The revolutionary impact of AI in dermatopathology: enhancing accuracy or overstepping boundaries

Artificial intelligence (AI) is transforming the field of dermatopathology, presenting both opportunities and challenges. This comprehensive review examines the integration of AI, particularly deep learning algorithms, into dermatopathology and its implications for disease diagnosis, research, and the future of clinical practice.

AI, especially through convolutional neural networks (CNNs), has shown exceptional promise in dermatopathology. CNNs are capable of processing complex image data, breaking it down into pixels, and making accurate diagnostic predictions. Studies have demonstrated CNNs’ ability to achieve diagnostic accuracy comparable to, or even surpassing, that of human dermatopathologists. For instance, AI systems have achieved 99.3% accuracy in diagnosing dermal nevus and 100% in identifying seborrheic keratoses. One of the significant applications of AI in dermatopathology is the diagnosis of melanoma. Advanced deep learning algorithms have outperformed pathologists in distinguishing between benign and malignant melanocytic lesions. In a notable study, the diagnostic accuracy of CNNs for melanoma was significantly higher than that of experienced dermatopathologists.

AI's potential extends beyond melanoma to non-melanoma skin cancers and other dermatological conditions like psoriasis and atopic dermatitis. However, the diagnostic complexity of non-melanoma skin cancers poses a greater challenge due to the diverse range of benign and malignant conditions. Current AI systems require extensive training and high-quality data inputs to achieve reliable accuracy in these cases. Despite its promising capabilities, AI in dermatopathology is not without limitations. The classification system remains narrow, restricting AI’s ability to accurately diagnose a broad spectrum of dermatological conditions. Additionally, there is significant variability among pathologists, complicating the training process for AI systems. The limited availability of comprehensive image databases and the challenges of data sharing further hinder the development and implementation of robust AI algorithms.