Overview

• CrossNN analyzes tumor epigenetic signatures to differentiate more than 170 cancer types without relying on invasive tissue sampling.
• Training on more than 8,000 reference tumors and testing on over 5,000 samples yielded 97.8 percent accuracy overall.
• A streamlined neural network design makes the AI’s predictions explainable and traceable for clinical use.
• Compatibility with liquid biopsies, including cerebrospinal fluid analyzed by nanopore sequencing, enables noninvasive tumor diagnosis.
• Upcoming clinical trials at all eight German Cancer Consortium centers will assess CrossNN’s routine and intraoperative applications.