Skin condition estimation method

Abstract

To provide an estimation method capable of estimating skin tendency quickly.SOLUTION: An estimation method estimates skin tendency by identifying one or more of a cell area, a delaminated area and a single-cell area in a stratum corneum image, and comparing stratum corneum parameters obtained from these areas with a specific cutoff value.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a method for estimating skin tendency. [Background technology]

[0002] Skin tendencies, such as susceptibility to skin trouble caused by cosmetics and changes that occur after sunburn, vary from person to person. As methods for identifying skin tendencies, for example, Patent Document 1 proposes a method for distinguishing skin types using seasonal variations in secreted lipids as an index, and Patent Document 2 proposes a method for distinguishing sensitive skin using the regularity of the arrangement of stratum corneum cells as an index. The assessment method of Patent Document 1 requires long-term sample collection and takes time to obtain results, while the assessment method of Patent Document 2 requires the collected stratum corneum cells to be visually ranked for regularity in their arrangement, which may result in variability depending on the assessor. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2000-279417 [Patent Document 2] Japanese Patent Application Laid-Open No. 2010-194267 Summary of the Invention [Problem to be solved by the invention]

[0004] An object of the present invention is to provide an estimation method that can quickly estimate skin tendency. [Means for solving the problem]

[0005] The present invention was made after extensive research to solve the above problems and the discovery that skin conditions can be estimated using the stratum corneum structure obtained from an observation image of the stratum corneum as an index.

[0006] Specifically, the means for solving the problems of the present invention are as follows. 1. Identify one or more of the cellular region, the delaminated region, and the single cell region in the stratum corneum image. calculating one or more stratum corneum parameter values ​​selected from the group consisting of a single cell area, circularity, a rectangular area circumscribing a single cell, a ratio of the length and width of the rectangle circumscribing a single cell, a single cell average brightness, a single cell standard deviation brightness, a single cell median brightness, a single cell maximum brightness, a single cell minimum brightness, and an approximate tetragon-hexagon; The estimation method is characterized by estimating the likelihood of skin roughness occurring after application of cosmetics by comparing the stratum corneum parameter value with a cutoff value. 2. Identify one or more of the cellular region, the delaminated region, and the single cell region in the stratum corneum image. calculating one or more stratum corneum parameter values ​​selected from the group consisting of a single cell area, a single cell perimeter, circularity, an area of ​​a rectangle circumscribing a single cell, a ratio of the length to the width of a rectangle circumscribing a single cell, and an approximate tetragon to hexagon; The estimation method is characterized by estimating the likelihood of redness and itching occurring after applying cosmetics by comparing the stratum corneum parameter value with a cutoff value. 3. Identify one or more of the cellular region, the delaminated region, and the single cell region in the stratum corneum image; calculating one or more stratum corneum parameter values ​​selected from the group consisting of a cell region area and a rectangular area circumscribing one cell; The estimation method is characterized in that the likelihood of reddening after sunburn is estimated by comparing the stratum corneum parameter value with a cutoff value. 4. Identify one or more of the cellular region, the delaminated region, and the single cell region in the stratum corneum image; calculating one or more stratum corneum parameter values ​​selected from the group consisting of single cell area, single cell perimeter, circularity, area of ​​a rectangle circumscribing a single cell, ratio of sides of the rectangle circumscribing a single cell, mean single cell luminance, standard deviation single cell luminance, median single cell luminance, minimum single cell luminance, and approximate tetragon-hexagon; The estimation method is characterized in that the likelihood of tanning after sunburn is estimated by comparing the stratum corneum parameter value with a cutoff value. 5. Identify one or more of the cellular region, the delaminated region, and the single cell region in the stratum corneum image; Based on the correlation between one or more stratum corneum parameter values ​​selected from the group consisting of the area of ​​a single cell, the perimeter of a single cell, and the area of ​​a rectangle circumscribing a single cell as explanatory variables and the likelihood of sunburn as a target variable, An estimation method for estimating the likelihood of sunburn from an image of the stratum corneum. 6. An estimation method according to any one of 1. to 5., characterized in that one or more of the cellular region, multilayered peeling region, and single-cell region are identified using a machine learning model that has been trained in advance on multiple training images of the stratum corneum and one or more of the cellular region, multilayered peeling region, and single-cell region determined by visual evaluation in each training image of the stratum corneum. [Effects of the Invention]

[0007] The estimation method of the present invention uses a numerical value obtained from the stratum corneum structure as an index, and can estimate skin conditions very quickly and easily. [Brief explanation of the drawings]

[0008] [Figure 1] 10 is a graph showing the relationship between the minimum brightness value and the number of people in each of two groups with different skin tendencies who gave different answers to skin questionnaire Q1 in the example. [Figure 2] This graph in Figure 1 shows the number of people who are judged to be positive or negative when the ROC optimal value is used as the cutoff value. DETAILED DESCRIPTION OF THE INVENTION

[0009] The first estimation method of the present invention comprises: Identifying one or more of a cellular region, a delaminated region, and a single cell region in the stratum corneum image; calculating one or more stratum corneum parameter values ​​selected from the group consisting of a single cell area, circularity, a rectangular area circumscribing a single cell, a ratio of the length and width of the rectangle circumscribing a single cell, a single cell average brightness, a single cell standard deviation brightness, a single cell median brightness, a single cell maximum brightness, a single cell minimum brightness, and an approximate tetragon-hexagon; By comparing this stratum corneum parameter value with the cutoff value, the likelihood of skin roughness occurring after applying cosmetics is estimated.

[0010] The second estimation method of the present invention comprises: Identifying one or more of a cellular region, a delaminated region, and a single cell region in the stratum corneum image; calculating one or more stratum corneum parameter values ​​selected from the group consisting of a single cell area, a single cell perimeter, circularity, an area of ​​a rectangle circumscribing a single cell, a ratio of the length to the width of a rectangle circumscribing a single cell, and an approximate tetragon to hexagon; By comparing this stratum corneum parameter value with the cutoff value, the likelihood of redness and itching occurring after applying cosmetics can be estimated.

[0011] The third estimation method of the present invention comprises: Identifying one or more of a cellular region, a delaminated region, and a single cell region in the stratum corneum image; calculating one or more stratum corneum parameter values ​​selected from the group consisting of a cell region area and a rectangular area circumscribing one cell; The susceptibility to reddening after sunburn is estimated by comparing this stratum corneum parameter value with the cutoff value.

[0012] The fourth estimation method of the present invention comprises: Identifying one or more of a cellular region, a delaminated region, and a single cell region in the stratum corneum image; calculating one or more stratum corneum parameter values ​​selected from the group consisting of single cell area, single cell perimeter, circularity, area of ​​a rectangle circumscribing a single cell, ratio of sides of the rectangle circumscribing a single cell, mean single cell luminance, standard deviation single cell luminance, median single cell luminance, minimum single cell luminance, and approximate tetragon-hexagon; The susceptibility to sunburn is estimated by comparing this stratum corneum parameter value with the cutoff value.

[0013] The fifth estimation method of the present invention comprises: Identifying one or more of a cellular region, a delaminated region, and a single cell region in the stratum corneum image; Based on the correlation between one or more stratum corneum parameter values ​​selected from the group consisting of the area of ​​a single cell, the perimeter of a single cell, and the area of ​​a rectangle circumscribing a single cell as explanatory variables and the likelihood of sunburn as a target variable, Estimate the likelihood of tanning after sunburn from images of the stratum corneum.

[0014] The estimation method of the present invention will be described in detail below. Although the stratum corneum can be collected for imaging by either biopsy or tape stripping, tape stripping is preferred because it places less strain on the subject. Tape stripping involves applying adhesive tape to the skin and then peeling it off to collect the surface layer of the skin.

[0015] The corneal cells collected by tape stripping are imaged using transmitted light (including differential interference contrast, phase contrast, and dark-field observation) or reflected light to observe the cell morphology. Corneal cells are preferably observed unstained, but can also be stained if necessary. Cells can be observed using a microscope capable of cell observation, such as the Digital Microscope VHX-500 manufactured by Keyence Corporation or the Digital Microscope Dino-Lite manufactured by AnMo Electronics Corporation. There are no limitations on the observation conditions as long as the details of the cells can be confirmed, but examples include conditions such as a resolution of 1.0 μm / pixel or more and approximately 200,000 pixels or more.

[0016] The captured stratum corneum image is processed using an image processing system to identify one or more of the cellular region, the layered peeled region, and the single-cell region of the stratum corneum image. Then, based on the identified one or more of the cellular region, the layered peeled region, and the single-cell region, one or more of the following is quantified depending on the skin tendency to be estimated and the estimation method: the cellular region area, the single-cell area, the single-cell perimeter, the circularity, the area of ​​a rectangle circumscribing a single cell, the ratio of the length and width of the rectangle circumscribing a single cell, the average luminance of a single cell, the standard deviation of luminance of a single cell, the median luminance of a single cell, the maximum luminance of a single cell, and the minimum luminance of a single cell. The stratum corneum image may be a color image, an image converted to grayscale by image processing, or both.

[0017] The cell region area is the total area of ​​the stratum corneum cell region in the stratum corneum image. The area of ​​multilayer peeling is the area of ​​the region where two or more layers of the stratum corneum have been peeled off. The layer delamination rate is the ratio of the "layer delamination area" to the "cell region area" (layer delamination area / cell region area). The area of ​​one cell is the average area of ​​each individual stratum corneum cell in the stratum corneum image. The cell perimeter is the average value of the perimeters of individual stratum corneum cells in an image of the stratum corneum. Circularity is the average value of the circularity of each individual stratum corneum cell area (calculated by 4π×(area) / (square of perimeter)) for all cells in one image. The rectangular area circumscribing one cell is the average area of ​​the smallest rectangle that covers the entire individual stratum corneum cell in the stratum corneum image across all cells. The ratio of the long and short sides of the rectangle circumscribing one cell is the average ratio of the long and short sides (long side / short side) of the rectangle with the smallest area that covers the entire individual stratum corneum cell in the stratum corneum image for all cells.

[0018] The average brightness of one cell is the average brightness value within the cell region of each individual stratum corneum cell in the stratum corneum image, averaged over all cells. The standard deviation of brightness per cell is the value obtained by calculating the standard deviation of brightness within the cell region of each stratum corneum cell in the stratum corneum image. The single-cell median brightness is the average of the median brightness values ​​within the cell region of each individual stratum corneum cell in the stratum corneum image across all cells. The maximum brightness value of one cell is the average of the maximum brightness values ​​within the cell region of each individual stratum corneum cell in the stratum corneum image across all cells. The minimum luminance value of one cell is the average of the minimum luminance values ​​within the cell region of each individual stratum corneum cell in the stratum corneum image across all cells. The approximate tetrahexagon is a regular tetrahexagon placed within both stratum corneum cell regions, with the center of gravity of the shape of each individual stratum corneum cell at the center, and the angle and size are adjusted so that the area of ​​the area protruding from the regular tetrahexagon is minimized. When the area of ​​the protruding area is minimized, the area of ​​the protruding area is divided by the stratum corneum cell area, and the n that gives the smallest value among regular n-gons (n ​​is 4 to 6) is defined as the approximate n-gon, and the n of each stratum corneum cell in the stratum corneum image is averaged across all cells. Hereinafter, the stratum corneum parameter value refers to a numerical value of one or more values ​​selected from the group consisting of cell region area, single cell area, single cell perimeter, circularity, area of ​​a rectangle circumscribing a single cell, ratio of sides of a rectangle circumscribing a single cell, mean single cell luminance, standard deviation single cell luminance, median single cell luminance, maximum single cell luminance, and minimum single cell luminance.

[0019] The digitization using an image processing system can be performed using a known image processing system, such as the image processing software provided with the digital microscope described above, commercially available image processing software, etc. Alternatively, an image processing system can be used that includes a plurality of training stratum corneum images and a machine learning model that has been trained to learn one or more of the following in each training stratum corneum image: a cellular region (a region where cells are present), a multilayer delamination region (a region where two or more layers of stratum corneum have been delaminated) determined by visual evaluation, and a single-cell region (a region of an individual cell) determined by visual evaluation. The stratum corneum image to be quantified is input into this machine learning model, and one or more of the cellular area, delaminated area, and single-cell area of ​​this stratum corneum image are output. Based on the cellular area, delaminated area, and single-cell area output by the machine learning model, stratum corneum parameter values ​​(cellular area area, single-cell area, single-cell perimeter, circularity, area of ​​a rectangle circumscribing a single cell, ratio of sides of a rectangle circumscribing a single cell, average single-cell brightness, standard deviation single-cell brightness, median single-cell brightness, maximum single-cell brightness, and minimum single-cell brightness) are calculated, thereby enabling the rapid calculation of stratum corneum parameter values ​​equivalent to those based on visual evaluation. Furthermore, one or more of the stratum corneum area, delaminated area, and single-cell area output by the image processing system equipped with the machine learning model can also be visually corrected by a human to determine stratum corneum parameter values.

[0020] The first to fourth estimation methods of the present invention were developed based on the finding that there is a statistically significant difference between the skin tendency described in each estimation method and the stratum corneum parameter value. That is, there is a significant difference in the average values ​​of the stratum corneum parameter between a group with a specific skin tendency and a group without this specific skin tendency. Then, by comparing the stratum corneum parameter value of a subject with a cutoff value, for example, if the stratum corneum parameter value is greater than the cutoff value, it can be estimated that the subject has the specific skin tendency.

[0021] Here, the cutoff value is a value used in the pharmaceutical field to distinguish between disease and non-disease groups, and is determined to determine disease as follows: a value below the cutoff value is negative, a value above the cutoff value is positive, or a value below the cutoff value is positive, and a value above the cutoff value is negative. Ideally, the distributions of measured values ​​between the diseased and non-disease groups should not overlap, but they usually do. Therefore, when a subject is assessed using a cutoff value, there will be "false negatives" where the subject is judged to be negative despite belonging to the diseased group, and "false positives" where the subject is judged to be positive despite belonging to the non-disease group (Table 1). [Table 1] Here, the value of true positives / (true positives + false negatives) is called sensitivity, and the value of true negatives / (false positives + true negatives) is called specificity. By changing the cutoff value up or down, the sensitivity and specificity values ​​also change. The usefulness of the cutoff value is evaluated by sensitivity and specificity, and it is preferable that both sensitivity and specificity are high.

[0022] In the present invention, examples of cutoff values ​​include the average of the mean stratum corneum parameter values ​​of a group with a particular skin tendency and the mean stratum corneum parameter values ​​of a group without that particular skin tendency, the value closest to a true positive rate (sensitivity) of 100% and a false positive rate (1-specificity) of 0% on an ROC curve), a value resulting in a particular sensitivity (e.g., 70%, 80%, 90%), or a value resulting in a particular specificity (e.g., 70%, 80%, 90%). However, it is preferable that both the sensitivity and specificity are 50% or higher, more preferably both are 60% or higher, and even more preferably both are 70% or higher.

[0023] The fifth estimation method of the present invention estimates the susceptibility to tanning after sunburn from an image of the stratum corneum based on a correlation in which one or more stratum corneum parameter values ​​selected from the group consisting of the area of ​​a single cell, the perimeter of a single cell, and the area of ​​a rectangle circumscribing a single cell are used as explanatory variables, and the susceptibility to tanning after sunburn is used as the response variable. The method for estimating the likelihood of sunburn using the quantified stratum corneum parameter values ​​as indicators is not particularly limited, and known methods can be used. For example, the stratum corneum parameter values ​​may be subjected to simple regression analysis or multiple regression analysis. The multiple regression analysis may be linear multiple regression analysis or nonlinear multiple regression analysis. Furthermore, parameters other than the single-cell area, single-cell perimeter, and rectangular area circumscribing a single cell may be included as explanatory variables in the multiple regression analysis. [Example]

[0024] <Machine learning sample> 412 women (aged 18-87 years, average age 46.8 years) were subjected to tape stripping. Corneal cells were collected from the face. The collected stratum corneum cells were photographed unstained using a digital microscope (Keyence Corporation, VHX-5000) with 8-bit (RGB color), 0.41 μm / pixel resolution, and 1600 x 1200 pixels using transmitted light. Two to three fields of view were photographed for each sample and used for machine learning.

[0025] Using the image annotation software Labelme, cellular regions, delaminated regions, and single-cell regions were visually identified to obtain training images of the stratum corneum. We machine-learned each training stratum corneum image and the cellular area, multilayered peeled area, and single-cell area based on visual evaluation, and obtained a machine learning model that can output the cellular area, multilayered peeled area, and single-cell area.

[0026] <Sample> Corneal cells were collected from the faces of 227 women (28-87 years old, average age 47.2 years) using the tape stripping method. Three to five samples were collected from each person, for a total of 966 samples. The sample preparation method was the same as for the machine learning samples. The sample images were analyzed using an image processing device equipped with a machine learning model, and the cell region area, single cell area, single cell perimeter, circularity, area of ​​the rectangle circumscribing a single cell, ratio of the length and width of the rectangle circumscribing a single cell, mean single cell brightness, standard deviation single cell brightness, median single cell brightness, maximum single cell brightness, and minimum single cell brightness were quantified for each stratum corneum image to determine stratum corneum parameter values.

[0027] For 966 samples, the cell area and layer-delaminated area determined by visual evaluation were compared with the cell area and layer-delaminated area analyzed by an image processing device equipped with a machine learning model. The results were as follows: cell area area determined by visual evaluation / cell area area analyzed by an image processing device equipped with a machine learning model = 95%, and the layer-delaminated area area determined by visual evaluation / layer-delaminated area area analyzed by an image processing device equipped with a machine learning model = 86%, indicating that the cell area and layer-delaminated area determined by visual evaluation were equivalent to the cell area and layer-delaminated area analyzed by an image processing device equipped with a machine learning model.

[0028] <Skin condition survey> The questionnaire shown in Table 2 below was administered to the 227 women sampled above. [Table 2]

[0029] The first to fourth estimation methods of the present invention will be explained using as an example the relationship between the result of Q1 and the minimum luminance value, which is a stratum corneum parameter value. Figure 1 is a graph showing the relationship between the minimum brightness value and the number of people in each of the two groups with different skin tendencies who answered differently in Q1. As shown in Figure 1, it was confirmed that there was a significant difference in the average minimum brightness value between the group who had experienced skin irritation from cosmetics and the group who had not experienced skin irritation from cosmetics.

[0030] As shown in Figure 2, when the ROC optimal value (minimum brightness value = 43.14) was used as the cutoff value, of the 127 people who answered that they had experience, 101 were judged to be positive (true positive) and 26 were judged to be negative (false negative).Of the 286 people who answered that they had no experience, 77 were judged to be positive (false positive) and 209 were judged to be negative (true negative), resulting in a sensitivity (= true positive / (true positive + false negative)) of 80% and a specificity (= true negative / (false negative + true negative)) of 73%. Based on these results, if a stratum corneum sample is taken from a subject, the minimum luminance value is measured, and compared with the cutoff value (43.14), it can be determined with a sensitivity of 80% and a specificity of 73% that if the measured minimum luminance value is greater than the cutoff value, the subject is more likely to experience skin irritation after applying cosmetics, and if the measured minimum luminance value is less than the cutoff value, the subject is less likely to experience skin irritation after applying cosmetics.

[0031] Table 3 shows the mean values ​​and p-values ​​for each group for the stratum corneum parameter values ​​that were significantly different between groups that answered differently in the questionnaire. Table 4 also shows the mean value (the value of [(a) + (b)] / 2, where (a) is the mean value of the stratum corneum parameter values ​​of the negative subjects and (b) is the mean value of the stratum corneum parameter values ​​of the positive subjects) for each stratum corneum parameter value that showed a significant difference, the ROC optimum value, and the threshold value (cutoff value), sensitivity, and specificity when the value at which sensitivity was 90% was used as the cutoff value. [Table 3]

[0032] [Table 4] It was confirmed that the first to fourth estimation methods of the present invention make it possible to determine skin tendencies from stratum corneum parameter values.

[0033] "Simple regression analysis" Simple regression analysis was performed on each stratum corneum parameter value and the value obtained in Q5 representing the degree of skin darkening after sunburn. For correlations observed, the correlation coefficient r, p-value, and simple regression line equation (Y = aX + b, where X is each stratum corneum parameter value and Y is the questionnaire value) are shown in Table 5. [Table 5] As shown in Table 5, the stratum corneum parameter values ​​selected from the group consisting of the area of ​​one cell, the perimeter of one cell, and the rectangular area circumscribing one cell were correlated (|r| ≧ 0.2) with the degree of skin darkening after sunburn.

Claims

1. Identifying one or more of a cellular region, a layered peeled region, and a single cell region in the stratum corneum image; calculating one or more stratum corneum parameter values ​​selected from the group consisting of the area of ​​a rectangle circumscribing one cell, the ratio of the length and width of the rectangle circumscribing one cell, the average brightness of one cell, the standard deviation of brightness of one cell, the median brightness of one cell, the maximum brightness of one cell, the minimum brightness of one cell, and an approximate tetragon or hexagon; The estimation method is characterized by estimating the likelihood of skin roughness occurring after application of cosmetics by comparing the stratum corneum parameter value with a cutoff value.

2. Identifying one or more of a cellular region, a layered peeled region, and a single cell region in the stratum corneum image; calculating one or more stratum corneum parameter values ​​selected from the group consisting of the perimeter of a single cell, the area of ​​a rectangle circumscribing a single cell, the ratio of the long and short sides of the rectangle circumscribing a single cell, and an approximate tetragon or hexagon; The estimation method is characterized by estimating the likelihood of redness and itching occurring after application of cosmetics by comparing the stratum corneum parameter value with a cutoff value.

3. Identifying one or more of a cellular region, a layered peeled region, and a single cell region in the stratum corneum image; calculating one or more stratum corneum parameter values ​​selected from the group consisting of a cell region area and a rectangular area circumscribing one cell; The estimation method is characterized in that the likelihood of reddening after sunburn is estimated by comparing the stratum corneum parameter value with a cutoff value.

4. Identifying one or more of a cellular region, a layered peeled region, and a single cell region in the stratum corneum image; calculating one or more stratum corneum parameter values ​​selected from the group consisting of a single cell area, a single cell perimeter, circularity, an area of ​​a rectangle circumscribing a single cell, a ratio of the length and width of the rectangle circumscribing a single cell, a single cell average luminance, a single cell standard deviation luminance, a single cell median luminance, a single cell minimum luminance, and an approximate tetragon or hexagon; The estimation method is characterized in that the likelihood of tanning after sunburn is estimated by comparing the stratum corneum parameter value with a cutoff value.

5. Identifying one or more of a cellular region, a layered peeled region, and a single cell region in the stratum corneum image; Based on a correlation between one or more stratum corneum parameter values ​​selected from the group consisting of a single cell area, a single cell perimeter, and a rectangular area circumscribing a single cell as explanatory variables and the likelihood of sunburn as a response variable, An estimation method for estimating the likelihood of sunburn from an image of the stratum corneum.

6. The estimation method according to any one of claims 1 to 5, wherein one or more of the cellular region, the multilayered peeled region, and the single-cell region are identified using a machine learning model that has previously undergone machine learning training using a plurality of training stratum corneum images and one or more of the cellular region, the multilayered peeled region, and the single-cell region determined by visual evaluation in each training stratum corneum image.

Images