Mechanical learning enhances laser treatment planning

Evaluation of skin features such as the type of Fitzpatrick skin, over -coloring, redness and severity of wrinkles are vital to determine the appropriate laser treatment for patients. With the growing demand for laser processes, the exact evaluation of these features is essential.¹ According to the US Dermatological Surgery Society (ASDS), the percentage of consumers who consider cosmetic procedures increased from 30% in 2013 to 70% in 2023, with laser treatments appearing as the most sought after option.² However, the growing popularity of these procedures has pointed out concerns about the quality of care, especially in doctors’ not -operating arrangements, such as medical spa baths.³

Challenges in skin evaluation and laser procedures

In the US, the availability of trained dermatologists remains limited, with only 1 dermatologist per 29,000 citizens. This lack has led to an increase in non -physicists, including nurses, cosmetics and cosmetologists, performing skin assessments and the administration of laser treatments.⁴ Medical thermal thermal baths, which often use non -medical providers, now exceed cosmetic practices based on doctors in 73% of major US cities. Unfortunately, studies have shown that medical thermal baths have poorer security files and patient results compared to practices under the guidance of doctors. Inaccurate skin evaluations can lead to complications such as burns, scars, infections and even loss of vision.⁵

AI’s role in dermatological evaluations

To address the deficiencies in the accuracy of skin evaluation and to improve the effects of treatment, the researchers have explored the potential of artificial intelligence (AI) and mechanical learning in dermatology. AI has shown impressive capabilities in sorting images, detecting objects and splitting. Previous research has been successfully trained models of mechanical learning to classify the type of Fitzpatrick skin with accuracy rates between 81% and 96%.⁶ Other studies have explored the classification of skin -based skin diseases, such as oil and dryness.⁷ However, to date, no previous study has tried to develop a model of mechanical learning capable of evaluating multiple different skin characteristics at the same time.

The Skinanalysis and Machine Learning data model

A recent study aims to bridge this gap by creating a new set of data, Skinanalysis, consisting of 3,662 images marked by Fitzpatrick skin types, over -coloring, redness and severe wrinkles.⁸ These images came from available in common data sets, ensuring diversity in skin tones, ages, backgrounds and lighting conditions. A dual committee dermatologist commented on the data set using standard dermatological scales, making it a strong foundation for the training of mechanical learning models.

To analyze the data set, the study used 3 established mechanical learning architectures: VGG-16, Reset-50 and Efficiynet, which all preceded the Imagenet. The final model was trained using a specialized loss function, Skinceloss, which was aligned with dermatological scales and prevented the inconsistent predictions. Data growth techniques, such as random translations, rotations and reversals, were applied to strengthen the health of the model.

Results and effects

The best performance model of the study, an effective V2M architecture, achieved an average accuracy of 85.02% and Auroc score 0.8191. The results of the total testing confirmed the powerful capabilities of the model, with an average accuracy of 85.41% and Auroc 0.8306. The study also identified a trend where the model had better performance at extreme prices of each dermatological scale, but showed lower precision in medium -range classifications, possibly due to increased clinical ambiguity.

These findings indicate that dermatological evaluation models based on AI may provide dermatologist level expertise in skin analysis, possibly by enhancing the accuracy of laser treatment design. By incorporating these models into non -medical -based medical arrangements, medical thermal baths and other aesthetic practices may improve the effectiveness of safety and treatment.

Conclusion

This study represents a significant progress in AI -based dermatology, developing a model of mechanical learning capable of evaluating multiple skin features at the same time. The high accuracy and generality of the model suggests that AI has the ability to support non -barbaric providers to carry out more accurate skin evaluations, thereby reducing the complications associated with inappropriate lasers. Researchers have suggested that future research should focus on expanding the set of data, improving the design of therapy based on AI and the integration of AI -based tools to enhance the effectiveness of patient safety and treatment. Utilizing AI in dermatological evaluations, the field can proceed with more personalized, effective and safer cosmetic procedures.

References

1. Butani A, Dudelzak J, Goldberg DJ. Recent developments in laser dermatology. J Cosmet Laser Ther. 2009; 11 (1): 2-10. DOI: 10.1080/14764170802524411
2. Consumer research for cosmetic dermatological procedures. ASDS. 2021. Access March 27 2025. https://www.asds.net/medical-professionals/practice-resources/consumer-on.
3. Valiga A, Albornoz Ca, Chitsazzadeh V, et al. Medical installations spa and non -aesthetic operators. Clin Dermatol. 2022, 40 (3): 239-243. DOI: 10.1016/J.Clindermatol.2021.11.007
4. Professional employment and salaries. US statistics, professional statistics, no. 2. Published in 2019, access to March 27, 2025. https://www.bls.gov/oe/current/OES535011.htm#nat.
5. Rossi AM, Wilson B, HIBLER BP, DRAKE LA. Non -physical practice of cosmetic dermatology: a prospect of a patient and a physician of results and unwanted actions. Dermatol Surg. 2019; 45 (4): 588-597. DOI: 10.1097/DSS.0000000000001829
6. CHANG CC, HSING S, CHUANG Y, et al. Of course skin type classification using concentration of neural networks. 2018 13th IEEE Congress for Industrial Electronics and Applications (ICIAA), Wuhan, China, 2018, pp. 2011-2014, DOI: 10.1109/iciea.2018.8398040.
7. Saiwaeo S, Arwatchananukul S, Mungmai L, Preedalikit W, Aunsri N. Human Skin Sorting Using image processing and deep learning approaches. Heliyon, Volume 9, Issue 11, 2023, E21176, ISSN 2405-8440. DOI: 10.1016/J.HELIYON.2023.E21176.
8. Draelos RL, Kesty Ce, Kesty Kr. Artificial intelligence predicts the type of Fitzpatrick skin, coloring, redness and severity of wrinkles from color photos of the face. J Cosmet Dermatol. 2025; 24 (4): E70050. Doi: 10.1111/Jocd.70050