Artificial Intelligence-based Image Analysis for Early Cancer Detection: Review

Abstract

Background: Early detection of melanoma is critical for improving patient outcomes, as late-stage diagnosis is associated with increased mortality. Traditional imaging techniques like reflectance confocal microscopy (RCM), optical coherence tomography (OCT), and dermoscopy require significant expertise, leading to variability in diagnostic accuracy. Artificial intelligence (AI) offers a promising solution to enhance the objectivity and consistency of skin cancer diagnosis.

Methods: A systematic literature review was conducted using PubMed/Medline, Embase, and Cochrane databases, focusing on studies published between 2016 and 2023. The review assessed AI methodologies applied to images of malignant melanoma obtained via RCM, OCT, and dermoscopy. Key search terms included “melanoma,” “neural network,” and “artificial intelligence.”

Results: The analysis revealed a significant advancement in AI-driven techniques for melanoma classification, with deep learning models demonstrating performance equal to or exceeding that of dermatologists. Various studies reported high accuracy rates, with some models achieving an area under the receiver operating characteristic curves (AUC) above 0.90. Notably, methods incorporating diverse datasets showed improved diagnostic reliability across varying skin tones.

Conclusion: AI-based image analysis significantly enhances the early detection of melanoma, offering a robust alternative to traditional diagnostic methods. This review underscores the need for continued research to develop inclusive datasets and refine AI algorithms, ensuring equitable healthcare access and improved patient outcomes.