Visualization of Cutibacterium acnes from facial images using deep learning

Abstract

In order to empower individuals to self-evaluate the effect of daily habit modifications on acne prevention and the state of Cutibacterium acnes (C. acnes) on the face, research was conducted into a technology capable of visualizing C. acnes solely from facial images captured using visible light. Deep learning methodologies, commonly utilized in the field of engineering, were applied to the practice of visualizing C. acnes with ultraviolet light, a technique used in dermatological fields. A deep-learning model was trained to identify the characteristics between pairs of images, captured by illuminating the face with both ultraviolet and visible light. For the sake of enhancing learning efficiency, procedures for cropping these paired images and for emphasizing areas of C. acnes luminescence in ultraviolet images were examined. As a result of this research, we successfully established a deep learning model capable of visualizing C. acnes using only visible light images without the need for specialized equipment. Through the application of this model in our endeavor to make a difference in everyday lives by redesigning habits, we anticipate contributing to the improvement of quality of life by enabling better habit formation for acne prevention and allowing individuals to regain confidence in their facial appearance.