Are dermatologists a well-informed audience for artificial intelligence? A knowledge, attitude, and practice survey

Dear Editor,

Due to its huge clinical, dermoscopic, and dermatopathological picture library, dermatology has taken the lead in the medical field’s artificial intelligence (AI) implementation. Although consumer access to this technology has been made widely possible, there is barely any data regarding dermatologists’ opinions on AI. A recently published report by Janda and Soyer highlighted the various advantages, disadvantages, and nine ethical considerations of using AI in conjunction with a physician’s practice.^[1]

In this survey, we created an anonymous online poll with 17 questions aimed at achieving a better understanding of a dermatologists’ knowledge (K1-K5), attitude (A1-A7), and practice (P1-P6) toward the implementation of AI in the field of dermatology [Table 1]. The correct or incorrect responses were determined based on established knowledge and guidelines in the field of AI as it applies to dermatology. The correct responses were predefined and validated through cognitive interviews, content validity, and pilot testing. The survey was delivered to dermatologists through a number of online means, including emails, WhatsApp, Facebook, and Instagram, and the study took place between July 2023 and October 2023. The total score was graded from 0 to 16 (0–5 for K, 0–6 for A, and 0–5 for P). A 7^th question in attitude asked the respondents about which field, according to them, would flourish the most with AI (clinical, dermatoscopy, or dermatopathology) [Table 1]. To verify the identity of each respondent and prevent multiple responses by the same respondent, their unique registration number and email addresses were used.

The study included 103 dermatologists with 64 post-graduate training students, 26 junior dermatologists (<10 years of specialist experience), and 13 senior dermatologists (≥10 years of experience). We found that only 59 (57.28%) of respondents had a high or excellent understanding of AI in dermatology (score ≥13). In this study, trainees and young dermatologists (P = 0.0001) had a significantly higher knowledge of AI and machine learning (ML) [Table 2]. Only 36 (34.95%) knew the difference between AI and ML. Up to 75 were well informed of the applications of AI in dermatology, and only 16 (15.34%) dermatologists had been provided training for AI as part of their medical curriculum or continuing medical education.

On analysis of attitude score, we observed that 77 (74.76%) and 72 (69.90%) dermatologists believed that AI could transform the field of medicine and dermatology, respectively. In fact, 57 (55.34%) dermatologists even believed that AI could replace human dermatologists in the foreseeable future, and 28 (27.18%) were scared by the prospect of AI in dermatology, with 59 (57.28%) respondents agreeing that AI could increase the rate of misdiagnosis. However, a majority of subjects (75; 72.84%) were optimistic regarding its utilization in the field. We discussed three potential uses of AI in dermatology with our subjects. Dermatoscopic images were thought to have the most potential (48; 46.60%), which was much higher than the potential for dermatopathological (17; 16.50%) or clinical images (6; 5.82%). A favorable attitude was connected with having more knowledge of AI in the field of dermatology (P < 0.0001). Younger dermatologists expressed less trepidation and more excitement about the employment of AI in dermatology and other fields of medicine as a whole.

Overall, the dermatologists participating in this study performed poorly in the domain of practice. Only trainees or young dermatologists used AI-related tools. Just 12 (11.65%) dermatologists (all trainees) used AI software or mobile applications to diagnose skin cancer. In our study, the AI software or mobile applications used by these dermatologists were Skinvision, ChatGPT, Skin Scanner, Skinive, and Cureskin. The apps that the respondents said they were using were; https://skinive.com/, https://cureskin.com/, https://www.skinvision.com/, https://chatgpt.com/, https://ai-derm.com

A recent report by Polesie et al.^[2] found that dermatologists have a positive attitude toward AI, Lim, and Flaherty^[3] suggested that AI should be carefully incorporated into routine clinical practice after dermatologists have received proper training for the usage of AI. According to a recent survey by Pinto Dos Santos et al., most undergraduate medical students are aware of the possible uses and implications of AI in medicine and they are not worried that AI will eventually replace human doctors.^[4] A similar observation was made by us as well, with a majority of subjects being optimistic about its use in dermatology. In a similar survey by Oh et al., 60% of physicians agreed that they had a decent familiarity with AI, and 73.4% considered that technology had useful implications in the medical profession.^[5]

Dermatology’s most desirable AI aim right now is skin tumor diagnosis. However, our findings suggest that Indian dermatologists from the western part of the country are not well aware of this usage. The gap in the correct knowledge with inadequate practices was seen in senior dermatologists with a very poor understanding of AI and ML in most senior doctors. Our study had several limitations. These include small sample size, exclusion of physicians from other medical streams, and significant variation in the number of trainees, junior dermatologists, and senior dermatologists, with a notably low number of senior dermatologists. In addition, differences in clinical experience and familiarity with technological advances among participants may have introduced bias.

The survey’s results imply that although the Indian dermatology fraternity has a positive view of AI, its usage is very limited in the country. Young dermatologists are especially optimistic about its usage and well aware of its applications, but their exposure to the field is limited by the absence of inclusion of this subject in their current curriculum. In addition, most subjects had limited access to AI and had not used it in the field of dermatology. The majority of dermatologists who participated in this survey believe that AI will strengthen our ability to diagnose, but with a considerable possibility of misdiagnosis. A small percentage of respondents, especially senior dermatologists, expressed anxiety about being replaced by AI and felt threatened by it.