AI palmistry: Haut.AI’s “breakthrough” hand analysis predicts age with personalized product recommendations

07 May 2024 --- Skin diagnostics platform Haut.AI releases findings from a “breakthrough” study demonstrating how images of hands can be analyzed using AI to accurately predict age. 

Personal Care Insights speaks to Anastasia Georgievskaya, CEO at Haut.AI, about how the new approach presents a viable alternative to traditional facial photo methods while enhancing personalized product recommendations.

Hand analysis is intended to facilitate more personalized beauty product recommendations for consumers while reducing biases often associated with conventional systems, particularly toward people of color of Indian descent. 

“There are two AI-based models developed for this study: HandAge and FaceAge,” she highlights. “Both platforms can be used to offer personalized skin care recommendations and objectively track the effectiveness of anti-aging products and treatments over time by measuring changes in skin features.”

“While both are designed to predict age, HandAge stands out because it uses images of the dorsal side of the hand to accurately estimate chronological age, which is a more novel approach,” notes Georgievskaya.

Haut.AI claims analyzing hand images for age predictions offers “distinct advantages” over typical facial images. “For example, hand images are less likely to raise privacy concerns, as they do not reveal personal identity in the same way as facial images,” says Georgievskaya.

“Furthermore, hands show consistent signs of aging, such as wrinkles, structure of knuckles, vein prominence and changes in skin texture. These markers are generally less affected by makeup, facial hair or cosmetic procedures. Hands also tend to have less variability due to expressions or facial movements, providing a more stable dataset for analysis.”

AI models trained on hand images achieve “comparable accuracy” to those using facial images.Significant findings for ethnic skin
The study, titled “Predicting human chronological age via AI analysis of dorsal hand versus facial images: A study in a cohort of Indian females,” shows AI models trained on hand images achieve “comparable accuracy” to those using facial images, with an average error of 4.1 and 4.7 years in predicting chronological age.

“This research is particularly significant for ethnic skin, as it was trained using the Indian population dataset and represents the first AI model for age prediction specifically designed with a diverse dataset that includes a wide range of skin tones,” Georgievskaya tells us.

The R&D process for this research paper unfolded in two phases. The initial iteration, which laid the groundwork for the core functionalities, took approximately a year to complete. 

“Following that, we embarked on a series of refinements and optimizations based on real-world testing and user feedback. This iterative improvement phase spanned an additional year,” Georgievskaya details.

Haut.AI previously published research revealing beauty tech biases that are blocking inclusivity. We spoke to Georgievskaya alongside EveLab Insight about outsmarting AI beauty bias amid a rapidly evolving landscape.

Beyond age prediction
We ask Georgievskaya if Haut.AI envisions this hand analysis being used beyond simply predicting age. 

“Absolutely. The system can identify premature aging by analyzing age-related features and comparing them to typical patterns for various age groups,” she tells us.

“Using Deep Convolutional Neural Networks (CNNs), a type of AI architecture known for its ability to analyze complex image data, we can assess features like wrinkles, pigmentation, vein prominence and skin texture.”

Any of these markers that are more pronounced than expected for a given age can signify premature aging, Georgievskaya notes.

“The technology also has the potential to track changes over time, detecting if age-related features are developing faster than usual.”

Haut.AI suggests that specific age-related markers can be used by individuals and beauty professionals to take preventive measures, recommend lifestyle changes or choose specific skin care routines to address the issues and slow the aging process.

It can also alert users to seek medical advice if the detected changes indicate underlying health issues.

Dermatologists and cosmetic professionals can use AI-based hand analyses to assess the progress of anti-aging treatments.Supporting added protection
Additionally, Georgievskaya notes dermatologists and cosmetic professionals can use AI-based hand analyses to assess the progress of anti-aging treatments. 

“By predicting a patient’s age before and after treatment, practitioners can quantify the treatment’s impact. This information can help to guide treatment plans,” she highlights.

“Importantly, this technology can also highlight the need for sun protection beyond the face. Sun exposure is a major contributing factor to premature aging and skin cancer, and the hands are often neglected areas. By showcasing the impact of UV rays on hand appearance, we hope to raise awareness and encourage broader sunscreen application.”

Georgievskaya adds that further combining the datasets of hand and face imagery opens a “world of possibilities.” This dual approach introduces the possibility of learning more about how different body parts age over time.

“By studying different skin features of both the hands and face, we can better understand how these areas are affected by the natural aging process, lifestyle and external factors. This information can help us come up with better ways to stop or slow down the signs of getting older,” she tells us.

“Also, this combined approach has the potential to detect early signs of health issues. For example, if the combined model identifies discrepancies between the aging patterns of hands and face, it could indicate underlying health concerns that are worth looking into further. This aspect adds an extra layer of diagnostic value to the technology.”

Haut.AI has already begun exploring this combined data approach. Overall, leveraging the strengths of both data sets will create a more “complete picture” of the aging process, resulting in improved accuracy in age prediction, Georgievskaya underscores.

“It creates a more robust framework for a wide range of applications while reducing biases. A combined model also has the advantage of cross-validation between facial and hand images. If one set of features has inconsistencies or missing data, the other can fill in the gaps, leading to more reliable results.”

Haut.AI previously teamed up with French company Naos — made up of skin care brands Bioderma, Etat Pur and Institut Esthederm — to launch a digital tool, Skin Observer, which offers “quick and precise” skin analysis and customized beauty rituals.

Haut.AI also published research revealing beauty tech biases that are blocking inclusivity. We spoke to Georgievskaya alongside EveLab Insight about “outsmarting AI beauty bias” amid a rapidly evolving tech landscape.

By Benjamin Ferrer