An image analysis method for analyzing an image (78) of a
tissue collected from a subject using a deep learning algorithm (60) of a convolutional neural network comprising an input layer (50a), an output layer (50b) and an intermediate layer (50c) configured by a plurality of layers, the image analysis method comprising: generating analysis data (80) from an analysis target image (78) that includes an analysis target tissue by generating a plurality of analysis data (80) for each region in the analysis target image (78) having a predetermined number of pixels, the analysis data (80) comprising, for each of the predetermined number of pixels, color density value data (81) for one or more primary colors; inputting the analysis data (80) of each region having the predetermined number of pixels to the deep learning algorithm (60) in order to generate, for each region having the predetermined number of pixels, a label value (83) indicating that a pixel located at the center of the analysis data of that region is a nucleus region and estimated value indicating a layer structure of one of a plurality of layer structures or that the pixel located at the center of the analysis data of that region is a region other than a cell nucleus; and generating a layer structure nucleus region weighted image (84) indicating the nucleus region of each layer structure, wherein the deep learning algorithm (60) is trained by: first training data and second training data, and the convolutional neural network learns the relationship between the first and second training data, wherein the second training data includes a label value which indicates a nuclear region of a layer structure and the label value is a value estimating the layer structure. 2. The image analysis method according to claim 1, wherein the analysis target image (78) is an image of a sample for histological diagnosis.
3. The image analysis method according to claim 1 or 2,
wherein the analysis target image (78) is a bright field image of a stained analysis target tissue.
4. The image analysis method according to claim 1, wherein the
plurality of layer structures comprises at least one selected from an epithelial cell layer, an endothelial cell layer, a connective tissue layer, and a muscle layer.
5. The image analysis method according to any one of claims 1
to 4, wherein the tissue is taken from the stomach, small intestine, or large intestine, and the plurality of layer structures includes at least one selected from among a mucosal epithelial cell layer, a mucosal lamina propria, a muscularis mucosae, a submucosal layer, an intrinsic muscle layer, and a serosal layer.
6. The image analysis method according to any one of claims 1
to 4, wherein the tissue is collected from a trachea or a bronchus, and the plurality of layer structures comprises at least one selected from among a mucosal epithelial cell layer, a mucosal lamina propria, a smooth muscle layer, a submucosal layer.
7. The method according to any one of claims 1 to 4, wherein
the tissue is collected from the uterine body, and the plurality of layer structures comprises at least one selected from a mucosal epithelial cell layer, a mucosal lamina propria, a muscle layer, an adventitia layer. 8. The image analysis method according to any one of claims 1 to 4, wherein the tissue is collected from a gall bladder; and the plurality of layer structures includes at least one selected from a mucosal epithelial cell layer, a mucosal lamina propria, a muscle layer, and a serosal layer. 9. The image analysis method according to any one of claims 1 to 4, wherein the tissue is a glandular tissue collected from tissues other than the stomach, small intestine, large intestine and uterus, and the plurality of layer structures comprises at least one selected from a gland epithelial cell layer, a muscle layer, and a connective tissue layer. 10. The method according to any one of claims 1 to 4, wherein the tissue is a heart, and the plurality of layer structures comprises at least one selected from among an endocardial layer, a myocardium layer, an epicardial layer, and an epicardial tissue. 11. The image analysis method according to any one of claims 1 to 4, wherein the tissue is a blood vessel, and the plurality of layer structures comprises at least one selected from an endothelial cell layer and a muscle layer. 12. An image analysis apparatus (100) for analyzing an image (78) of a tissue collected from a subject using a deep learning algorithm (60) of a convolutional neural network comprising an input layer (50a), an output layer (50b) and an intermediate layer (50c) configured by a plurality of layers, the image analysis apparatus (100) comprising: a processor (10) configured to: generate analysis data (80) from an analysis target image (78) that includes a tissue or cell to be analysed by generating a plurality of analysis data (80) for each region in the analysis target image (78) having a predetermined number of pixels, the analysis data (80) comprising, for each of the predetermined number of pixels, color density value data (81) for one or more primary colors; input the analysis data (80) of each region having the predetermined number of pixels to a deep learning algorithm (60) in order to generate, for each region having the predetermined number of pixels, a label value (83) indicating that a pixel located at the center of the analysis data of that region is a nucleus region and estimated value indicating a layer structure of one of a plurality of layer structures or that the pixel located at the center of the analysis data of that region is a region other than a cell nucleus; and generate a layer structure nucleus region weighted image (84) indicating the nucleus region of each layer structure, wherein the deep learning algorithm (60) is trained by: first training data and second training data, and the convolutional neural network learns the relationship between the first and second training data, wherein the second training data includes a label value which indicates a nuclear region of a layer structure and the label value is a value estimating the layer structure.