AI Diagnoses Brain Tumors in Minutes Rather Than Weeks

Why It Matters

By slashing turnaround time and reducing reliance on expensive molecular assays, Hetairos can accelerate treatment decisions and broaden access to precision neuro‑oncology, especially in resource‑constrained hospitals.

Key Takeaways

• Hetairos classifies 102 CNS tumor subtypes from standard H&E slides.
• Diagnostic accuracy reaches 68% overall, 84% within top three predictions.
• Turnaround time drops from weeks to minutes, under two days total.
• High‑confidence predictions achieve ~87% accuracy, cutting need for costly assays.
• AI aids resource‑limited settings by using existing slides, reducing $300‑$400 test costs.

Pulse Analysis

The diagnosis of brain tumors has long hinged on DNA methylation profiling, a gold‑standard test that maps epigenetic signatures to WHO tumor categories. While highly accurate, the assay requires specialized labs, sophisticated equipment, and weeks of processing, driving costs into the several hundred‑euro range (roughly $330‑$400 per case). These barriers limit adoption in many hospitals, delaying critical therapeutic choices for patients with aggressive CNS malignancies.

Hetairos sidesteps these constraints by extracting molecular information directly from routine H&E‑stained sections, the same slides pathologists already generate. Leveraging a deep‑learning architecture trained on a globally sourced dataset of 11,000 digitized slides, the system distinguishes 102 tumor subtypes with an overall accuracy of 68% and up to 87% confidence on high‑certainty cases. When the top three predictions are considered, concordance climbs to 84%, far surpassing the roughly 30% accuracy achieved by expert neuropathologists reviewing only histology. The AI delivers its analysis in about twelve minutes after slide digitization, enabling a full diagnostic workflow in under two days—a dramatic acceleration that can reshape treatment planning.

Beyond speed, Hetairos promises a democratizing effect on precision oncology. By relying on existing histology slides, it eliminates the need for costly methylation assays, making molecular‑level classification feasible in low‑resource environments. Hospitals can prioritize cases for full genomic work‑up, optimize tissue usage, and reduce overall diagnostic expenditures. As the model continues to ingest larger, more diverse datasets, its performance on rare tumor entities is expected to improve, further solidifying AI’s role as a decision‑support tool rather than a replacement for human expertise. The commercial potential is significant, with digital pathology platforms poised to integrate Hetairos‑type solutions, driving a new wave of AI‑enabled, cost‑effective cancer care.