System for providing diagnostic script for scalp and hair loss condition based on artificial intelligence algorithm
A system for providing diagnostic script for scalp and hair loss condition, includes a data receiving unit that receives scalp photographed image data of a patient to be diagnosed with a scalp condition and a hair loss condition; a classification unit that inputs the received scalp photographed image data of the patient into a first learning model that has been learned to acquire result values for a plurality of measurement items and classifies the scalp condition and the hair loss condition using the acquired result values; and a diagnosis result providing unit that inputs information on the classified scalp condition and hair loss condition into a second learning model and outputs diagnosis script data.
1 . A system for providing diagnostic script, comprising:
a data receiving unit configured to receive scalp photographed image data of a patient to be diagnosed with a scalp condition and a hair loss condition;
a classification unit configured to input the received scalp photographed image data of the patient into a first learning model to acquire result values for a plurality of measurement items and classify the scalp condition and the hair loss condition using the acquired result values;
a diagnosis result providing unit configured to input information on the classified scalp condition and hair loss condition into a second learning model and output diagnosis script data for the patient;
a data collection unit configured to collect a plurality of scalp photographed images and a plurality of expert diagnosis scripts;
a labeling data collection unit configured to collect labeling data including scores and keywords for each measurement item corresponding to the scalp photographed images and the expert diagnosis scripts; and
a learning unit configured to:
build the first learning model using the scalp photographed images and the labeling data;
build the second learning model using the expert diagnosis scripts and labeling data;
cause the first learning model to learn to output the result values of the plurality of measurement items; and
cause the second learning model to learn to generate a diagnostic script using a classification result which is output through the first learning model,
wherein, to acquire a hair thickness among the measurement items, the classification unit is configured to:
extract a plurality of first pixels having a value greater than a preset reference pixel value, wherein the reference pixel value represents a pixel value to be identified as a hair;
count the extracted first pixels to extract the number of the first pixels; and
acquire the hair thickness by using the number of the extracted first pixels and a length of the first pixel, wherein the length of the first pixel is calculated using a ratio of a hair thickness in the scalp photographed image to an actual hair thickness.
2 . The system for providing diagnostic script according to claim 1 , wherein the measurement items include at least one of keratin, oil, sensitivity, the hair thickness, the number of hairs per hair follicle, erythema, and pustules.
3 . The system for providing diagnostic script according to claim 1 , wherein the learning unit is configured to cause the first learning model to learn to extract result values for keratin, oil, sensitivity, the number of hairs per hair follicle, the hair thickness, erythema, and pustules in a form of score using the scalp photographed images and the labeling data.
4 . The system for providing diagnostic script according to claim 1 , wherein the learning unit is configured to perform a tokenization pre-processing on the collected expert diagnosis scripts using a tokenizer, and cause the second learning model to learn to generate the diagnosis script using pre-processed expert diagnostic scripts and labeling data.
5 . The system for providing diagnostic script according to claim 1 , wherein the classification unit is configured to:
acquire a result value for keratin among the measurement items according to presence or absence of foreign substances in a scalp area,
acquire a result value for oil among the measurement items according to presence or absence of oil or foreign substances in a hair area; and
acquire a result value for sensitivity among the measurement items by comparing pixel values of the scalp area with a reference pixel value which is a pixel value corresponding to red color.
6 . The system for providing diagnostic script according to claim 1 , wherein the classification unit is configured to acquire a result value for the number of hairs grown in each of hair follicles by applying image processing business logic to a hair follicle area.
7 . The system for providing diagnostic script according to claim 1 , wherein, the classification unit is configured to:
acquire result values for keratin, oil, sensitivity, erythema, and pustules from the first learning model,
calculate an average value for the acquired the result values for each of the keratin, the oil, the sensitivity, the erythema, and the pustules or extract a highest score among the result values for each of the keratin, the oil, the sensitivity, the erythema, and the pustules to classify a severity for each of the keratin, the oil, the sensitivity, the erythema, and the pustules as high, medium, and low; and
classify the scalp condition into at least one of dry, oily, sensitive, dandruff, atopic, seborrheic, complex, inflammatory, pustular, and folliculitis using the classified severity.
8 . The system for providing diagnostic script according to claim 6 , wherein the classification unit is configured to:
set reference values for the number of hairs per hair follicle and the hair thickness; and
compare and analyze the result values acquired from the first learning model and the reference values to classify the hair loss condition as at least one of normal, suspected hair loss, and progressing hair loss.