Innovative AI Model Enhances Tumor Diagnosis with Over 170 Cancer Types Detectable

Recent advancements in cancer diagnostics have introduced a groundbreaking artificial intelligence (AI) model capable of identifying more than 170 different tumor types with exceptional accuracy. Developed by researchers at Charité – Universitätsmedizin Berlin, this model analyzes the tumor's epigenetic fingerprint—specific modifications to the genetic material that influence gene activity without altering the DNA sequence—to classify tumors rapidly and reliably. This approach is particularly valuable in cases where traditional tissue biopsies pose significant risks or are challenging to perform, such as brain tumors where surgery can be hazardous.

The AI model, known as crossNN, utilizes machine learning techniques to compare the epigenetic profiles of unknown tumors against an extensive database of over 8,000 reference profiles. It achieves a classification accuracy of 99.1% for brain tumors and 97.8% across a wide array of tumor types from various organs, surpassing previous diagnostic solutions. Its architecture, based on a simple neural network, ensures transparency and explainability—key factors for clinical application approval.

One of the notable advantages of this technology is its ability to work with data derived from less invasive samples, such as cerebrospinal fluid, which can be obtained through a liquid biopsy. This enables tumor classification without the need for surgical tissue extraction, reducing patient risk and discomfort. For instance, in a recent case, the model facilitated an accurate diagnosis of a central nervous system lymphoma from cerebrospinal fluid, leading to prompt initiation of targeted chemotherapy.

The methodology involves analyzing epigenetic modifications—such as DNA methylation patterns—that act as biological switches, offering a unique and consistent fingerprint for each tumor. These modifications allow differentiation between tumor types, even when the profiles are partial or obtained via different sequencing methods. Experts aim to validate the model further through clinical trials across multiple German cancer centers and explore intraoperative applications.

This innovation represents a significant step toward precision medicine in oncology, enabling more accurate, faster, and less invasive tumor diagnosis, ultimately improving patient outcomes. The research underscores the importance of molecular profiling in modern cancer care and highlights AI’s growing role in transforming diagnostic processes.

Source: https://medicalxpress.com/news/2025-06-tumor-diagnostics-ai-cancer.html