The rise of artificial intelligence in dermatology: Perspectives and potentials

DERMATOLOGISTS PLAY A SUBSTANTIAL ROLE IN MEDICINE The skin, as the body's largest immune organ, provides a visible impression of one's health status at first glance. The appearance of the skin could reflect internal medical problems, including diseases related to the kidneys, liver, endocrine system, thyroid, pancreas, and immune system. In addition, skin rash is the most common adverse drug reaction, and warned clinical physicians are reminded to promptly identify and discontinue culprit drugs. Furthermore, by observing the condition of the skin, we can gain insight into a person's emotional state. Factors such as their eating habits, sleep patterns, and romantic relationships can be discerned to some extent through the appearance of their skin. Dermatology, being an inherently visual specialty, is well suited for AI-driven innovations. We acknowledge the vital role dermatologists play in medicine and recognize the challenges we face amid the rise of artificial intelligence (AI).1-3 What perspectives and opportunities emerge for dermatologists in this evolving landscape? ARTIFICIAL INTELLIGENCE-ASSISTED DIAGNOSIS OF SKIN CANCERS AND OTHER DISEASES AI research has made significant strides in diagnosing and treating skin cancers, acne, rosacea, onychomycosis, psoriasis, vitiligo, atopic dermatitis, eczema, and others.4-6 AI algorithms demonstrate remarkable improvement in classifying clinical, dermoscopic, or optical coherence tomography7 images of skin cancers.8-10 During the COVID-19 pandemic, the integration of telemedicine and AI within hospital information systems offers dermatologists a platform to combating the novel virus and addressing academic, regional, and global health-care needs.11 Various popular image-based applications are available on both the iPhone and Android App stores, allowing patients to upload photos for analysis. Patient-directed AI platforms facilitate the monitoring of benign skin lesions, promoting beneficence, and mitigating maleficence associated with delayed diagnoses.12 Large language models enable patients and health-care workers to interact with AI models using natural language. Notable examples include ChatGPT, Bing Chat, and Google Bard.2,13 By leveraging vision-language models capable of processing both text and image inputs, dermatology can achieve more effective and personalized care. This innovation has the potential to reduce burnout and stress among dermatologists, as it streamlines the diagnostic and treatment process.13 CURRENT VENTURE CAPITAL PLATFORMS IN DERMATOLOGY FOR ARTIFICIAL INTELLIGENCE A wide range of potential applications of AI in dermatology include image-based diagnostic devices (VisualDx and Piction Health) and devices equipped with real-time imaging capabilities (3Derm, Anapix Medical, and Dermasensor). Other examples include image-based direct-to-patient triaging tools (Triage and Eskindoctor), clinical workflow tools and decision support systems (Aiklu and Magnosco), clinical analytics platforms (BoomerangFX and Qubole), AI-enabled skincare recommendation services directly to patients (Daisies, Exely, and IQONIC.AI), and telemedicine services enhanced by AI (SolDoc and Yuko AI).3,14 LIMITATIONS OF ARTIFICIAL INTELLIGENCE IN DERMATOLOGY Many skin diseases still cannot be diagnosed at first glance using images. Examples include diverse drug eruptions, viral exanthems, vasculitis, vasculopathies, and others. Moreover, the practical effectiveness of AI-assisted diagnosis in clinical settings remains uncertain as most studies evaluate accuracy in restricted settings rather than real-world scenarios.8,15 The sensitivity and specificity of AI are not 100%, and even dermatopathology, considered the gold standard for dermatologic diagnosis, lacks perfect interobserver consistency despite advanced imaging techniques.16 Certain dermatology AI models underperform with dark skin tones and uncommon diseases.17 Recommendations by Phung et al. address crucial aspects such as patient consent, privacy, image acquisition, labeling, curation, and storage, aiming to enhance transparency, explainability, training datasets, and accuracy.18 Acknowledging these limitations is essential for AI's integration into clinical practice. ETHICAL CONCERNS OF ARTIFICIAL INTELLIGENCE IN DERMATOLOGY AI technologies in dermatology are still imperfect and can lead to misdiagnoses of lesions, potentially causing harm to patients.12,19 Patients may receive AI-generated results before clinicians may lead to potential confusion and unwarranted anxiety. Inaccurate diagnoses could trigger unnecessary angst and psychological maleficence.12,16,17 Patients getting wrong diagnoses from these apps might ask for biopsies, leading to more unnecessary procedures, higher health-care costs, and patient maleficence.12 Furthermore, AI systems are particularly less effective for patients with darker skin tones and those with rare diseases. It is important to ensure that AI does not unintentionally reinforce preexisting racial or cultural biases in global health.17 AI techniques diagnose benign or malignant lesions without physician participation, and AI data collection, sorting, and interpretation can be challenging. Part of the results of AI interpretations were not convincing to clinicians, and they hesitated to use these technologies. Regardless of whether AI techniques have been used, patients should be informed to see a dermatologist if they notice any lesion that bleeds, does not heal, changes color, becomes itchy or painful, or shows signs of evolution, as these signs could sign a higher risk of malignancy. The emergence of claims about AI's capabilities in dermatology makes it challenging to differentiate between hype and reality.20 It is important to realize the limitations and ethical issues of AI in clinical practice, and future research should prioritize investigations in real-world settings, emphasizing AI-assisted approaches in dermatology.9,16 Data availability statement Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study. Financial support and sponsorship Nil. Conflicts of interest There are no conflicts of interest.