Can Deep Learning Replace Skin Testing?

Diagnosing with 98.3% accuracy… The era of AI skin irritation assessment is here

The era where AI can read skin conditions has arrived. Recent research findings demonstrate that AI technology could surpass the traditional methods that relied solely on human experts. We examine just how advanced this AI technology has become in automatically diagnosing skin irritation.

Asking AI about skin irritation diagnosis

Deep learning is pushing the boundaries of cosmetic science further. The recent development of ‘AI-based automatic skin irritation diagnosis technology’ has attracted significant attention in the field of dermatology. It holds great significance beyond a mere trend or experimental approach, proving clinical accuracy. This technology is particularly notable for its ability to automatically analyze patch test results that previously required expert judgment.

Source of Article

This study was published in the online edition of the official journal of the European Society of Contact Dermatitis (ESCD), ‘Contact Dermatitis’. The research team developed a model to automatically analyze skin reactions in patch test images using state-of-the-art object detection algorithms, specifically YOLOv5x.

“By introducing deep learning into patch test interpretation, we were able to reduce subjectivity and enhance the consistency and speed of diagnoses.”

AI trained on more than 80,000 data cases… Results in 98.3% accuracy

This AI model is more than a mere image analysis tool. Its performance shines through the amount and precision of the data it has been trained on. The study utilized 83,629 patch test images with irritation scores for training, and tested accuracy with 1,312 evaluation data and 1,536 validation data.

The AI’s assessment showed a high accuracy rate of 98.3% for both 24-hour and 48-hour readings of irritation response, with a sensitivity rate of 99.7% for no-irritation (0 score) judgments, proving its clinical applicability.

“Stable analysis by AI in irritation tests requiring observation over time provides accurate results without repeated experiments, marking a significant technological advancement.”

This result indicates a potential shift from traditional reliance on human observation in scoring skin irritation to a more objective AI-based approach.

Fewer Experts, Higher Accuracy… The Change in AI Diagnosis

Patch tests involve applying various substances to the skin to induce irritations, then observing reactions after a set time to determine sensitivity or allergy potential. The challenge has been the time consumption and variation in interpretations among experts.

However, this AI-based model reduces reliance on experts while achieving high consistency and objectivity. It allows for faster evaluations, garnering attention for both practicality and diagnostic efficiency.

Beyond technical challenges, AI’s role is expected to expand in securing cosmetic safety across product quality control and human application trials.

Continuous Technological Evolution Turns Challenges into Opportunities

This technology, exceeding a singular function, has significant potential to develop into more complex analytical systems in the future. For instance, given the proven accuracy in photo-based interpretations, it can be expanded to simulated skin reaction identification and test environment automation.

The research team emphasized that automatically classifying irritant reactions and focusing on abnormal reactions boost research and cost efficiency. This serves as a catalyst for acquiring skin safety data quickly and on a wide scale.

“When AI-based analysis complements human judgment, it will significantly impact consumer trust and product quality in the long run.”

---

To foster continual growth in the cosmetic industry, YURICO5 is developing AI technologies with high reliability and efficiency to enhance the safety and quality of cosmetics. Such technological advancements will positively impact both industrial growth and future economic development.