that did not contain human recombinant BMP4.

[0121] Three non-neural retinal tissues were selected from the prepared cell aggregates and obtained as samples SPL15 to SPL17. The method described in Observation Method 3 was applied to these samples to obtain inverted and upright observation images. The images obtained for samples SPL15 to SPL17 are shown in Figures 23 to 25, respectively. In each figure, the upper image is the upright observation image, and the lower image is the inverted observation image.

[0122] The morphology and color of the cell aggregates were observed and evaluated based on Figures 23 to 25. Morphological evaluation began by observing the entire cell aggregate in upright observation, which allows for a wider vertical focus range, to confirm whether the aggregate exhibited a characteristic bright outer layer and dark inner layer. Next, the periphery was observed in inverted observation, which allows for more detailed structural observation, to confirm whether the tissue, in which individual cells are clearly distinguishable as grains, was arranged in a regular, continuous manner in the tangential direction. Next, the absence of non-target cells in the cell aggregate was confirmed by checking whether the cell aggregate contained normal retinal tissue regions by checking whether non-target cells were present above the cell aggregate in upright observation and below the cell aggregate in inverted observation. When samples SPL15 to SPL17 were examined using the above procedure, neither upright nor inverted observation images confirmed the characteristics of neural retina (a continuous epithelial structure and a two-layer structure consisting of ONbL and INbL), confirming that they were non-neural retina. Thus, it was confirmed that bidirectional observation using upright and inverted images can be applied to non-neural retinal tissues obtained by manufacturing methods different from those for the neural retina, and that such tissues can be clearly identified as non-neural retinal tissues. In other words, it was confirmed that bidirectional observation using upright and inverted images can be applied to appropriately identify not only the retina but also three-dimensional tissues of various regions and organs throughout the body.

[0123] <Observation Method 6: Study of Observation Methods for Inverted and Upright Observation Images> Further study was conducted on the specific microscope settings for the observation methods. Below, general settings for upright and inverted microscopes and their features are explained.

[0124] An upright microscope (bright-field observation) is characterized by a wide range of focus and the ability to acquire information across a wide range in the Z direction. Furthermore, a stereo microscope allows for stereoscopic viewing with both eyes. Furthermore, when using a zoom microscope, it is characterized by the ability to easily switch to an appropriate magnification. For upright microscope observation, for example, an objective lens with a magnification of 0.5 to 2.0 times and a magnification of 2 to 200 times can be used. Furthermore, either vertical incident light illumination or oblique illumination can be used as the illumination method for the light source.

[0125] Inverted microscopes (bright-field observation, transmitted light) have a narrow focus range but are characterized by their ability to observe subjects in detail. The following combinations of magnification and imaging method can be used: 4x phase contrast, 4x differential interference, 4x oblique, 10x Hoffmann interference, 10x phase contrast, 10x differential interference, 10x oblique, 20x Hoffmann interference, 20x phase contrast, 20x differential interference, and 20x oblique. The following describes the main features of the combinations of magnification and imaging method: 4x phase contrast: Most commonly used for cell observation. It allows for excellent observation of structures within epithelial tissue. It provides stable and clear observation of both retinal cells and non-target cells with moderate contrast. The low magnification makes it easy to observe and evaluate the entire cell aggregate. 4x oblique: It allows for excellent observation of shadows and cell arrangement within epithelial tissue. The low magnification makes it easy to evaluate the entire cell aggregate.・10x Hoffmann interference: Produces an image similar to differential interference, with a light-dark gradient across the entire sample, making it suitable for observing thick, three-dimensional tissues. It provides the best observation of structures within epithelial tissue. The large light-dark gradient in the observed image makes it easier to spot the black-stained RPE. However, due to the large light-dark gradient, there are slight differences in the appearance of the same type of cell (such as the difference in brightness between the outside and inside, the visibility of the boundary, and the ease of capturing the granularity of photoreceptors) depending on the vertical, horizontal, and vertical positions of the observation area. Furthermore, the large 10x objective lens magnification may make it impossible to observe the entire cell aggregate, and the narrow focal range makes it difficult to focus on the entire periphery of the cell aggregate. Similar images can also be observed with differential interference. ・10x phase contrast: Facilitates observation of structures within epithelial tissue. However, focusing is difficult. ・10x polarized: Facilitates observation of the shadows and cell arrangement of epithelial tissue.

[0126] Cell aggregates were prepared using the method described in Observation Method 1, and two of them were selected and acquired as samples SPL18 and SPL19. Sample SPL18 was observed using an upright microscope (Carl Zeiss, STEREO DISCOVERY V12) at six zoom magnifications: 12, 20, 32, 50, 80, and 100x. Sample SPL18 was also observed using an inverted microscope (Olympus, IX-83) at three different magnifications: bright-field imaging (phase contrast, 4x objective), bright-field imaging (phase contrast, 10x objective), and bright-field imaging (Hoffmann interference, 10x objective, 2x2 tiling). The resulting images are shown in Figures 26 to 34, respectively. The following table shows the correspondence between the observed images and the figures. Note that the scales are not consistent between the figures.

[0127]

[0128] 26 to 31 show upright images at different magnifications. Each sample exhibits a distinctive color tone, with the outer surface bright and the inner surface dark, confirming the presence of neural retina. Each sample also contains a dark interior, confirming the presence of RPE. However, the upright images at 12x and 20x magnifications shown in FIGS. 26 and 27 do not allow for a sufficient image of the sample. Therefore, magnifications ranging from 32x to 100x, as shown in FIGS. 28 to 31, are preferred.

[0129] When examining the inverted observation images in Figures 32 to 34, regardless of the observation method or magnification, the characteristic color tone of the neural retina, which is bright on the outside and dark on the inside, can be confirmed, and the presence of a black RPE inside can also be confirmed.

[0130] Sample SPL19 was observed using an upright microscope (Carl Zeiss, STEREO DISCOVERY V12) at three zoom magnifications: 25x, 50x, and 80x. Sample SPL19 was also observed using an inverted microscope (Olympus, IX-83) at three different magnifications: bright-field image (phase contrast image, objective lens: 4x), bright-field image (phase contrast image, objective lens: 10x), and bright-field image (Hoffmann interference, objective lens: 10x, 2x2 tiling). The resulting images are shown in Figures 35 to 40, respectively. The following table shows the correspondence between the observed images and the figures. Note that the scales are not consistent between the figures.

[0131]

[0132] The upright observation images at different magnifications shown in Figures 35 to 37 show that all samples exhibit a bright outer and dark inner color tone, confirming the presence of neural retina. Furthermore, all samples exhibit a dark inner color tone, confirming the presence of RPE. While the upright observation image at 25x magnification shown in Figure 35 shows a slightly smaller sample size, the color tone can be seen in some detail. Magnifications of 50x to 80x, as shown in Figures 36 to 37, are more suitable.

[0133] When examining the inverted observation images in Figures 38 to 40, regardless of the observation method or magnification, it is possible to confirm the characteristic color tone of the neural retina, which is bright on the outside and dark on the inside, and the presence of a black RPE inside.

[0134] 26 to 40, it was confirmed that it was possible to determine whether the cell aggregate contained neural retina and RPE under any of the observation conditions. It was confirmed that, in order to observe the entire cell aggregate with higher accuracy, a zoom magnification of 32 to 100x is preferable for upright observation, and a 2x2 tiling with a 4x or 10x objective lens is preferable for inverted observation.

[0135] These results demonstrate that observing three-dimensional tissues with both an upright microscope and an inverted microscope and obtaining both upright and inverted images enables efficient and accurate observation of the entire cell aggregate. Taking advantage of the unique features of both the upright and inverted microscopes, the characteristics of various types of tissue contained in the cell aggregate can be detected with greater precision.

[0136] <Configuration of the Normal Cell Aggregate Determination System 100> As described above, it has been confirmed that the combined use of upright and inverted observation images enables highly accurate identification of cell aggregates containing normal retinal tissue regions. Further investigation of a computer-based machine-based method for evaluating cell aggregate images confirmed that normal retinal tissue can be extracted by analyzing upright and inverted observation images of cell aggregates containing neural retina. It has also been confirmed that combining two types of images (upright and inverted) for determination improves the accuracy of the determination, and that integrating the analysis of the two types of images after positional correspondence enables more precise analysis and location-specific detection of retinal tissue. Furthermore, it has been confirmed that this combined analysis of upright and inverted observation images is a highly versatile method applicable not only to three-dimensional tissues including the retina, but also to various three-dimensional tissues such as non-retinal neural tissue and RPE. A system for determining normal retinal tissue regions from upright and inverted observation images of cell aggregates using machine-based determination will now be described. A normal cell aggregate determination system 100 according to an embodiment of the present invention will be described with reference to the drawings. 41 is a schematic diagram showing the configuration of a normal cell aggregate determination system 100 according to an embodiment of the present invention. The normal cell aggregate determination system 100 typically comprises a normal cell aggregate determination server 110, an upright microscope 200, an inverted microscope 210, and an operation terminal 220, which are connected to each other via a local area network LAN. Note that instead of the upright microscope 200 and the inverted microscope 210, it is also possible to use a digital camera 215 (not shown) equipped with an optical system capable of taking magnified photographs.

[0137] Figure 42 shows a block diagram of the normal cell aggregate determination server 110. The normal cell aggregate determination server 110 is typically in the form of a server computer on which a computer program that realizes its functions is installed. The normal cell aggregate determination server 110 is broadly composed of a control unit 111, a memory 120, and a communication interface 130. The control unit 111 includes, as components having more specific functions, an image data acquisition means 112, a candidate region identification means 113, and a normal region determination means 114. The normal region determination means 114 includes, as components having even more specific functions, a region portion evaluation means 114A and a continuity length-based normal region determination means 114B. The memory 120 stores a normal cell aggregate determination program 120A and normal cell aggregate determination data 120B.

[0138] The control unit 111 is a processing circuit for executing various functions that control the operation of the normal cell aggregate determination server 110, and is typically an information processing circuit composed of a processor (not shown) that causes the normal cell aggregate determination server 110 to perform the specified functions by executing a computer program for realizing the specified functions, and a RAM (not shown) that serves as a temporary data storage area for the processor.

[0139] The memory 120 is typically a nonvolatile storage device such as a hard disk drive or a solid-state drive (SSD), and stores computer programs and data referenced during execution of the programs. When a computer program is executed, basic software such as an operating system (OS) is typically used. However, the functions of the OS are included in the function of the control unit 111 to execute the computer program, and therefore will not be described here. The functions of the normal cell aggregate determination server 110 are realized by the computer program being executed by the control unit 111, thereby configuring a specific information processing circuit corresponding to the function. The memory 120 stores a normal cell aggregate determination program 120A as a computer program, and stores normal cell aggregate determination data 120B as data referenced during execution of the normal cell aggregate determination program 120A. The normal cell aggregate determination data 120B includes specific values ​​of various determination criteria, and is read from the memory 120 and referenced by the control unit 111 when a determination based on the criteria is performed. Note that normal cell aggregate determination data 120B can also be incorporated as part of the program logic within normal cell aggregate determination program 120A, in which case normal cell aggregate determination data 120B does not necessarily have to be stored as individual data in memory 120. When control unit 111 reads and executes normal cell aggregate determination program 120A from memory 120, a specific information processing circuit is configured that realizes various functions related to normal cell aggregate determination, and this circuit executes operations that realize those functions. That is, this configures a specific information processing circuit comprising image data acquisition means 112, candidate region identification means 113, and normal region determination means 114 including region portion evaluation means 114A and continuity length-based normal region determination means 114B. Although not shown, memory 120 also has a data storage area for storing arbitrary data such as image data and calculation results.

[0140] Image data acquisition means 112 is an information processing circuit that realizes the function of acquiring image data of an observation image of a cell aggregate to be determined. More specifically, image data acquisition means 112 is an information processing circuit that realizes the function of acquiring image data of the cell aggregate to be determined via a LAN as image data of an upright observation image and an inverted observation image taken using upright microscope 200 and / or inverted microscope 210, respectively, and storing the image data in a data storage area of ​​memory 120. Note that when an image of a cell aggregate to be determined is taken using digital camera 215, image data acquisition means 112 acquires image data of the observation image taken using digital camera 215.

[0141] The candidate region identification means 113 is an information processing circuit that realizes the function of identifying candidate regions included in regions that are candidates for normal retinal tissue within a cell aggregate in an observation image represented by image data. More specifically, the candidate region identification means 113 is an information processing circuit that realizes the function of identifying, within an image region of the cell aggregate in an inverted observation image and / or an upright observation image (a region corresponding to a portion where the cell aggregate is present in the inverted observation image or the upright observation image), a region that is determined to satisfy candidate region criteria for the inverted observation image and / or the upright observation image based on candidate region criteria related to colors that correspond to the observation method (inverted observation, upright observation, etc.), as a candidate region included in a region that is a candidate for normal retinal tissue.

[0142] The images of cell aggregates in the inverted observation image and / or the upright observation image may be in chromatic colors or achromatic colors such as grayscale. When an image expressed in chromatic colors including blue, green, and red is used, brightness, hue, and saturation can be used as color-related candidate region determination criteria. When an image expressed in grayscale is used, brightness alone can also be used as color-related candidate region determination criteria.

[0143] A candidate region for a normal retinal tissue region is a region that can be determined as a normal retinal tissue region at a later stage by satisfying the candidate region determination criteria. A candidate region is a region that is included in part or all of such a candidate region for a normal retinal tissue region, and if it satisfies the predetermined criteria at a later stage, it can be determined as a normal retinal tissue region, or both it and its adjacent regions can be determined as normal retinal tissue regions. A candidate region is typically an ONbL, and if it satisfies the predetermined criteria at a later stage, the ONbL and the INbL adjacent to it on the inside (toward the center of the cell aggregate) can be determined as normal retinal tissue regions.

[0144] The normal region determination means 114 is an information processing circuit that realizes the function of determining whether a cell aggregate has a normal retinal tissue region. More specifically, the normal region determination means 114 is an information processing circuit that realizes the function of determining, in the image region of the cell aggregate in an inverted observation image and / or an upright observation image, a region portion of a candidate region that is determined to satisfy the normal region determination criteria corresponding to the observation method (inverted observation, upright observation, etc.), as a normal retinal cell region, based on normal region determination criteria related to brightness and / or morphology corresponding to the observation method. The normal region determination means 114 includes, as information processing circuits that realize more specific functions, a region portion evaluation means 114A and a continuity length-based normal region determination means 114B.

[0145] The region portion evaluation means 114A is an information processing circuit that realizes a function of calculating an evaluation score for a region portion included in a candidate region of epithelial tissue. More specifically, the region portion evaluation means 114A is an information processing circuit that realizes a function of calculating an evaluation score for each region portion present in a range obtained by dividing a candidate region into multiple regions along a curve representing the extension direction of the candidate region, based on evaluation criteria related to color and / or morphology corresponding to the observation method, for each candidate region identified in an image region of a cell aggregate, positionally associated as necessary, in each of an inverted observation image and / or an upright observation image. The evaluation criteria can be width-related criteria such as the width of the candidate region when viewed along the extension direction of its shape, the width of the overlap between the candidate region and a line passing through the center of the cell aggregate or its reference ellipse, or length-related criteria such as the continuous length of a region portion having a predetermined width or more. Because the candidate region is located near the outline (periphery) of the cell aggregate, the curve representing the extension direction of the candidate region usually follows the outline or periphery of the cell aggregate.

[0146] Positional correspondence is a process performed when both inverted and upright observation images are used to calculate the evaluation score for a region. It involves associating an arbitrary position within the image region of a cell aggregate in each of the inverted and upright observation images of the same cell aggregate so that it corresponds to the same position on the cell aggregate as viewed from the vertical direction. To achieve this, first, one of the inverted and upright observation images is mirror-inverted so that the positions of the cell aggregate in the image region corresponding to the cell aggregate in the upright and inverted observation images match. Furthermore, it is preferable that the inverted and upright observation images have the same magnification (i.e., the same object is represented by an image region with the same number of pixels). In this case, one of the images is enlarged or reduced as necessary to achieve the same magnification.

[0147] If necessary, image transformation operations such as translation and rotation are performed on at least one of the image regions of the cell aggregates so that the positions of the same pixels in each image region corresponding to the cell aggregate correspond to the same positions on the cell aggregates viewed from a vertical direction. This operation may be performed using a known feature point matching method consisting of feature point detection, feature description, and matching steps so that the contours of the image regions of the cell aggregates match. Alternatively, it may be performed visually by an operator. Furthermore, by integrating an inverted microscope and an upright microscope and configuring them to capture inverted and upright observation images without moving the container holding the cell aggregate, the amount of translation and rotation (and magnification / reduction) can be constant, and the above-mentioned image transformation operations can be performed as predetermined transformation operations.

[0148] Furthermore, even if transformation operations such as translation and rotation are not performed on the images, functions representing those transformation operations (direction / distance of translation, angle of rotation) may be stored and applied when comparing those images. If the inverted observation image and the upright observation image have different magnifications, it is advisable to store the enlargement / reduction ratio for one of the images as a function of the transformation operation so that the images have the same magnification when compared.

[0149] The continuous length-based normal region determination means 114B is an information processing circuit that realizes the function of determining a region as a normal retinal tissue region when a region having an evaluation score equal to or greater than a reference value exists continuously for a predetermined reference length or more. More specifically, the continuous length-based normal region determination means 114B is an information processing circuit that realizes the function of determining, as a normal retinal tissue region, a region having an evaluation score equal to or greater than a normal region determination reference value corresponding to the observation method (inverted observation, upright observation) in an inverted observation image and / or an upright observation image, the region being determined to exist continuously for a predetermined normal region determination reference length or more in a direction along a curve representing the extension direction of the region. The predetermined normal region determination reference length is not limited to the length of the curve representing the extension direction of the shape of the region, but can also be the angle of the range in which the region exists as viewed from the center of a reference ellipse, which is the smallest ellipse that touches the cell aggregate at the most points (the central angle when the region is approximated as an arc). Note that the predetermined reference length should be determined based on the actual physical length, but it may also be calculated by converting it to other units, such as the number of consecutive pixels in the image.

[0150] The communication interface 130 is an interface for connecting the normal cell aggregate determination server 110 to a local area network LAN to transmit and receive data. The normal cell aggregate determination server 110 is connected to the local area network LAN by the communication interface 130 in a wired or wireless manner, and can perform data communication with other components such as the upright microscope 200, the inverted microscope 210, and the operation terminal 220 that are connected to the local area network LAN.

[0151] The upright microscope 200 is a microscope equipped with an optical system for performing upright magnification observation and a camera for capturing magnified images. Image data captured by the camera can be transmitted to the normal cell aggregate determination server 110 via a LAN from a communication interface built into the upright microscope 200 (or a terminal such as a laptop PC connected to the upright microscope 200).

[0152] The inverted microscope 210 is a microscope equipped with an optical system for performing inverted magnification observation and a camera for capturing magnified images. Image data captured by the camera can be transmitted to the normal cell aggregate determination server 110 via a LAN from a communication interface built into the inverted microscope 210 (or a terminal such as a laptop PC connected to the inverted microscope 210). Images are not limited to those obtained from the inverted microscope 210 and the upright microscope 200. Instead of such microscopes, a digital camera 215 (e.g., a CMOS (Complementary Metal Oxide Semiconductor) camera or a CCD (Charge-Coupled Device) camera) having an optical system capable of capturing magnified images can also be used. The digital camera 215 preferably has an autofocus function and a focus position adjustment function. In this case, the cell aggregate is photographed using the digital camera from both directions of the cell aggregate along the approximately vertical direction, and image data is acquired. As a result, for example, a Z-stack image reconstructed from stack images taken by gradually shifting the focal plane from the initial autofocused focus can be used as an image containing three-dimensional information. Furthermore, the normal cell aggregate determination server 110 may also have an image selection unit (not shown) that selects an image to be used for determination from multiple images taken by gradually shifting the focal plane in the Z direction based on predetermined criteria or user selection, and may be configured to select an appropriate image through its operation. The selection criteria may be based on various perspectives, such as the clarity of the entire image or specific regions, the presence or absence of objects of a specific shape, and their shape and color.

[0153] The operation terminal 220 is a component that accepts operation inputs for the operator to operate the normal cell aggregate determination server 110, transmits the inputs to the normal cell aggregate determination server 110, and outputs the determination results transmitted from the normal cell aggregate determination server 110. The operation terminal 220 is equipped with input means such as a keyboard and a mouse and output means such as a display, and is also capable of displaying a graphical user interface on the display to allow the operator to input and output information to and from the normal cell aggregate determination server 110.

[0154] Furthermore, the normal cell aggregate determination system 100 may be connected to an automatic cutting device as part of an automatic retinal tissue cutting device. By transmitting position information of the normal retinal tissue region determined by the normal cell aggregate determination system 100 to the automatic cutting device, the automatic cutting device can automatically cut out the normal retinal tissue region from the cell aggregate. The automatic cutting device has a control unit that controls various operations for cutting based on the position information obtained from the normal cell aggregate determination system 100, and the control unit is connected to at least a cutting device such as a cutter whose operation can be controlled, and a holding tool for holding the cell aggregate whose operation can be controlled, and the control unit controls the holding tool and cutter based on the position information of the normal retinal tissue region to hold the cell aggregate and cut out the retinal tissue. The automatic cutting device can also be configured to have a well position control device that moves the desired well to a working area for cutting, and by the normal cell aggregate determination system 100 transmitting information about wells holding cell aggregates including normal retinal tissue areas to the automatic cutting device, the automatic cutting device can sequentially move only the wells holding cell aggregates including normal retinal tissue areas to the working area and cut out the normal retinal tissue areas from there.

[0155] The present invention will be specifically described below using four examples, but the present invention is not limited to these. Cell aggregates were prepared using the method described in Observation Method 1, and 12 of them were obtained as Samples SPL20 to SPL31. The procedures of the examples described below were carried out for each of them. In all examples, machine judgment was performed on the observed images of the cell aggregates, but different procedures were carried out in Examples 1 to 4 depending on the combination of observation methods and observed image processing methods. The following table lists descriptions of the four examples and the corresponding figures.

[0156]

[0157] Example 1: Application of machine judgment to inverted observation images of cell aggregates Whether automatic judgment is possible by applying machine judgment to inverted observation images was examined. The overall operation flow is shown in Figure 43, and the detailed operation flow related to normal region judgment is shown in Figure 44. Samples SPL20 to SPL31 were used as cell aggregates to be judged, and the following steps were carried out.

[0158] [1] Image Data Acquisition Step: A container holding a cell aggregate to be evaluated is placed on the inverted microscope 210, and the cell aggregate is photographed using the inverted microscope 210. Preferably, the cell aggregate is photographed while suspended in a liquid, such as a culture medium, in a plastic container having a downwardly convex or recessed shape and / or one or more isolated culture areas. The image data of the resulting inverted observation image is acquired by the image data acquisition means 112 from the inverted microscope 210 via a LAN (step S101A). Specifically, the operator observes a bright-field image (phase-contrast image, objective lens: 4x magnification) of the cell aggregate to be evaluated, held in a culture medium in an appropriate container, using the inverted microscope 210, and photographs the image to obtain an inverted observation image. The inverted microscope 210 transmits image data representing the inverted observation image to the normal cell aggregate evaluation server 110 via the LAN, and the image data acquisition means 112 acquires and stores the image data in a data storage area in the memory 120. The upper part of FIG. 51 shows the acquired inverted observation image of the cell aggregate to be evaluated.

[0159] Next, the acquired inverted observation image is subjected to an analysis of the inverted observation image, which will be described in detail below, to obtain the analysis results of the inverted observation image. In the steps (1) to (6) described below, image processing and image analysis were performed using Pipeline Pilot Version 22.1.0, a software construction tool from Dassault System, as part of the normal cell aggregate determination program 120A.

[0160] [2] Candidate Region Identification Step: The candidate region identification means 113 identifies, within the image region of the cell aggregate in the inverted observation image, regions determined to satisfy the candidate region criteria for color based on the candidate region criteria as candidate regions included in the region that is a candidate for the normal retinal tissue region (step S102A). In this step, an image region corresponding to the cell aggregate (i.e., the image of the cell aggregate itself) is identified, and a candidate region is identified from among the identified image regions based on the candidate region criteria for color. Identification of the image region corresponding to the cell aggregate can be performed, for example, by image processing based on characteristics such as brightness and morphology. Furthermore, by excluding the periphery (near the four corners) of the observed image, it is possible to exclude regions clearly outside the cell aggregate, thereby identifying the image region corresponding to the cell aggregate. Here, the candidate region is typically the ONbL region. This is because the epithelial structure of the neural retina has a bright, arc-shaped ONbL on the outside and a dark INbL on the inside, and therefore the ONbL is included as part of the normal retinal tissue region. By establishing candidate region determination criteria based on color characteristics, such as bright color tone, it is possible to quickly identify ONbLs as candidate regions from image regions of cell aggregates. It is preferable to use only brightness as a color characteristic. A specific brightness criterion can be a predetermined degree (relative value, absolute value) higher than the maximum or average brightness of the central region or the entire region of the image region of the cell aggregate. It is also possible to identify candidate regions by first performing an identification process based on candidate region determination criteria related to color, and then identifying the image region of the cell aggregate and excluding other image regions. Specifically, the candidate region identification step can be performed by performing the following steps (1) to (3).

[0161] (1) The candidate area identification means 113 converts the image data of the inverted observation image, which is represented as an RGB image, into a black-and-white image with 256 gradations. Since the brightness information of the color is typically used as the criterion for determining the candidate area, this procedure removes the hue and saturation information from the image data.

[0162] (2) The candidate region identification means 113 uses known image processing techniques to reduce noise in the inverted observation image, identify regions that meet a certain brightness standard, and merge adjacent regions. Regions near the center of the cell aggregate are characterized by dark color and little gradation change due to the presence of INbL, while regions near the periphery of the cell aggregate are characterized by bright color and little gradation change due to the presence of ONbL. Through this procedure, the candidate region identification means 113 extracts regions with large gradation changes as the circumferential outline of the cell aggregate, and identifies relatively bright regions within this outline as potential candidate regions, identifying and marking these as contiguous regions. Here, when merging regions, in addition to brightness, morphological characteristics based on a predetermined viewpoint, such as a certain number of contiguous pixels, can also be used as criteria. For example, bright regions that are not contiguous by two or more pixels in any of the vertical, horizontal, or diagonal directions can be excluded from region merging. This makes it possible to remove extremely small noise, etc.

[0163] Furthermore, since candidate regions typically exist along the periphery of cell aggregates, a morphological feature that the outermost contour of the cell aggregate image region of the region to be identified substantially matches the contour of the cell aggregate image region can be used as a candidate region determination criterion. Furthermore, since candidate regions typically exist along the periphery of cell aggregates without significant angular change, a morphological feature that the tangential angle of the contour of the region to be identified is continuous can be used as a candidate region determination criterion. Furthermore, since candidate regions typically have a certain width or more, a morphological feature that the width of the region to be identified as viewed from the center of the cell aggregate image region (the width between the outermost and innermost regions) is greater than a predetermined reference width can be used as a candidate region determination criterion.

[0164] Furthermore, non-target cells cannot be present within the candidate region. Therefore, a non-target cell identification step can be performed to identify whether or not non-target cells are present based on non-target cell identification criteria for determining whether non-target cells are present in the region to be identified, and the absence of non-target cells can be used as a morphological characteristic as a criterion for determining the candidate region. Examples of non-target cell identification criteria that can be used include disordered cell orientation, similarity to the characteristics of typical non-target cells in the eyestalk, RPE, telencephalon tissue, spinal cord tissue, etc., and a darker color than the average of the entire cell aggregate.

[0165] (3) The candidate region identification means 113 uses known image processing techniques to remove the outermost regions of the entire image, i.e., regions that are in contact with the four sides and have been marked as potential candidate regions, as noise, and removes the markings. This is to eliminate the adverse effects of objects that are not cell aggregates being captured on the outer edges of the image and that could potentially become noise. Since the previous procedures may leave large amounts of such noise adjacent to the four sides of the image, this procedure removes regions that are in contact with the four sides and could potentially become noise from the potential candidate regions.

[0166] These procedures allow the identification of candidate regions within the image region of the cell aggregate. The lower part of Figure 51 shows an inverted image of the cell aggregate to be evaluated, with the candidate regions marked slightly brighter. It is preferable to create an image in which the candidate regions are overlaid on the original inverted image, and store the image data in a data storage area in memory 120 as an output image for visual confirmation.

[0167] [3] Normal Region Determination Step Next, the normal region determination means 114 determines, in the candidate region of the inverted observation image, a candidate region to be identified that is determined to satisfy the normal region determination criteria for the inverted observation image based on the normal region determination criteria for the brightness and / or morphology of each candidate region of the inverted observation image, as a normal retinal tissue region (step S103A). A cell aggregate having such a normal retinal tissue region is determined to have no quality problems. This normal region determination step is executed in detail by steps [3A] and [3B], which will be described below. This will be described with reference to the detailed operational flow for normal region determination shown in Figure 44.

[0168] [3A] Region Part Evaluation Stage The region part evaluation means 114A calculates an evaluation score for each of the region parts present in the ranges obtained by dividing the candidate region into multiple parts along a curve representing the extension direction of the candidate region, based on evaluation criteria related to color and / or shape corresponding to the inverted observation technique, for the candidate region identified in the image region of the cell aggregate in the inverted observation image (step S151A). The evaluation criteria can be related to shape such as width or length, but it is preferable to use width as the evaluation criterion at this stage.

[0169] In this region sub-evaluation step, the candidate region is further divided into multiple regions, and an evaluation score is calculated for each of the region sub-regions in order to narrow down the candidate region to regions that are more likely to contain normal retinal tissue. This step is specifically performed by executing the steps (4) and (5) described below.

[0170] Here, the concepts of the length and width of a candidate region of a cell aggregate and its region will be explained. The actual physical width, expressed in units such as μm, determines the characteristics of a cell aggregate, and thus serves as the basis for judgment. However, in image analysis, the amount of calculation can be reduced by converting the actual physical length and width into the number of pixels. When performing analysis based on the number of pixels, if the direction in which the width extends is not a direction in which pixels are linearly aligned, such as the X-axis or Y-axis, it is possible to express the length in any direction in terms of the number of pixels by rotating the direction in which the width extends so that it is parallel to the X-axis or Y-axis and then counting the number of pixels. Hereinafter, when performing some judgment using the number of pixels, similarly, the actual physical length, etc. is the basis for the judgment. When analyzing image data of the observation image, the analysis is performed using the number of pixels corresponding to the reference physical length, etc., to reduce the amount of calculation. It is also possible to perform analysis by converting the number of pixels into the actual length or width, expressed in units such as μm.

[0171] Refer to the conceptual diagram for measuring the width W of the candidate region of the cell aggregate 10 shown in Figure 55. The candidate region of the cell aggregate 10 typically has an elongated shape, typically approximating an elliptical arc along the contour of the cell aggregate 10. In Figure 55, bright arc-shaped candidate regions can be seen in four locations: the upper left, upper right, lower right, and lower left. The curves representing the extension direction of the shape of each candidate region are indicated by dashed lines. The length of the candidate region CR1 indicated by the dashed-dotted line in the upper left of Figure 55 corresponds to the length of the curve CL1 representing the extension direction of the elongated shape. The curve CL1 is a curve connecting the central portions of the candidate region CR1 in the width direction. In this way, the curve representing the extension direction of the elongated shape can be, for example, a curve passing through the central portion of the shape in the width direction.

[0172] Figure 55 shows a reference ellipse RE, which is the smallest ellipse that lies within the contour of the cell aggregate 10 and touches the contour at the greatest number of points. The reference ellipse RE is typically an ellipse whose major axis is approximately aligned with the longitudinal direction of the entire cell aggregate. It is preferable that the reference ellipse RE touches the cell aggregate at least near four points: two intersections of the major axis and the ellipse, and two intersections of the minor axis and the ellipse. Such a reference ellipse RE closely approximates the overall shape of the corresponding cell aggregate. Here, the length of the candidate region may be expressed in units of length, such as the number of pixels along the length of a curve representing the direction of extension of the shape (curve CL1 for candidate region CR1), as described above. Alternatively, it may be expressed as the angle of the range in which the candidate region exists as viewed from the center of the cell aggregate (the central angle of the arc when the shape of the candidate region is approximated by an arc). Assume that lines R2 and R4 passing through the center CTR of the reference ellipse RE indicate the range in which candidate region CR1 exists. In this case, the length of the candidate region CR1 may be expressed as the angle β between the lines R2 and R4, which define the range in which the candidate region exists as viewed from the center CTR of the base ellipse RE. The angle β corresponds to the central angle when the shape of the candidate region CR1 (or the curve CL1) is considered to be an elliptical arc.

[0173] The width of a candidate region at a specific position is the width of the curve approximating it (curve CL1 for candidate region CR1) at that specific position (the width at which a line perpendicular to the tangent to the curve at that position overlaps with the candidate region). Here, the specific position of the candidate region can be expressed as an angle α, where a line R1 extending horizontally to the right from the center of the cell aggregate (i.e., the center of the reference ellipse RE) is set as the reference angle of 0 degrees, and a line R3 rotated leftward by a predetermined angle α from that reference angle, for example, is the intersection of the candidate region and the line R3. In Figure 55, candidate region CR1 has width W at the position of the candidate region at angle α.

[0174] (4) The region portion evaluation means 114A expresses a specific position along the extension direction of the candidate region as a counterclockwise angle, with a line extending horizontally to the right from the center of the cell aggregate, i.e., the center of a reference ellipse that approximates its shape, as a reference angle of 0 degrees, and determines the width of the candidate region at that specific position from 0 degrees to 360 degrees counterclockwise by incrementing the angle α by a predetermined angle, such as 1 degree. If the width is measured by incrementing the angle by 1 degree, the number of measurement positions is 360 degrees / 1 degree = 360.

[0175] (5) The region portion evaluation means 114A evaluates regions where the candidate region width is 15 pixels or more and the angle is 1 degree consecutively, as a Level 1 evaluation score, and regions where the candidate region width is 15 pixels or more and the angle is 2 degrees or more consecutively, as a Level 2 evaluation score. Because a candidate region must be wide enough to be included in the normal retinal tissue region, a width of 15 pixels or more in the inverted observation image is a necessary condition, and therefore, a Level 1 evaluation score is assigned to such regions. Furthermore, the longer the length of a region satisfying this width condition, the higher the likelihood that the candidate region will be included in the normal retinal tissue region. Therefore, a higher evaluation score is assigned to regions where the width condition of 15 pixels or more is satisfied for two or more consecutive degrees, resulting in a Level 2 evaluation score. As mentioned above, the width of the ONbL in the epithelial structure of retinal tissue can be determined by a width of 0.03 mm, and this was used as the width evaluation standard. In observation using an inverted microscope at the magnification used in this example, the actual size of the target region corresponding to one pixel was 2.07 μm square. Therefore, the judgment standard of 0.03 mm corresponds to approximately 15 pixels. The region portion evaluation means 114A then measured the width of the candidate region along the periphery and calculated the evaluation score. The calculated evaluation score is shown in a line graph in the region indicated by B in Figure 56. In the graph of the region indicated by B in Figure 56, the horizontal axis represents the angle of the measured region portion as seen from the center of the cell aggregate, with a line extending horizontally from the center of the cell aggregate to the right as the reference angle of 0 degrees, and increments of 1 degree in a counterclockwise direction over a range of 360 degrees. The vertical axis of the region indicated by B in Figure 56 represents the width of the region portion and the evaluation score, with the scale for the width of the region portion indicated on the left end and the scale for the evaluation score indicated on the right end. The width of the region portion is the width of the portion where the line passing through the center of the cell aggregate overlaps with the region portion, and is indicated by a dashed line in the graph. The evaluation score is indicated by a solid line in the graph, taking values ​​of 0, 1, or 2. The graph shown here can also be displayed on the display of the operation terminal 220 so that the operator can check it.

[0176] [3B] Continuous Length-Based Normal Region Determination Step: If the portions of a candidate region whose evaluation scores in the inverted observation image are equal to or greater than the normal region determination criterion value extend continuously along a curve representing the direction of extension for a length equal to or greater than a predetermined normal region determination criterion, the continuous length-based normal region determination means 114B determines the candidate region as a normal retinal tissue region (step S152A). That is, at this step, the candidate regions determined to have evaluation scores equal to or greater than the normal region determination criterion value are further narrowed down to regions that are more likely to contain normal retinal tissue, and a determination is made as to whether such candidate regions extend for a length equal to or greater than the predetermined normal region determination criterion. Specifically, this step is performed by executing the following procedure (6).

[0177] (6) When a region within a cell aggregate in an inverted image has a width of 15 pixels or more and its continuous length is an arc length corresponding to an angle of 45 degrees or more from the center of a reference ellipse approximating the cell aggregate (i.e., when the shape of the region is approximated by an arc, the central angle is 45 degrees or more), the candidate region can be determined to be a sufficiently large normal retinal tissue region with a length (the periphery of the candidate region along the outline of the cell aggregate) of approximately 0.4 mm or more in a cell aggregate with a diameter of 1 mm or more. Therefore, this was used as the normal region determination criterion length, and a cell aggregate having a candidate region with a length equal to or greater than this normal region determination criterion length was determined to contain the neural retina (normal retinal tissue region). According to this criterion, the continuous length-based normal region determination means 114B determines whether a region portion of a candidate region with an evaluation score of 2 or more has a length corresponding to a predetermined normal region determination criterion length of 45 degrees along a curve representing the extension direction of the candidate region, and determines a candidate region that meets this criterion as a normal retinal tissue region. The solid line indicates the angular range in which candidate regions with a width of 15 pixels or more and a continuous length corresponding to an arc length of 45 degrees or more are present as the determination result for the region indicated by A in Figure 56 . The dashed line indicates the corresponding position on the evaluation score graph for the solid line indicating the candidate region determination result for the region indicated by A in Figure 56 and the region indicated by B in Figure 56 . It can be seen that regions determined to be normal retinal tissue regions that meet the criteria exist at five positions corresponding to angles of approximately 28 degrees to 84 degrees, approximately 117 degrees to 167 degrees, approximately 172 degrees to 242 degrees, approximately 246 degrees to 296 degrees, and approximately 302 degrees to 351 degrees. Cell aggregates that contain one or more candidate region regions determined to be normal retinal tissue regions are determined to be of satisfactory quality. This determination result, along with information identifying the cell aggregate being determined, is stored in a data storage area in memory 120 for later reference. At this time, the angle at which the area portion determined to satisfy the criteria exists can be stored as information specifying the position of the normal retinal tissue area.

[0178] The results of applying the above-mentioned steps to 12 cell aggregates from sample SPL20 to sample SPL31 and assessing them are shown in Table 1. In the assessment results, 8 of the 12 samples were judged to have "good" quality since they had one or more normal retinal tissue regions, and 4 were judged to have "bad" quality since they did not have normal retinal tissue regions.

[0179]

[0180] To verify the validity of the results of the machine-based assessment shown in Table 1, we compared the results with those of visual inverted observation to determine whether the cell aggregate contained a normal retinal tissue region. As a result, samples with at least one candidate region with a width of 15 pixels and an arc length corresponding to a central angle of 45 degrees or more were judged to have "good" quality, similar to the machine-based assessment. Samples with zero such candidate regions were judged to have "bad" quality. Thus, the machine-based assessment results shown in Table 1 strongly correlated with actual visual assessment, confirming that the criterion of a candidate region with a width of 15 pixels and an arc length corresponding to a central angle of 45 degrees or more is appropriate as a criterion for determining whether the cell aggregate contained a normal retinal tissue region. Thus, applying machine-based assessment confirmed that the appropriateness of a cell aggregate could be determined from an inverted observation image. Therefore, machine-based assessment confirmed that the presence of a normal retinal tissue region could be accurately determined. However, because Example 1 did not take into account upright observation images, there is a possibility that the condition of the surface behind the surface of the cell aggregate observed by inverted observation may not be reflected in the assessment results.

[0181] Example 2: Application of machine judgment to upright observation images of cell aggregates Next, we investigated whether machine judgment can be applied to upright observation images. The overall operation flow is shown in Figure 45, the operation flow for constructing a machine learning model is shown in Figure 46, and the detailed operation flow for normal region judgment is shown in Figure 47. Samples SPL20 to SPL31 were used as cell aggregates to be judged, and the following steps were carried out, respectively.

[0182] [1] Image Data Acquisition Step First, a container holding a cell aggregate to be evaluated is placed in upright microscope 200, and the cell aggregate is photographed using upright microscope 200. A Carl Zeiss AxioZoom V16 was used as the upright microscope. Image data of the resulting upright observation image is acquired by image data acquisition means 112 from upright microscope 200 via LAN (step S101B). Specifically, similar to step S101A in Example 1, the operator uses upright microscope 200 to observe a bright-field image (zoom magnification: 33.3x) of a cell aggregate to be evaluated, which is held in a culture medium in an appropriate container, and photographs the image to obtain an upright observation image. Upright microscope 200 transmits image data representing the upright observation image to normal cell aggregate evaluation server 110 via LAN, and image data acquisition means 112 acquires the image data and stores it in a data storage area in memory 120. The upper part of Figure 53 shows the acquired upright observation image of the cell aggregate to be evaluated.

[0183] Next, the acquired upright observation image is subjected to an analysis of the upright observation image, which will be described in detail below, to obtain the analysis results of the upright observation image. In the steps (1) to (6) described below, image processing and image analysis were performed using Pipeline Pilot Version 22.1.0, a software construction tool from Dassault Systems, as part of normal cell aggregate determination program 120A.

[0184] [2] Candidate Region Identification Step: The candidate region identification means 113 identifies, in the image region of the cell aggregate in the upright and inverted observation image, regions determined to satisfy the candidate region determination criteria for color as candidate regions included in regions that are candidates for normal retinal tissue (step S102B). The purpose of this step is the same as step S102A in Example 1. However, in the upright observation image, compared to the inverted observation image, the image conditions, such as brightness and contrast, vary significantly due to differences in shooting conditions. Therefore, a procedure using the same criteria as in Example 1 could not produce good results. However, in the upright observation image, it was confirmed that good results could be obtained by performing discrimination using machine learning. Therefore, in order to automatically extract the ONbL region, which is a bright, arc-shaped region in the cell aggregate, machine learning was used to enable analysis and image processing even in upright observation images where the color tone is unclear. In this step, the image region corresponding to the cell aggregate is identified, and within that region, candidate regions are identified based on a machine learning model trained on the candidate region determination criteria for color. Here, brightness was the target of training for the machine learning model. As in Example 1, the candidate region can be an ONbL region. Alternatively, the candidate region can be identified by first performing an identification process based on a machine learning model that has learned candidate region determination criteria related to color, and then identifying the image region of the cell aggregate and excluding other image regions. Specifically, this candidate region identification step can be performed by performing the preparatory procedure shown in (1) below and the procedures (2) and (3) below.

[0185] (1) As a preparatory process for step S102B, a machine learning model for image identification processing is constructed. Prior to this, training data for training the machine learning model is prepared. The training data includes a plurality of upright observation images and corresponding training marking images. Here, three images are prepared for each. The upper side of FIG. 52 shows one upright observation image prepared as training data. The lower side of FIG. 52 shows the training marking image corresponding to that upright observation image. The training marking image was created by visually checking the upright observation image and marking the ONbL region, which is a bright arc-shaped portion, by filling it with white and maximizing the brightness so that it can be clearly distinguished. In other words, the ONbL region is identified and marked by brightness. Note that if the bright arc-shaped portion does not have a constant width, it is possible to include information based on shape as well as brightness in the training marking image by, for example, not marking that portion. In bright ONbL regions in upright observation images, the change in brightness across the width is ambiguous, and this can be approximated as following a Gaussian distribution. Therefore, a Gaussian image recognition model was used as the machine learning model, taking into account the color distribution characteristics of upright observation images. Gaussian image recognition is a technique in which the color distribution of a certain size of an image is learned, and a region in the image is searched for that region whose color distribution approximates the modeled Gaussian distribution of the target region, assuming that color changes follow a Gaussian distribution. It is possible to select other machine learning models related to appropriate image analysis techniques depending on the characteristics of the image.

[0186] The operational flow for constructing a machine learning model is shown in FIG. 46 . To train the machine learning model, multiple pairs of upright observation images and training marking images are sequentially input to the machine learning model. One upright observation image is input to the machine learning model of the candidate area identification means 113 (step S102B1). Next, a corresponding training marking image in which the ONbL region of the upright observation image is marked is input to the machine learning model of the candidate area identification means 113 (step S102B2). The machine learning model of the candidate area identification means 113 then compares the brightness distribution characteristics of the region in the upright observation image that should be determined as a candidate region corresponding to the ONbL region with the brightness distribution characteristics of the region indicated as an ONbL region by marking in the training marking image, thereby constructing a machine learning model for identifying a candidate area from the upright observation image (step S102B3). The determination mechanism realized by this trained machine learning model corresponds to the candidate area determination criterion. Once the machine learning model is constructed, no further steps for constructing the machine learning model are required.

[0187] (2) The candidate region identification means 113 uses a machine learning model to identify regions that are likely to be candidate regions that are likely to be determined to be included in ONbL in the upright observation image of the cell aggregate to be determined, and marks these as continuous regions. Through this procedure, the candidate region identification means 113 extracts regions with large changes in gradation as the circumferential outline of the cell aggregate, and identifies relatively bright regions within this outline whose brightness changes in the width direction like a Gaussian distribution as regions that are likely to be candidate regions, and identifies and marks these as continuous regions.

[0188] (3) The candidate region identification means 113 uses known image processing techniques to remove as noise the regions marked as potential candidate regions at the outermost part of the entire image, i.e., regions that contact all four sides, and removes the markings. Through these procedures, candidate regions within the image region of the cell aggregate are identified. The lower part of Figure 53 shows an upright observation image of the cell aggregate to be evaluated, in which the candidate regions are marked slightly brighter. It is preferable to create an image in which the candidate regions are overlaid on the original inverted observation image, and store the image data in a data storage area in the memory 120 as an output image for visual confirmation.

[0189] [3] Normal Region Determination Step Next, the normal region determination means 114 determines, as a normal retinal tissue region, a candidate region to be identified that is determined to satisfy the normal region determination criteria for the upright observation image based on the normal region determination criteria for the brightness and / or morphology of the upright observation image (step S103B). A cell aggregate having such a normal retinal tissue region is determined to have no quality problems. This normal region determination step is executed in detail by steps [3A] and [3B], which will be described below. This will be described with reference to the detailed operational flow for normal region determination shown in FIG.

[0190] [3A] Region Part Evaluation Stage The region part evaluation means 114A calculates an evaluation score for each of the region parts present in the range obtained by dividing the candidate region into multiple parts along a curve representing the extension direction of the candidate region, based on evaluation criteria related to color and / or shape corresponding to the upright observation technique, for the candidate region identified in the image region of the cell aggregate in the upright observation image (step S151B). In this region part evaluation stage, in order to further narrow down the candidate region to regions that are more likely to contain normal retinal tissue, an evaluation score for each of the region parts present in the range obtained by dividing the candidate region into multiple parts is calculated. Specifically, this stage is performed by executing the following steps (4) and (5).

[0191] (4) The region portion evaluation means 114A expresses a specific position along the extension direction of the candidate region as a counterclockwise angle, with a straight line extending horizontally to the right from the center of the cell aggregate, i.e., the center of a reference ellipse that approximates its shape, as the reference angle of 0 degrees, and increments the angle α by a predetermined angle, such as 1 degree, to determine the width of the candidate region at the specific position from 0 degrees to 360 degrees counterclockwise.

[0192] (5) The region evaluation unit 114A evaluates a candidate region by assigning a Level 1 evaluation score to regions where regions with a width of 75 pixels or more are contiguous at one degree, and a Level 2 evaluation score to regions where regions with a width of 75 pixels or more are contiguous at an angle of 2 degrees or more. This is because a candidate region must be relatively wide to be included in the normal retinal tissue region. Actual upright observation confirmed that the width was approximately 0.15 mm, which corresponds to 75 pixels in an upright observation image. Therefore, a width of 75 pixels or more, which is the standard width corresponding to the upright observation technique, was set as a necessary condition, and such regions were assigned a Level 1 evaluation score. Note that the 75-pixel criterion is higher than the 15-pixel criterion for inverted observation images. However, this difference is due to factors such as the low contrast in upright observation images, which causes bright ONbL regions to appear wider. Furthermore, the longer the length of the region portion satisfying such width conditions, the higher the likelihood that the region portion of the candidate region is included in the normal retinal tissue region. Therefore, a higher evaluation score is given when a region portion satisfying the width condition of 75 pixels or more exists consecutively for two or more degrees, resulting in a Level 2 evaluation score. The reference width corresponding to the observation method (upright observation) in Example 2 was approximately 75 pixels. The region portion evaluation means 114A then measured the width of the candidate region along its periphery to determine the evaluation score. The region indicated by B in Figure 57 shows the obtained evaluation score as a line graph. In the graph of the region indicated by B in Figure 57, the horizontal axis represents the angle of the measured region portion as seen from the center of the cell aggregate, calculated by incrementing the angle by one degree counterclockwise over a 360-degree range, with the line extending horizontally to the right from the center of the cell aggregate as the reference angle of 0 degrees. The vertical axis of the region indicated by B in Figure 57 represents the region portion width and evaluation score, with the region portion width scale indicated on the left end and the evaluation score scale indicated on the right end. The width of the region is the width of the area where the line passing through the center of the cell aggregate overlaps with the region, and is shown by a dashed line in the graph. The evaluation score is shown by a solid line in the graph, taking the value of 0, 1, or 2.The graph shown here can also be displayed on the display of the operation terminal 220 so that the operator can check it.

[0193] [3B] Continuous Length-Based Normal Region Determination Step If the portions of a candidate region whose evaluation scores in the upright observation image are equal to or greater than the normal region determination criterion value extend continuously along a curve representing the direction of extension for a length equal to or greater than a predetermined normal region determination criterion, the continuous length-based normal region determination means 114B determines the candidate region as a normal retinal tissue region (step S152B). That is, at this step, the candidate regions determined to have evaluation scores equal to or greater than the normal region determination criterion value are further narrowed down to regions that are more likely to contain normal retinal tissue, and a determination is made as to whether such candidate regions extend for a length equal to or greater than the predetermined normal region determination criterion. Specifically, this step is performed by executing the following procedure (6).

[0194] (6) In an upright observation image, when a region within a cell aggregate with a width of 75 pixels or more exists continuously in a cell aggregate with a diameter of approximately 1 mm or more, the continuous length of the region is an arc length corresponding to an angle of 45 degrees or more from the center of a reference ellipse approximating the cell aggregate (when the shape of the region is approximated by an arc, the central angle is 45 degrees or more), the candidate region can be determined to be a sufficiently large normal retinal tissue region with an actual length of approximately 0.4 mm or more. Therefore, this is set as the normal region determination criterion length, and a cell aggregate having a candidate region with a length equal to or greater than this normal region determination criterion length is determined to contain the neural retina (normal retinal tissue region). According to this criterion, the continuous length-based normal region determination means 114B determines whether a region portion of a candidate region with an evaluation score of 2 or more continues along a curve representing the direction of extension of the region and has a length corresponding to an angle of 45 degrees, which is the predetermined normal region determination criterion length, and determines a candidate region that meets this criterion as a normal retinal tissue region. The solid line indicates the angular range in which candidate regions with a width of 75 pixels or more and a continuous length corresponding to an arc length of 45 degrees or more are present as the determination result for the region shown in FIG. 57A. The dashed line indicates the corresponding position on the graph of the evaluation score for the region shown in FIG. 57B with the solid line indicating the candidate region determination result for the region shown in FIG. 57A. It can be seen that regions determined to be normal retinal tissue regions meeting the criteria are present at three positions corresponding to angles of approximately 2 to 102 degrees, approximately 121 to 182 degrees, and approximately 189 to 356 degrees. Cell aggregates containing candidate regions determined to be normal retinal tissue regions are determined to be normal retinal tissue regions with no quality issues. This determination result, along with information identifying the cell aggregate being determined, is stored in a data storage area in memory 120 for later reference. At this time, the angle at which the region determined to meet the criteria exists can also be stored as information identifying the position of the normal retinal tissue region.

[0195] The results of applying the above-mentioned steps to 12 cell aggregates from sample SPL20 to sample SPL31 and assessing them are shown in Table 2. In the assessment results, 12 of the 12 samples were assessed as having one or more normal retinal tissue regions and therefore were judged to have "good" quality, and none of the samples were assessed as having no normal retinal tissue regions and therefore were judged to have "bad" quality.

[0196]

[0197] It was confirmed that the trend in the number of samples judged to have a quality of "good" in Table 2 correlates with the judgment results of the inverted observation images of Example 1 shown in Table 1. That is, all of the samples judged to have a quality of "good" in the judgment results of the inverted observation images of Example 1 shown in Table 1 were also judged to have a quality of "good" in Example 2 shown in Table 2. In this way, it was confirmed that by applying machine judgment, the suitability of cell aggregates can be judged from upright observation images. However, because inverted observation images are not taken into consideration in Example 2, there is a possibility that the condition of the surface behind the surface of the cell aggregate observed by upright observation will not be reflected in the judgment results.

[0198] Example 3: Determination by Combining Individual Determination Results of Upright Observation Image and Inverted Observation Image of Cell Aggregate It is believed that a more accurate determination can be made by combining the determination result from the inverted observation image in Example 1 and the determination result from the upright observation image in Example 2. In Example 3, if the cell aggregate to be determined has a normal retinal tissue region in both the determination result from the inverted observation image in Example 1 and the determination result from the upright observation image in Example 2, the cell aggregate to be determined is finally determined to have a normal retinal tissue region. The overall operational flow of Example 3 is shown in Figure 48. Using samples SPL20 to SPL31 as the cell aggregates to be determined, the following steps were each performed.

[0199] Steps S101A to S103A of Example 1 were performed, and it was determined whether the cell aggregate sample had a normal retinal tissue region based on the inverted observation image (step S111). The area indicated by D in Figure 58 shows a line graph of the evaluation score obtained in step S103A, and the area indicated by B in Figure 58 shows the candidate region determination results obtained in step S103A. The solid line of the candidate region determination results for the area indicated by B in Figure 58 and the corresponding position on the graph of the evaluation score for the area indicated by D in Figure 58 are indicated by dashed lines.

[0200] Steps S101B to S103B of Example 2 were performed, and it was determined whether the cell aggregate sample had a normal retinal tissue region based on the upright observation image (step S112). The area indicated by C in Figure 58 shows a line graph of the evaluation score obtained in step S103B, and the area indicated by A in Figure 58 shows the candidate region determination results obtained in step S103B. The solid line of the candidate region determination results for the area indicated by A in Figure 58 corresponds to the position on the graph of the evaluation score for the area indicated by C in Figure 58, indicated by a dashed line.

[0201] If the cell aggregate to be evaluated is determined to have a normal retinal tissue region in both the evaluation results based on the inverted observation image obtained in steps S101A to S103A of Example 1 and the evaluation results based on the upright observation image obtained in steps S101B to S103B of Example 2, the cell aggregate to be evaluated is finally determined to have a normal retinal tissue region (step S113). Cell aggregates that have regions of these candidate regions evaluated as normal retinal tissue regions are determined to have no quality problems. This evaluation result, along with information identifying the cell aggregate to be evaluated, is stored in a data storage area in memory 120 for later reference. At this time, the angle at which the region determined to meet the criteria exists can also be stored as information identifying the position of the normal retinal tissue region.

[0202] The results of applying the above-mentioned steps to twelve cell aggregates from sample SPL20 to sample SPL31 and assessing them are shown in Table 3. In the assessment results, eight of the twelve samples were judged to have one or more normal retinal tissue regions and therefore had a quality of "good," while four samples were judged to have no normal retinal tissue regions and therefore had a quality of "bad."

[0203]

[0204] The judgment results for the combination of the upright observation image and the inverted observation image in Example 3 shown in Table 3 were the same as the judgment results for the inverted observation image in Example 1. However, in Example 3, unless the judgment results for both the upright observation image and the inverted observation image are "good," the quality is not ultimately judged to be "good." Therefore, it is thought that a stricter judgment can be made than when either observation image is judged individually. In other words, Example 3 makes it possible to more strictly judge whether the quality is "bad." In this way, by combining the individual judgment results from the upright observation image and the inverted observation image, it is possible to judge with higher accuracy whether a cell aggregate has a normal retinal tissue region than when either observation image is judged individually.

[0205] Example 4: Simultaneous Assessment of Cell Aggregates in Upright and Inverted Observation Images In Example 3, the assessment was performed by combining the assessment results obtained by separately analyzing the upright and inverted observation images. However, it is believed that even more accurate assessment can be achieved by simultaneously assessing both the upright and inverted observation images as much as possible. In Example 4, the analysis of the inverted observation image according to Example 1 and the analysis of the upright observation image according to Example 2 are performed separately up to the stage of calculating the evaluation score of the candidate region, and then the assessment of whether the cell aggregate to be assessed has a normal retinal tissue region is performed by simultaneously considering the evaluation scores of the candidate region portions in both the inverted and upright observation images. The overall operational flow of Example 3 is shown in FIG. 49 , and the detailed operational flow for normal region assessment is shown in FIG. 50 . Using samples SPL20 to SPL31 as the cell aggregates to be assessed, the following steps were performed.

[0206] Steps S101A to S102A of Example 1 are executed to obtain an image in which candidate regions are marked in an inverted observation image of the cell aggregate to be evaluated, and step S151A1 of Example 1 is executed to calculate an evaluation score for the region portion of the candidate region based on the evaluation criteria corresponding to the inverted observation technique (step S121). The upper part of Fig. 54 shows the image in which the candidate regions are marked in the inverted observation image obtained in step S102A.

[0207] Steps S101B to S102B of Example 2 are executed to obtain an image in which candidate regions are marked in the upright observation image of the cell aggregate to be evaluated, and step S151B1 of Example 2 is executed to calculate an evaluation score for the region portion of the candidate region based on the evaluation criteria corresponding to the upright observation technique (step S122). The lower part of Figure 54 shows the image in which the candidate regions are marked in the upright observation image obtained in step S102B.

[0208] The continuous length-based normal region determining means 114B determines a candidate region as a normal retinal tissue region when the evaluation scores of the candidate region in both the inverted observation image and the upright observation image are simultaneously equal to or greater than the normal region determination criterion value and the region portions of the candidate region extend continuously along a curve representing the direction of the region portions for a predetermined normal region determination criterion length or longer (step S161). This step is specifically performed by executing the following procedure (1).

[0209] (1) The contiguous length-based normal region determination means 114B determines whether both a region portion of a candidate region with an evaluation score of 2 in an inverted observation image (i.e., a region portion with a width of 15 pixels or more and a continuous arc length corresponding to a central angle of 2 degrees or more) and a region portion of a candidate region with an evaluation score of 2 in an upright observation image (i.e., a region portion with a width of 75 pixels or more and a continuous arc length corresponding to a central angle of 2 degrees or more) have an arc length corresponding to an angle of 45 degrees, which is a predetermined normal region determination criterion length, along a curve representing the extension direction, and determines a candidate region that satisfies this criterion as a normal retinal tissue region. Cell aggregates having regions of these candidate regions determined to be normal retinal tissue regions are determined to be of satisfactory quality.

[0210] In the graph of the region indicated by C in Figure 59, the horizontal axis represents the angle of the region of the cell aggregate in the measured inverted observation image, measured in increments of 1 degree over a 360-degree counterclockwise range, with a line extending horizontally to the right from the center of the cell aggregate (the center of a reference ellipse that approximates its shape) as the reference angle of 0 degree. The vertical axis of the region indicated by C in Figure 59 represents the width of the region and the evaluation score, with the scale for the width of the region indicated on the left and the scale for the evaluation score indicated on the right. The width of the region is the width of the portion where the line passing through the center of the cell aggregate overlaps with the region, and is indicated by a dashed line in the graph. The evaluation score based on the continuous length expressed in angle for region portions with a width of 15 pixels or more is indicated by a solid line (an angle of 1 degree is an evaluation score of 1, and an angle of 2 degrees or more is an evaluation score of 2).

[0211] In the graph of the region indicated by B in Figure 59, the horizontal axis represents the angle of the region of the cell aggregate in the measured upright observation image, incremented by 1 degree over a 360-degree counterclockwise range, with a line extending horizontally to the right from the center of the cell aggregate (the center of a reference ellipse that approximates its shape) as the reference angle of 0 degree. The vertical axis of the region indicated by B in Figure 59 represents the width of the region and the evaluation score, with the scale for the width of the region indicated on the left edge and the scale for the evaluation score indicated on the right edge. The width of the region is the width of the portion where the line passing through the center of the cell aggregate overlaps with the region, and is indicated by a dashed line in the graph. The evaluation score based on the continuous length expressed in angle for region portions with a width of 75 pixels or more is indicated by a solid line (an angle of 1 degree is an evaluation score of 1, and an angle of 2 degrees or more is an evaluation score of 2).

[0212] The solid lines indicate the range of angles within the region indicated by A in Figure 59, where the candidate region portion has a width of 15 pixels or more in the inverted image and a continuous length whose arc length corresponds to a central angle of 45 degrees or more, and where the candidate region portion has a width of 75 pixels or more in the upright image and a continuous length whose arc length corresponds to a central angle of 45 degrees or more. The solid lines indicate the candidate region determination results for the region indicated by A in Figure 59, and the corresponding positions on the graphs of the evaluation scores for the regions indicated by B and C in Figure 59 are indicated by dashed lines. It can be seen that regions determined to be normal retinal tissue regions that meet the criteria exist at five positions corresponding to angles of approximately 28 degrees to 84 degrees, approximately 121 degrees to 167 degrees, approximately 189 degrees to 242 degrees, approximately 246 degrees to 296 degrees, and approximately 302 degrees to 351 degrees. Therefore, cell aggregates having regions determined to be normal retinal tissue regions are determined to be normal retinal tissue regions with no quality issues. The determination result, together with information identifying the cell aggregate to be determined, is stored in a data storage area in memory 120 so that it can be referenced later. At this time, the angle at which the area determined to satisfy the criteria exists can also be stored as information identifying the position of the normal retinal tissue area.

[0213] The results of applying the above-mentioned steps to the 12 cell aggregates from sample SPL20 to sample SPL31 and assessing them are shown in Table 4. In the assessment results, 7 of the 12 samples were judged to have "good" quality since they had one or more normal retinal tissue regions, and 5 samples were judged to have "bad" quality since they did not have normal retinal tissue regions.

[0214]

[0215] In the results of simultaneous evaluation of the upright and inverted observation images in Example 4 shown in Table 4, five images were judged to be of "unacceptable" quality. This is a stricter result than the results of the evaluation of the inverted observation images in Example 1 and the evaluation of the combination of the upright and inverted observation images in Example 3, in which four images were judged to be of "unacceptable" quality. In the simultaneous evaluation of the upright and inverted observation images, the number of normal retinal tissue regions in sample SPL21 was five, which is more than the three in the upright observation image. This is because several smaller regions within the region judged as normal retinal tissue regions in the upright observation image were judged as normal retinal tissue regions in the inverted observation image and the simultaneous evaluation of the upright and inverted observation images. This results in a more precise result in that non-retinal tissue included in the region judged as normal retinal tissue regions in the upright observation image was excluded. Although the angle at which an evaluation score of 2 is obtained may differ between the upright observation image and the inverted observation image, it was confirmed that stricter evaluation results could be obtained without any problems by simultaneously using these evaluation scores. Furthermore, by calculating the correlation coefficient and the p-value of the significance test result for the width of the candidate region determined to be neural retina between the upright observation image and the inverted observation image of a cell aggregate as reference values ​​for quality measurement and evaluating the difference between the upright observation image and the inverted observation image, it was confirmed that there is a correlation between them. Therefore, appropriate determination results should be obtained by simultaneously using the evaluation scores for the same angle in the upright observation image and the inverted observation image. Furthermore, by determining that areas at angles where the evaluation scores differ do not meet the criteria, it was confirmed that it is possible to determine whether a cell aggregate contains a normal retinal tissue region with higher accuracy than in any of Examples 1 to 3.

[0216] <Preparation of neural retinal sheet for transplantation> When a cell aggregate determined to have a normal retinal tissue region in any of Examples 1 to 4 consists of multiple lobes, a neural retinal sheet for transplantation can be prepared by cutting out the lobe containing the normal retinal tissue region. Cell aggregates determined to have a normal retinal tissue region in Example 4 in particular have undergone the most sophisticated analysis and are therefore suitable for preparing neural retinal sheets for transplantation. When cutting out the lobe containing the normal retinal tissue region, an appropriate lobe can be selected by referring to the data stored in the data storage area of ​​memory 120, which indicates the angle at which the region determined to meet the criteria exists. When the cell aggregate consists of a single lobe, it can be used as a neural retinal sheet for transplantation.

[0217] <Reference Parameter Values> In the above example, the condition for a normal retinal tissue region was that an ONbL with a width of 0.03 mm or more was present continuously at an angle of 45 degrees or more from the center of the cell aggregate in an inverted observation image of the cell aggregate. In this case, for a cell aggregate with a diameter of approximately 1 mm, the length of the ONbL along the periphery was approximately 0.4 mm or more. However, the quality criteria for visual confirmation can be slightly changed and the parameter values ​​can be changed accordingly. This allows the parameter value conditions to be within a certain range. For example, with regard to the width reference value, if the quality criteria were more lenient, the width of the region that must be present continuously at an angle of 45 degrees or more could be set to 0.02 mm, 0.025 mm, etc. Furthermore, if the quality criteria were more strict, the width of the region that must be present continuously at an angle of 45 degrees or more could be set to 0.06 mm, 0.07 mm, etc. Furthermore, with regard to the angle reference value, if the quality standard is more lenient, the central angle of the range in which ONbLs with a width equal to or greater than the predetermined reference value should exist can be set to 30 degrees, 35 degrees, 40 degrees, etc. Furthermore, if the quality standard is more strict, the central angle of the range in which ONbLs with a width equal to or greater than the predetermined reference value should exist can be set to 50 degrees, 60 degrees, 75 degrees, etc. Furthermore, as the diameter of the cell aggregate increases, if the continuous angle at which ONbLs exist as viewed from the center is the same, the length along the periphery of the ONbLs also increases proportionally. Therefore, as the diameter of the cell aggregate increases, the reference angle at which ONbLs exist as viewed from the center can be reduced. On the other hand, by fixing the reference angle at which ONbLs exist as viewed from the center to 45 degrees and setting the diameter of the cell aggregate to be evaluated to approximately 1 mm or greater, the presence of ONbLs with a length along the periphery of approximately 0.4 mm or greater can be determined regardless of the size of the cell aggregate.

[0218] The center point serving as the reference for the central angle can also be changed as appropriate. That is, instead of the center point serving as the reference for the central angle being the angle from the center point of the smallest ellipse tangent to the contour of the image of the cell aggregate portion, for example, an ellipse approximating each lobe, such as an ellipse including the contour, can be set, and the angle from the center point of that ellipse can be used. In this case, it is preferable to modify the parameter value for the central angle condition based on the size (e.g., the length of the major axis) of the ellipse approximating each lobe. That is, for example, as the size of the ellipse approximating each lobe becomes smaller, it is preferable to increase the central angle condition in inverse proportion to the size.

[0219] As described above, the basic technical idea of ​​the present invention is to determine the appropriate width and length of ONbL based on some criteria so that an area where ONbL and INbL are adjacent in a cell aggregate can be determined to be a normal retinal tissue area.

[0220] <Other Evaluation Parameters> In addition, evaluation parameters from the following perspectives can also be used in the evaluation of retinal tissue. Specifically, evaluation parameters for a normal retinal tissue region can include the continuity of the epithelial structure, the normality of the shape of the cells in the epithelial structure, the normality of the color tone of the epithelial structure, the normality of the texture of the cells inside the epithelial structure, the appropriateness of the width of the epithelial structure, and the smoothness of the periphery of the epithelial structure. The morphology and color evaluation parameters for a normal retinal tissue region can be determined by calculating the similarity between an image of the evaluation region of the cell aggregate to be evaluated and an image of a region of the cell aggregate having a typical normal morphology using a known method. Furthermore, evaluation parameters for non-target cells can include the color tone of the epithelial structure, the texture inside the epithelial structure, the thickness of the epithelial structure, the smoothness of the periphery of the epithelial structure, the presence or absence of pigmented cells, and the presence or absence of attachments. The evaluation parameters for non-target cells can be determined by calculating the similarity between an image of the evaluation region of the cell aggregate to be evaluated and an image of a region of a typical specific non-target cell in the cell aggregate using a known method.

[0221] As a specific example of the evaluation of the evaluation parameters, the following criteria can be used: That is, the evaluation parameters related to retinal tissue and non-target cells each have criteria for evaluating the score, and each of the criteria can be as follows:

[0222] Regarding the continuity of the epithelial structure, the higher the degree of continuity of the epithelial structure, the higher the evaluation score can be. Regarding the normality of the shape of cells in the epithelial structure, the fewer cells with abnormal shapes, the higher the evaluation score can be. Regarding the normality of the color tone of the epithelial structure, the closer it is to a standard state in which the color tone gradually changes from bright white to light brown to black from the outside to the inside in the normal direction of the epithelial structure, the higher the evaluation score can be. Regarding the normality of the texture of cells inside ... more uniformly rounded, similarly sized cells are arranged inside the epithelial structure and the more granular the texture, the higher the evaluation score can be. Regarding the appropriateness of the width of the epithelial structure, the wider the width of the outer bright ring-shaped region as viewed in the normal direction of the epithelial structure, the higher the evaluation score can be. Regarding the smoothness of the outer periphery of the epithelial structure, the fewer inflection points where the angle changes abruptly in the outline of the cell aggregate and the higher the continuity of the angle change, the higher the evaluation score can be. Regarding the presence or absence of pigmented cells, the fewer pigmented cells there are, the higher the score can be given. Regarding the presence or absence of deposits, the fewer dead cells or non-cellular components derived from equipment, etc., the higher the score can be given.

[0223] <Method for Producing Retinal Tissue Transplantable into Animals> Retinal tissue transplantable into animals can be produced using a cell aggregate determined to have a normal retinal tissue region according to any of Examples 1 to 4. Below, a method for producing retinal tissue transplantable into animals using a cell aggregate determined to have a normal retinal tissue region in Example 4, which has particularly high determination accuracy, is described. First, a cell aggregate determined to have a normal retinal tissue region in Example 4 is prepared. From this, a region portion of the candidate region that meets the normal region determination criteria is excised as a normal retinal tissue region, making it possible to produce retinal tissue transplantable into animals.

[0224] Furthermore, by confirming the expression of neural retinal cell-related genes and non-neural retinal cell-related genes, it is possible to determine with greater accuracy the normal retinal tissue area that can be transplanted into an animal. To this end, for a normal retinal tissue area that is determined to simultaneously satisfy each of the normal area determination criteria by any of the methods of Examples 1 to 4, it is preferable to further perform the following steps: detect the expression of neural retinal cell-related genes and non-neural retinal cell-related genes in the normal retinal tissue area; and determine that the normal retinal tissue area can be transplanted into an animal if expression of neural retinal cell-related genes is observed but expression of non-neural retinal cell-related genes is not observed. Here, it is preferable that the non-neural retinal cell-related genes include one or more genes selected from the group consisting of spinal cord tissue marker genes and eye-related tissue marker genes. The presence of neural retinal cells can be confirmed by the presence or absence of expression of neural retinal cell-related genes (sometimes referred to as "neural retinal cell markers" or "neural retinal markers"). This confirmation allows for more precise determination of the normal retinal tissue area. The presence or absence of expression of a neural retinal cell marker, or the proportion of neural retinal cell marker-positive cells in a cell population or tissue, can be confirmed using known techniques such as techniques using antibodies, techniques using nucleic acid primers, and techniques using sequencing reactions.

[0225] Furthermore, by confirming the expression of neural retinal cell-related genes and non-neural retinal cell-related genes, it is possible to prepare retinal tissue that can be transplanted into animals and that contains a more favorable normal retinal tissue region. To this end, first, a cell aggregate determined to have a normal retinal tissue region in any of Examples 1 to 4 is prepared. The following steps [1] to [4] are performed on a region within the cell aggregate that is a candidate region that meets the normal region determination criteria.

[0226] [1] A portion or all of the region of the candidate region that meets the normal region criteria is extracted as a quality assessment sample. [2] Expression of neural retinal cell-related genes and non-neural retinal cell-related genes in the quality assessment sample is detected. [3] If expression of neural retinal cell-related genes is confirmed but expression of non-neural retinal cell-related genes is not confirmed, (1) neural retina (neural retina for transplantation) in the same cell aggregate as the cell aggregate containing the quality assessment sample, (2) neural retina (neural retina for transplantation) in the same cell aggregate lot as the cell aggregate containing the quality assessment sample, or (3) neural retina (neural retina for transplantation) in the same cell aggregate lot as the cell aggregate containing the quality assessment sample, which is the entirety of the quality assessment sample, is determined to be usable as a neural retina for transplantation. Here, the non-neural retinal cell-related genes preferably include one or more genes selected from the group consisting of spinal cord tissue marker genes and eyeball-related tissue marker genes. [4] The normal retinal tissue region determined to be usable as a neural retina for transplantation is excised. This makes it possible to prepare neural retinas for transplantation from cell aggregates in which the normal retinal tissue region has been determined more precisely. As described above, neural retinas for transplantation prepared by cutting out the cell aggregates using tweezers, scissors, a knife, or the like can be used as pharmaceutical compositions containing a cell population for transplantation in the form of a cell sheet as an active ingredient, or as therapeutic agents containing a cell population for transplantation in the form of a cell sheet, for treating diseases caused by disorders of retinal cells or damaged states of retinal cells. When cell aggregates of tissues other than the retina are used for the cell sheet, the cell sheet can be used as a therapeutic agent for that tissue.

[0227] Neuroretinal cell-related genes (target cell-related genes) refer to genes expressed by neuroretinal cells. Neuroretinal cell-related genes are preferably genes that are expressed at higher levels in photoreceptors (rod photoreceptors, cone photoreceptors), horizontal cells, amacrine cells, interneurons, retinal ganglion cells (ganglion cells), bipolar cells (cone bipolar cells, rod bipolar cells), Müller glial cells, or precursor cells of these cells, neuroretinal progenitor cells, etc., compared to non-target cells. Examples of neuroretinal cell-related genes include the above-mentioned neuroretinal cell markers, with RAX, Chx10, SIX3, SIX6, RCVRN, CRX, NRL, and NES being preferred.

[0228] In one embodiment, the cerebrospinal tissue marker gene may be one or more genes selected from the group consisting of a telencephalon marker gene, a diencephalon / mesencephalon marker gene, and a spinal cord marker gene. The diencephalon / mesencephalon marker gene may be one or more genes selected from the group consisting of a diencephalon marker gene, a mesencephalon marker gene, and a hypothalamic marker gene related to the hypothalamus, which is part of the diencephalon. In one embodiment, the eye-related tissue marker gene may be one or more genes selected from the group consisting of an optic stalk marker gene, a ciliary body marker gene, a lens marker gene, and a retinal pigment epithelium marker gene.

[0229] Telencephalic marker genes refer to genes expressed in the telencephalon. Telencephalic marker genes may include one or more genes selected from the group consisting of FoxG1 (also known as Bf1), Emx2, Dlx2, DIx1, and Dlx5. Other telencephalic marker genes include Emx1, LHX2, LHX6, LHX7, Gsh2, etc. Diencephalic / mesencephalic marker genes refer to genes expressed in the diencephalon and / or midbrain. Diencephalic / mesencephalic marker genes may include one or more genes selected from the group consisting of OTX1, OTX2, and DMBX1. Hypothalamic marker genes refer to genes expressed in the hypothalamus. Hypothalamic marker genes may include one or more genes selected from the group consisting of Rx, Nkx2.1, Dmbx1, OTP, gad1, FGFR2, and EFNA5. Spinal cord marker genes refer to genes expressed in the spinal cord. The spinal cord marker gene may comprise one or more genes selected from the group consisting of HoxB2, HoxA5, HOXC5, HOXD1, HOXD3 and HOXD4.

[0230] The term "optic stalk marker gene" refers to a gene expressed in optic stalk. The term "optic stalk marker gene" may include one or more genes selected from the group consisting of GREM1, GPR17, ACVR1C, CDH6, Pax2, Pax8, GAD2, and SEMA5A. The term "lens marker gene" refers to a gene expressed in the lens. The term "lens marker gene" may include one or more genes selected from the group consisting of CRYAA and CRYBA1. The term "ciliary body marker gene" refers to a gene expressed in the ciliary body, ciliary margin, and / or ciliary body. The term "ciliary body marker gene" may include one or more genes selected from the group consisting of ZC1, MAL, HNF1beta, FoxQ1, CLDN2, CLDN1, GPR177, AQP1, and AQP4. The retinal pigment epithelium marker gene refers to a gene expressed in retinal pigment epithelium cells. Examples of the retinal pigment epithelium marker gene include the retinal pigment epithelium markers described above, and may include one or more genes selected from the group consisting of MITF, TTR, and BEST1.

[0231] <Determination by Artificial Intelligence Using Deep Learning> In Examples 1 to 4, machine determination was performed based on criteria based on specific morphological and color characteristics, such as the width and length of the bright regions of the cell aggregate to be determined. This method is highly effective when the criteria can be clearly defined based on a predetermined perspective. Furthermore, by constructing a machine learning model using deep learning that inputs an inverted or upright observation image of the cell aggregate and outputs the normal / abnormal state for each cell aggregate, it is also possible to perform machine determination using artificial intelligence using deep learning to determine whether a cell aggregate contains normal retinal tissue regions. Machine determination using deep learning can take into account properties that cannot be formulated as clearly defined criteria based on specific morphology and / or color based on the results of observation with the naked eye. Below, machine determination using artificial intelligence using deep learning is described.

[0232] First, a method for generating a machine learning model will be described. A method for generating a machine learning model to be applied to a discrimination device that discriminates, among cell aggregates that may contain pluripotent stem cell-derived retinal tissue, cell aggregates that have a normal retinal tissue region that can be transplanted into an animal can include: a training data acquisition step of acquiring, for each of a plurality of cell aggregates, a plurality of training data including image data that is acquired by photographing the cell aggregate using an inverted microscope and an upright microscope and that is tagged with a training signal (label) indicating whether an image region of the cell aggregate, included in image data acquired, has a normal retinal tissue region; and a model training step of using the training data to train a machine learning model that receives image data of a cell aggregate to be determined as input and outputs information indicating whether the cell aggregate to be determined has a normal retinal tissue region.

[0233] It is preferable to perform appropriate preprocessing on the image data included in the training data, such as contrast enhancement, contour enhancement, removal of unnecessary image areas, and standardization of the position and size of image areas of cell aggregates.

[0234] Here, both the inverted and upright observation images of the cell aggregate can be used as inputs to the machine learning model at the same time. That is, the image data of the inverted observation image and the upright observation image can be used as training data, along with a training signal indicating whether the image region of the cell aggregate contains a normal retinal tissue region. This makes it possible to generate a machine learning model that makes judgments taking into account both the inverted and upright observation images.

[0235] A method for generating a machine learning model to be applied to a discrimination device that discriminates between cell aggregates that may contain pluripotent stem cell-derived retinal tissue and cell aggregates that have a normal retinal tissue region that can be transplanted into an animal can include: a training data acquisition step of acquiring, for each of a plurality of cell aggregates, sets of image data of inverted and upright observation images of the same cell aggregate, each of which is obtained by photographing the cell aggregate using an inverted microscope and an upright microscope, and which is associated with the image data of the inverted and upright observation images so that positions in each of the inverted and upright observation images correspond to the same positions of the cell aggregate when viewed from a vertical direction; the training data acquisition step of acquiring, for each of the plurality of cell aggregates, sets of image data of inverted and upright observation images of the same cell aggregate, each of which is associated with a training signal that indicates whether an image region of the cell aggregate having a contour of a region of a predetermined length or more is a normal retinal tissue region; and a model training step of using the plurality of training data to train a machine learning model that inputs image data of the inverted and upright observation images of the cell aggregate to be determined and outputs information indicating whether the cell aggregate to be determined has a normal retinal tissue region.

[0236] Furthermore, a discrimination device can be configured that individually judges inverted observation images and upright observation images and then combines the judgment results. A discrimination device for a cell aggregate having a transplantable normal retinal tissue region for this purpose can include: a machine learning model generated by the above-described machine learning model generation method; an image data input unit that inputs image data of the inverted observation images and the upright observation images of the same cell aggregate to be judged to the machine learning model; output means that receives output from the machine learning model information indicating whether the cell aggregate to be judged in each of the inverted observation images and the upright observation images has a normal retinal tissue region; and discrimination means that discriminates that the cell aggregate to be judged has a normal retinal tissue region when both the inverted observation image and the upright observation image of the cell aggregate to be judged indicate that the cell aggregate to be judged has a normal retinal tissue region.

[0237] Furthermore, a method for discriminating between transplantable normal retinal tissue regions using this discrimination device can be implemented, which includes an acquisition step of acquiring image data of an inverted observation image and an upright observation image of the same cell aggregate to be determined; an image data input step of inputting the acquired image data of the inverted observation image and the upright observation image of the same cell aggregate to be determined into the discrimination device; and a discrimination result receiving step of receiving a discrimination result from the discrimination device that the cell aggregate to be determined has a normal retinal tissue region.

[0238] Furthermore, a discrimination device can be configured that simultaneously discriminates between inverted observation images and upright observation images. The discrimination device for a cell aggregate having a transplantable normal retinal tissue region for this purpose can include: a machine learning model generated by the above-described machine learning model generation method; an image data input unit that inputs a set of image data of an inverted observation image and an upright observation image of the same cell aggregate to be discriminated into the machine learning model; and an output means that receives, from the machine learning model, information indicating whether the cell aggregate to be discriminated has a normal retinal tissue region.

[0239] Furthermore, a method for discriminating between transplantable normal retinal tissue regions using this discrimination device can be implemented, which includes an acquisition step of acquiring a set of image data of an inverted observation image and an upright observation image of the same cell aggregate to be determined; an image data input step of inputting the acquired set of image data of the inverted observation image and the upright observation image of the same cell aggregate to be determined into the discrimination device; and a discrimination result receiving step of receiving from the discrimination device a discrimination result that the cell aggregate to be determined has a normal retinal tissue region.

[0240] Furthermore, a method for producing retinal tissue for transplantation can be carried out, which includes a step of cutting out a normal retinal tissue region from a cell aggregate containing a normal retinal tissue region identified based on any of the above-mentioned methods for identifying a transplantable normal retinal tissue region. While the above description has been given using retinal tissue as an example of a cell aggregate, the above-mentioned technique can also be used to observe human or animal tissues other than retinal tissue. In other words, by observing both inverted and upright images of a cell aggregate containing tissue other than retinal tissue derived from pluripotent stem cells, the present invention makes it possible to compensate for the disadvantages of each observation method and perform synergistic, more effective observations. Furthermore, the present invention can be configured to identify, in an image region of a cell aggregate other than retinal tissue, a region that is determined to satisfy the corresponding candidate region criteria, such as those related to color, in the inverted observation image and the upright observation image, as a candidate region included in a region that is a candidate for normal tissue, and to determine, in the candidate region of the inverted observation image and the upright observation image, a cell aggregate having a candidate region that is determined to satisfy the corresponding normal region criteria, such as those related to brightness and / or morphology, in the inverted observation image and the upright observation image, as a region that contains normal tissue. With this configuration, it is also possible to automatically extract, by information processing, cell aggregates other than retinal tissue that contain three-dimensionally good normal regions using image data observed from two directions, namely, above and below, using an upright microscope and an inverted microscope.

[0241] The present invention can be widely used to produce cell aggregates containing various tissues, particularly neural retinal tissues, in the field of transplantation therapy for humans and animals using regenerative medicine.

[0242] 10 Cell aggregate 21 Upright microscope 22 Inverted microscope 23 Container 24 Culture solution 31 Upright observation 32 Inverted observation 100 Normal cell aggregate determination system 110 Normal cell aggregate determination server 111 Control unit 112 Image data acquisition means 113 Candidate region identification means 114 Normal region determination means 114A Region portion evaluation means 114B Continuous length-based normal region determination means 120 Memory 120A Normal cell aggregate determination program 120B Normal cell aggregate determination data 130 Communication interface 200 Upright microscope 210 Inverted microscope 215 Digital camera 220 Operation terminal CL1 Curve CR1 Candidate region CTR Center point ET1 to ET3 Epithelial tissue LAN Local area network PLT1 to PLT6 Container R1 to R4: Straight line RE: Reference ellipse SPL1 to SPL31: Sample W: Width

Claims

1. A method for determining a normal retinal tissue region in a cell aggregate that may contain retinal tissue derived from pluripotent stem cells and that is transplantable into an animal, comprising: an image data acquisition step of acquiring image data for an inverted observation image and an upright observation image of the cell aggregate by photographing the cell aggregate using an inverted microscope and an upright microscope; a candidate region identification step of identifying, in the image region of the cell aggregate in each of the inverted observation image and the upright observation image, regions that are determined to satisfy the corresponding candidate region criteria for color, based on the corresponding candidate region criteria for the inverted observation image and the upright observation image, as candidate regions included in regions that are candidates for the normal retinal tissue region; and a normal region determination step of determining, in the candidate region in each of the inverted observation image and the upright observation image, the candidate region that is determined to satisfy the corresponding normal region criteria, based on the normal region criteria for brightness and / or morphology, that are corresponding to the inverted observation image and the upright observation image, as the normal retinal tissue region.

2. The method according to claim 1, wherein the inverted observation image and the upright observation image are positionally correlated so that positions within each of the inverted observation image and the upright observation image correspond to the same positions of the cell aggregates as viewed from a vertical direction, and the normal region determination step determines, in the candidate region of each of the inverted observation image and the upright observation image, a candidate region that is determined to simultaneously satisfy the corresponding normal region criteria for brightness and shape corresponding to each of the inverted observation image and the upright observation image, as the normal retinal tissue region.

3. The method of claim 2, wherein the normal region determination step includes: a region portion evaluation step of determining, for the candidate region identified in the image region of the cell aggregate that is positionally associated with each other in the inverted observation image and the upright observation image, an evaluation score for each of the region portions existing in a range obtained by dividing the candidate region into a plurality of regions along a curve representing the extension direction of the candidate region, based on evaluation criteria related to color and / or shape corresponding to each of the inverted observation image and the upright observation image; and a continuous length-based normal region determination step of determining, as the normal retinal tissue region, a region portion of the candidate region for which the evaluation scores in both the inverted observation image and the upright observation image are equal to or greater than the corresponding normal region determination criterion value, if it is determined that the region portion exists continuously along the curve representing the extension direction of the candidate region for a length equal to or greater than a predetermined normal region determination criterion length.

4. The method of claim 3, wherein the candidate region determination criteria in the candidate region identification step further include the brightness of the region to be identified being higher than the maximum or average brightness of the central region of the image region of the cell aggregate by a predetermined degree or more.

5. The method of claim 4, wherein the candidate region determination criteria in the candidate region identification step further include that, in the region to be identified, the outermost contour of the region portion contained therein, as viewed from the center of the image region of the cell aggregate, substantially matches the contour of the image region of the cell aggregate.

6. The method of claim 5, wherein the candidate region determination criteria in the candidate region identification step further include continuous change in the angle of the tangent of the outermost contour of the region to be identified.

7. The method of claim 6, wherein the candidate region determination criteria in the candidate region identification step further include the width between the outermost and innermost parts of the region to be identified as viewed from the center of the image region of the cell aggregate being greater than or equal to a predetermined reference width.

8. The method of claim 7, wherein the candidate region identification step further includes a non-target cell identification step of identifying whether or not non-target cells are present in the cell aggregate image region of each of the inverted observation image and the upright observation image based on predetermined non-target cell determination criteria based on color and morphology, and the candidate region determination criteria further includes the non-existence of the non-target cells in the region to be identified.

9. The method of claim 3, wherein the region portion evaluation step determines, for the candidate region identified in the image region of the cell aggregate that is positionally associated with each other in the inverted observation image and the upright observation image, an evaluation score for each of the region portions that exist in a plurality of divided ranges within the candidate region that exist within a range of a predetermined angle as viewed from the center of the image region of the cell aggregate, based on evaluation criteria related to the shape that correspond to each of the inverted observation image and the upright observation image; and the continuous length-based normal region determination step determines, as having the cell aggregate, the region portions that are determined to have the evaluation scores in both the inverted observation image and the upright observation image that are equal to or greater than the corresponding normal region determination criterion value and that exist consecutively for a predetermined normal region determination criterion angle or more.

10. The method of claim 7, wherein the reference width for the candidate region identification step is 0.03 mm for the inverted-viewed image and 0.15 mm for the upright-viewed image.

11. The method according to claim 3, wherein the normal region determination reference length in the continuous length-based normal region determination step is 0.4 mm.

12. The method according to claim 9, wherein the normal region determination reference angle in the continuous length-based normal region determination step is 45 degrees.

13. The method of claim 3, wherein the image data of the inverted and upright observation images of the cell aggregate, which may include the retinal tissue, are obtained by photographing the cell aggregate suspended in a liquid.

14. The method of claim 13, wherein the image data of the inverted and upright observation images of the cell aggregate, which may include the retinal tissue, are obtained by photographing the cell aggregate in a plastic container.

15. The method of claim 14, wherein the plastic container is a container having one or more isolated incubation areas.

16. The method of claim 15, wherein the plastic container is a spherical or cone-shaped container having a downward convex shape or depression.

17. A method for producing retinal tissue that can be transplanted into an animal by excising the normal retinal tissue area determined by the method of any one of claims 3 to 16.

18. A method for determining a normal retinal tissue area that can be transplanted into an animal, further comprising the steps of: detecting the expression of neural retinal cell-related genes and non-neural retinal cell-related genes in the normal retinal tissue area determined by the method of any one of claims 3 to 16; and determining that the normal retinal tissue area can be transplanted into an animal if expression of the neural retinal cell-related genes is observed and expression of the non-neural retinal cell-related genes is not observed.

19. A method for producing retinal tissue transplantable into an animal, further comprising the steps of: extracting a portion or all of the normal retinal tissue area determined by the method of any one of claims 3 to 16 as a quality assessment sample; detecting the expression of neural retinal cell-related genes and non-neural retinal cell-related genes in the quality assessment sample; and, when expression of the neural retinal cell-related genes is observed but expression of the non-neural retinal cell-related genes is not observed, determining that (1) the neural retina (neural retina for transplantation) in the same cell aggregate as the portion of the cell aggregate containing the quality assessment sample, (2) the neural retina (neural retina for transplantation) in a cell aggregate from the same lot as the portion of the cell aggregate containing the quality assessment sample, or (3) the neural retina (neural retina for transplantation) in a cell aggregate from the same lot as the entire cell aggregate of the quality assessment sample, as usable as a neural retina for transplantation; and cutting out the normal retinal tissue area determined to be usable as the neural retina for transplantation.

Images