University of Missouri AI Achieves 92% Accuracy in Skin Cancer Detection

Artificial intelligence is demonstrating a remarkable ability to detect skin cancer, with researchers at the University of Missouri achieving 92% accuracy in identifying melanoma using advanced image analysis. The team, led by associate research professor Kamlendra Singh, trained and tested AI models on a database of 400,000 images of skin abnormalities captured via 3D total body photography. This technology evaluates subtle visual patterns – size, shape, color, and density – to pinpoint potentially cancerous spots. “The goal is not for AI to replace doctors and other experts, but AI can help patients with limited access to dermatologists,” Singh said. This research represents a significant step towards expanding healthcare access and improving early detection rates, ultimately promising better health outcomes for patients.

AI Model Ensemble Improves Melanoma Detection Accuracy to 92%

Singh, a principal investigator in the Bond Life Sciences Center, is particularly focused on expanding healthcare access, noting that “It will be some time before this can be used as a tool by doctors in a health care setting, but this research is a promising proof of concept.” Further training with more diverse datasets—incorporating varying skin tones, lighting, and camera angles—is expected to continually refine the AI’s predictive capabilities and build trust among clinicians. “As researchers, if we can get better at explaining why and how AI comes to the conclusions it makes, more health care professionals will trust that it can be a helpful tool,” Singh said.

3D Total Body Photography Enables High-Resolution Skin Abnormality Analysis

The technology employs 3D total body photography, creating a high-resolution, three-dimensional digital map enabling analysis of subtle visual details across the entire body—a significant leap beyond traditional methods. The team, led by Kamlendra Singh, investigated the performance of multiple AI models, discovering a synergistic effect when combining them. Individually, each model achieved up to 88% accuracy in distinguishing melanoma from benign conditions, but a combined approach boosted performance, exceeding 92% accuracy. This isn’t about replacing dermatologists, but providing a powerful decision-support tool, particularly for patients facing limited access to specialist care. The research, published in Biosensors and Bioelectronics: X, underscores Mizzou’s commitment to the intersection of AI, precision medicine, and patient-centered care.