Potential for Automated Imaging Analysis in Facial Aesthetics

Dear Editor, I am writing to bring to your attention to the innovative potential of ChatGPT-4 (Open AI, San Francisco, CA, USA) in the realm of automated imaging analysis for facial aesthetics. ChatGPT-4 is the latest iteration of OpenAI’s Generative Pre-Training Transformer, a language model that uses deep learning to produce human-like text. ChatGPT has the ability to answer questions, tell stories, write web code, and even form concepts and ideas about extremely complicated topics (Gordon et al., 2024). I wish to highlight its application in distinguishing between sinker and sagger facial aging patterns. Individuals who are sinkers are characterized by fat compartment atrophy, especially within the lateral-temporal cheek fat compartments, which leads to a sunken facial appearance. Individuals who are saggers exhibit signs of fat repositioning from the midface to the lower third of the face that results in tissue drooping, bulging, and increased wrinkle formation (Swift et al., 2021). Recent studies have demonstrated the utility of ChatGPT and similar artificial intelligence (AI) models in various imaging applications, including medical image segmentation, radiology, and dermatological analysis (Esteva et al., 2017; Liu et al., 2020). For example, ChatGPT has been used to support automated diagnosis in radiographic imaging, identifying abnormalities with significant precision. Moreover, in dermatology, AI-based tools are able to detect and classify skin lesions from images, improving diagnostic accuracy and providing invaluable assistance to clinicians in early cancer detection (Li et al., 2022). These examples underscore the potential of ChatGPT-4 in analyzing facial imaging, not only for assessing facial aging characteristics but also for broader applications across aesthetic and medical fields. By harnessing these technologies, practitioners can improve both the precision of diagnostic assessments and the quality of patient care. This advancement aligns with insights from a recently published article titled Profhilo® Structura, Current Status and Future Perspectives: A Practical Review (Forte et al., 2024) that discusses the role of adipose tissue depletion and redistribution in facial aging. ChatGPT-4 presents a promising tool to further enhance the precision of facial imaging analysis and evaluation. Moreover, the app can be installed on any smartphone, enabling practitioners to conduct these analyses conveniently and flexibly during patient consultations. Further, ChatGPT-4, allows practitioners to develop specific prompts that can quantify and assess sinker or sagger characteristics in facial images. Such analysis could provide significant utility in aesthetic practices. Following is a predefined prompt that practitioners can copy and paste as a proposed approach for performing this imaging analysis: Analyze facial imaging to quantify the degree of sinker or sagger characteristics by evaluating fat compartment depletion, volume loss, and tissue sagging in the middle and lower third of the face. The analysis should express results as a percentage of sinker and sagger features based on anatomical and aesthetic criteria. Facial Region Identification: Segment the face into the following anatomical regions: lateral-temporal cheek, zygomatic area, nasolabial region, and jowl. Sinker Characteristics: Quantify the extent of fat depletion and hollowness in the lateral-temporal and preauricular areas, identify signs of volume loss leading to sunken cheeks, and calculate the percentage of facial surface area affected by fat loss and sunken appearance in the midface. Sagger Characteristics: Analyze the sagging or drooping of tissue, particularly in the jowl and lower third of the face, evaluate the shift or redistribution of fat from the zygomatic area to the lower face, and measure the degree of tissue laxity, ptosis, and bulging, focusing on the midface and jowl regions. Percentage Calculation: Assign a percentage score for sinker characteristics based on the degree of fat loss and hollowing, assign a percentage score for sagger characteristics based on the extent of tissue sagging and fat redistribution, and based on a visual assessment of the provided images and the outlined criteria, estimate an approximation of the sinker and sagger characteristics (e.g., 60% sagger, 40% sinker). To ensure consistent and accurate results, it is crucial that the images used for analysis meet specific standards of clarity and zoom, similar to those generated by the Canfield Scientific imaging systems (Parsippany, NJ, USA) and QuantifiCare (Suwanee, GA, USA). These qualities include maintaining a consistent distance, lighting, and resolution to capture detailed facial features and fat compartments. In cases where the quality of the images is suboptimal and the ChatGPT-4 is unable to generate a score, the practitioner can still proceed with the analysis by using the following prompt: Approximate score based on visible features. This approach allows for a degree of flexibility in cases where the image quality might hinder the automated analysis, ensuring that even lower-quality images can still provide valuable insights. Based on visual assessments of 60 images (utilizing both front and 45-degree views), the ChatGPT-4 has demonstrated the capacity to effectively evaluate facial aging with an accuracy of 0.775 (77.5%). These preliminary results suggest that ChatGPT-4 could become a valuable asset to aesthetic medicine, particularly for pretreatment assessments. Given that Profhilo® Structura (IBSA Farmaceutici Italia, Lodi, Italy) targets fat compartment atrophy and tissue sagging, this novel AI-driven imaging analysis could serve as a powerful tool for aesthetic physicians by streamlining patient evaluations, improving treatment planning, and offering a precise approach to facial rejuvenation, which is especially relevant in the context of hyaluronic acid-based treatments such as Profhilo® Structura. Incorporating AI technologies such as ChatGPT-4 into routine practice could not only enhance patient outcomes but also support physicians in making more data-driven decisions. Furthermore, this integration would align well with current advancements in aesthetic medicine, fostering more accurate predictions of treatment outcomes and potentially accelerating the adoption of such technologies within clinical practice. Sincerely, Luciano Lanfranchi, MDPlastic SurgeonDepartment of Plastic and Reconstructive SurgeryCentro Diagnostico ItalianoMilan, Italy