Systems and methods for skin color classification
Systems and methods for classification and/or analysis of skin color(s) are provided. The technology integrates artificial intelligence (AI) with dermatology to enhance skin color classification. The unique matching technique aligns skin images with the Fitzpatrick skin type palette through Euclidean distance matching, achieving higher precision in skin color classification compared to conventional methods.
1. A system for classification of skin color, the system comprising:
a processor; and
a machine-readable medium in operable communication with the processor and having instructions stored thereon that, when executed by the processor, perform the following steps:
i) receiving first data of at least one image of skin of a subject;
ii) transforming the at least one image of the skin of the subject from an original color space to a standardized color space to give standardized data;
iii) using an unsupervised learning model to perform clustering and matching on the standardized data to give clustered data; and
iv) matching cluster centers from the clustered data to cluster centers of a Fitzpatrick skin type (FST) to classify each image of the skin of the subject into an FST color,
step iii) comprising resizing images of the standardized data, reshaping the images of the standardized data, and iteratively applying the unsupervised learning model until the clustered data is obtained, and
step iv) comprising calculating color distances, between the cluster centers from the clustered data and the cluster centers of the FST, using a colorimetry metric, followed by matching each cluster centered from the clustered data with a nearest cluster center of the FST based on the calculated color distances.
2. The system according to claim 1 , the standardized color space being LAB color space.
3. The system according to claim 1 , the subject being a human subject.
4. The system according to claim 1 , the system further comprising a display in operable communication with at least one of the processor and the machine-readable medium, and
the instructions when executed further comprising:
v) generating a visualization of colors from the at least one image of the skin of the subject with corresponding respective FST colors; and
vi) displaying the visualization on the display.
5. The system according to claim 1 , the system further comprising a ground truth device configured for ground truth classification of the at least one image of the skin of the subject, and
the instructions when executed further comprising:
vii) providing the classification of each image to a clinician to compare to ground truth classifications of each image obtained from the ground truth device.
6. The system according to claim 1 , the unsupervised learning model being K-means, K-means-PCA, K-means mini-batch, DBSCAN, HDBSCAN, OPTICS, a hierarchical model, a Gaussian mixture model, fuzzy C-means, affinity propagation, mean shift, or spectral clustering.
7. The system according to claim 1 , the unsupervised learning model being K-means.
8. The system according to claim 1 , the instructions when executed further comprising:
viii) providing the classification of each image to a clinician to use in diagnosis of a medical condition of the subject.
9. The system according to claim 8 , the medical condition being at least one of diabetes and a diabetic foot ulcer.
10. A method for classification of skin color, the method comprising:
i) receiving first data of at least one image of skin of a subject;
ii) transforming the at least one image of the skin of the subject from an original color space to a standardized color space to give standardized data;
iii) using an unsupervised learning model to perform clustering and matching on the standardized data to give clustered data; and
iv) matching cluster centers from the clustered data to cluster centers of a Fitzpatrick skin type (FST) to classify each image of the skin of the subject into an FST color,
step iii) comprising resizing images of the standardized data, reshaping the images of the standardized data, and iteratively applying the unsupervised learning model until the clustered data is obtained, and
step iv) comprising calculating color distances, between the cluster centers from the clustered data and the cluster centers of the FST, using a colorimetry metric, followed by matching each cluster center from the clustered data with a nearest cluster center of the FST based on the calculated color distances.
11. The method according to claim 10 , the standardized color space being LAB color space.
12. The method according to claim 10 , the subject being a human subject.
13. The method according to claim 10 , further comprising:
v) generating a visualization of colors from the at least one image of the skin of the subject with corresponding respective FST colors; and
vi) displaying the visualization on a display.
14. The method according to claim 10 , further comprising:
vii) providing the classification of each image to a clinician to compare to ground truth classifications of each image obtained from a ground truth device.
15. The method according to claim 10 , the unsupervised learning model being K-means, K-means-PCA, K-means mini-batch, DBSCAN, HDBSCAN, OPTICS, a hierarchical model, a Gaussian mixture model, fuzzy C-means, affinity propagation, mean shift, or spectral clustering.
16. The method according to claim 10 , the unsupervised learning model being K-means.
17. The method according to claim 10 , further comprising:
viii) using the classification of each image to use in diagnosis of a medical condition of the subject.
18. The method according to claim 17 , the medical condition being at least one of diabetes and a diabetic foot ulcer.
19. A system for classification of skin color, the system comprising:
a processor;
a display in operable communication with the processor; and
a machine-readable medium in operable communication with the processor and having instructions stored thereon that, when executed by the processor, perform the following steps:
i) receiving first data of at least one image of skin of a subject;
ii) transforming the at least one image of the skin of the subject from an original color space to a standardized color space to give standardized data;
iii) using an unsupervised learning model to perform clustering and matching on the standardized data to give clustered data;
iv) matching cluster centers from the clustered data to cluster centers of a Fitzpatrick skin type (FST) to classify each image of the skin of the subject into an FST color;
v) generating a visualization of colors from the at least one image of the skin of the subject with corresponding respective FST colors;
vi) displaying the visualization on the display; and
vii) providing the classification of each image to a clinician to compare to ground truth classifications of each image obtained from a ground truth device,
the standardized color space being LAB color space,
the subject being a human subject,
the unsupervised learning model being K-means,
step iii) comprising resizing images of the standardized data, reshaping the images of the standardized data, and iteratively applying the unsupervised learning model until the clustered data is obtained, and
step iv) comprising calculating color distances, between the cluster centers from the clustered data and the cluster centers of the FST, using a colorimetry metric, followed by matching each cluster center from the clustered data with a nearest cluster center of the FST based on the calculated color distances.
20. The system according to claim 19 , the instructions when executed further comprising:
viii) providing the classification of each image to the clinician to use in diagnosis of a medical condition of the subject.