ABS 023: Mitigating the Effects of Skin Cancer Through Artificial Intelligence

Liyana Ahmed ¹

¹ University of Florida

The Van Wickle Journal (2026) Volume 2, ABS023

Introduction: Skin cancer, one of the most common forms of cancer, has become increasingly more common within recent years. With the rise in use of artificial intelligence (AI), researchers have developed AI algorithms, which are more convenient for many patients, to help mitigate this danger. Most commonly, numerous images of cancerous or precancerous skin are uploaded onto a database to train the algorithms, then these algorithms are used to detect abnormalities on patients’ skin.

Methods: The purpose of this investigation is to evaluate various AI tools in their effectiveness of reducing the impacts of skin cancer by detecting it at an early or precancerous stage, as well as how this differs across skin colors. To effectively accomplish this, the PubMed database was searched with keywords such as “AI detecting skin cancer,” “AI detecting precancerous lesions,” and “Skin color in AI detection.” Data from these studies was compiled, addressing the accuracy and limitations of AI detecting skin conditions as well as the existing disparity among skin colors. Data was collected from studies evaluating accuracy in both existing skin cancer and precancerous lesions.

Results: AI is generally efficient as there are multiple examples of AI performing similar to, or in one case better than, physicians trained in the field. However, some algorithms are not trained on sufficiently diverse databases and therefore are not necessarily effective for all populations. Meanwhile, some algorithms perform inadequately as compared to physicians and thus cannot be trusted in a clinical setting. Further evidence reveals that many AI tools accurately detect non-cancerous lesions, which may be treated before progressing into cancer. However, some tools work significantly better for lighter-skinned patients, indicating the importance of improving equality in AI-based skin detection.

Discussion: AI has great potential to be used in a clinical setting due to the accuracy currently seen, but still needs large improvements before this technological leap is made. While many AI tools are equally or more accurate than physicians, many still lack accuracy to be used clinically. As there is a pressing need for increased representation of darker skin in training these algorithms, future algorithms may focus on bridging this gap and increasing accuracy among a more diverse population. These may be beneficial supplementary tools for the future when better developed, rather than replacements for physician expertise.