Florida Atlantic Researchers Detected Skin Cancer via AI

A 2026 study combined Raman spectroscopy and machine learning to identify skin cancer samples non-invasively.

Updated on Oct. 5, 2026 in Cancer

A close-up of precision laser spectroscopy equipment scanning a medical specimen slide in a clean, clinical laboratory setting. — Florida Atlantic University researchers have developed an AI-driven Raman spectroscopy diagnostic system capable of identifying skin cancer tissue with high diagnostic precision. AI Illustration. Upload story photo >

Live Poll

Would you trust a non-invasive AI scan to diagnose skin lesions instead of a traditional biopsy?

Researchers at Florida Atlantic University developed a system that uses a 785-nanometer diode laser to analyze skin tissue samples. Published in 2026, the study leveraged machine learning classifiers to distinguish between cancerous and normal tissue.

Why it matters

The research aimed to develop rapid, non-invasive diagnostic tools that could help physicians reduce the number of unnecessary biopsies performed on patients. This could streamline the clinical workflow and improve diagnostic precision for common skin malignancies.

The study generated 1,000 Raman spectra from 50 samples of basal cell carcinoma, squamous cell carcinoma, and normal skin. A support vector machine model achieved 78.7% sensitivity and 88.6% specificity in classifying the tissue.

The players

Florida Atlantic University

A public research university located in Boca Raton that serves as the hub for the team behind this diagnostic research.

The details

The team employed a mobile Raman spectroscopy setup to detect stronger protein-related signals in cancerous tissue and higher lipid-related signals in healthy samples. Further analysis was conducted using shallow neural networks, which yielded a receiver operating characteristic area under the curve of 0.910.

In 2024, nearly 1.5 million skin cancer cases were diagnosed globally.
The study results were published in 2026.

The Big Picture

This study follows the precedent for disseminating novel optical diagnostics set by the Proceedings of SPIE. It represents a shift toward leveraging computational modeling to overcome the inherent limitations of traditional visual skin assessments.

This diagnostic approach could eventually provide patients with a non-invasive way to screen for cancer during routine dermatological exams. By increasing the accuracy of initial assessments, the method may limit the need for painful biopsies in healthy tissue.

The takeaway

Advancing machine learning in dermatology offers a path toward more efficient, non-invasive screening for millions of patients. Future iterations of this technology will likely focus on deeper neural network architectures to bridge the gap between bench research and clinical implementation.

What happens next

Researchers plan to conduct larger studies to further optimize their machine-learning models and aim to implement deep neural networks to improve future diagnostic performance.