United by Unique: Advancing Personalised Cancer Care on World Cancer Day

We were proud to present our collaborative research with AstraZeneca at the 2025 American Society of Clinical Oncology (ASCO) Annual Meeting. The study, focusing on AI-assisted training for HER2 IHC interpretation, was delivered by Marina De Brot on Friday, May 30th, in Hall D2, in Front of over 2000 people.

Enhancing Diagnostic Accuracy in HER2-Low and HER2-Ultralow Breast Cancer

The classification of HER2-low—and the emerging HER2-ultralow category - has become increasingly significant with the advent of HER2-targeted antibody-drug conjugates such as Enhertu. These therapies are providing new treatment options for patients with minimal HER2 expression, underscoring the importance of accurate diagnostic interpretation.

However, differentiating IHC 1+ from IHC 0, particularly in cases of faint membrane staining, remains a major diagnostic challenge. In line with the 2025 ASCO Annual Meeting theme introduced by ASCO President Dr. Robin Zon, our work aims to improve diagnostic precision and expand equitable access to personalized cancer treatments.

Global AI-Enabled Training Initiative

To address this challenge, we deployed an AI-supported training program through the Mindpeak GmbH ComPath Academy. The initiative engaged 287 pathologists across 11 countries, providing standardized education and decision support on HER2 IHC scoring.

Impactful Results for Clinical Practice

The study results were both clinically and statistically significant. Notably, the AI-supported training led to a 40% increase in accurate identification of HER2-ultralow cases—potentially expanding access to life-extending targeted therapies for a broader group of patients.

As highlighted in coverage by OncLive following the official ASCO press conference, this research reflects the evolving standards in cancer care. Accurate identification of HER2-low and HER2-ultralow patients is now a critical component of therapeutic decision-making in both metastatic and early-stage breast cancer settings.

Empowering, Not Replacing, Pathologists

This initiative is not about replacing pathologists with AI - it’s about empowering them. By combining expert training with advanced AI tools, we aim to improve diagnostic consistency and ensure that more patients receive the therapies they need.

We extend our thanks to everyone who joined us at ASCO 2025 and look forward to continuing our work toward a more precise, equitable future in cancer diagnostics.

‍

Continuing the Story: AstraZeneca and Mindpeak GmbH present new real-world results on AI-assisted HER2 scoring at ASCO 2025Following our oral presentation last Friday on the AI-assisted training for HER2 IHC interpretation, we are proud to share insights on the Monday’s poster session that took place at this year American Society of Clinical Oncology (ASCO) Annual Meeting.This time, it’s about  real-world diagnostics. In collaboration with six academic centers across the globe, we evaluated whether AI can truly enhance accuracy and consistency in daily pathology practice.And the answer is clear:AI support boosted diagnostic agreement  among pathologists to 86.4% for HER2-low and by 15% to 80.6% for HER2-ultralow - a major leap in consensus for some of the most challenging scoring decisions.Misclassification of HER2-null cases dropped by 64.5%, meaning more patients who may benefit from novel therapies like Enhertu can now be correctly identified.Accuracy rose across the board, especially in the critical 0 vs. 1+ distinction - an area long known for variability and clinical impact.Together, the oral and poster presentations offer a comprehensive view of how AI can support reproducible, high-quality HER2 assessment at scale - from training to clinical practice.As the clinical importance of HER2-low and ultralow scoring continues to grow, our findings provide strong evidence that AI can play a pivotal role in expanding access to targeted therapies like Enhertu - by enabling greater diagnostic precision.

We’re thrilled to share that our poster titled “Large-scale validation of a robust artificial intelligence algorithm for detection of invasive breast cancer on H&E images” has been awarded Best Poster at this year’s ESMO Breast Cancer Annual Congress 2025.

The joint study with Emory University, University Clinic Erlangen and University of Regensburg shows that our AI software Mindpeak Breast H&E can reliably distinguish invasive breast cancer from non-invasive in-situ lesions and healthy tissue with outstanding accuracy.

On one of the most diverse datasets to date, including data from 59 institutions, the algorithm achieved a AUC of 0.997 demonstrating near-perfect classification performance.

Our results show that Mindpeak Breast H&E is ready to enable effective case prioritization for pathologists to achieve faster diagnosis and treatment for patients in need. This capability will help empower pathologists worldwide in maintaining high standards of patient care amid increasing case volumes.

Congratulations to the whole team and contributors: Patrick Frey, Andre Kochanke, PhD, Sunil Badve, Prof. Dr. med. Dr. rer. biol. hum. Ramona Erber, Niklas Abele, Diego Calvopina, PhD, Khalid Daifalla, Ralf Banisch, David Mulder, Tobias Lang

‍

‍

‍

‍