Proscia: AI-Powered Biomarker Analysis in Concentriq
From the Proscia Blog:
Biomarker interpretation is high-volume and variable. Whether you’re harmonizing scoring across teams or reducing manual effort, consistency at scale matters.
Concentriq® is increasingly helping users standardize quantification with AI algorithms for several of the highest-value biomarkers. Powered by Mindpeak, these zero-click applications* surface results within existing review workflows to help reduce analysis time. Each application is designed for speed, with outputs aligned to established scoring frameworks.
Current biomarker applications available in Concentriq include:
PD-L1 (Gastric, NSCLC, ESCC, Urothelial Cancer): Mindpeak’s PD-L1 algorithms detect and classify tumor and immune cells to generate CPS and TPS outputs. Immune cell handling is configured by application to support consistent quantification across heterogeneous samples.
HER2 (Breast Cancer: HER2-Positive, HER2-Low and Ultralow Research): Mindpeak HER2 detects and classifies HER2-stained tumor cells, distinguishing negative, partial, weak complete, and strong complete membrane staining using commonly referenced scoring frameworks (including ASCO/CAP criteria). It also supports assessment of HER2-low and ultralow categories, where subtle staining differences can affect classification and downstream analysis.
Ki-67 (Breast Cancer): Mindpeak Ki-67 uses a global scoring workflow to automatically detect and classify Ki-67–stained tumor cells across up to four regions of interest. It produces both unweighted and weighted global scores to support consistent proliferation quantification.
ER/PR (Breast Cancer): Mindpeak ER/PR identifies and classifies ER- and PR-stained tumor cells within whole slide images or predefined regions of interest. The AI then calculates a positivity score to support consistent quantification.
Proven and Trusted
These applications have been evaluated across diverse, real-world datasets and validated in several peer-reviewed studies.
A recent Nature article reported strong performance of Mindpeak’s gastric PD-L1 AI, highlighting its consistency across heterogeneous samples.
A Modern Pathology study validated Mindpeak Ki-67 and ER/PR across multiple institutions, showing that:
- AI-assisted scoring was noninferior to manual scoring
- Pathologists achieved higher inter- and intra-observer agreement using AI
- Algorithms performed robustly across varied scanners, stains and protocols
Further comparative analysis demonstrated that Mindpeak’s HER2 AI delivered the highest accuracy of all evaluated models for HER2-low, achieving a balanced clinical score of 88.1% — surpassing pathologist concordance (85.5%).
At ASCO 2025, Mindpeak and AstraZeneca presented results from a HER2 ultralow AI validation study involving more than 2,000 samples from 15 institutions, reinforcing the algorithm’s potential to bring new clarity to one of the most challenging categories in breast cancer pathology.
AI at the Core of Concentriq
Proscia knows that scaling AI depends on efficient, centralized workflows where markups, overlays, and endpoints are instantly accessible.
With more than 120 applications seamlessly embedded into Concentriq—and more continuously introduced—our platform gives laboratories and biopharma organizations unmatched breadth and depth to leverage AI across the full spectrum of their work and scale with confidence.
This is just one of the many reasons why some of the most advanced laboratories, 16 of the top 20 pharmaceutical companies, and major CROs rely on the platform every day.
Learn more about Mindpeak’s AI-driven biomarker analysis and all AI applications, available in Concentriq here.
*These applications are not intended for diagnostic use with Concentriq. Contact the original manufacturer for indications in your region.
**Concentriq AP is for Research Use Only. Not for use in diagnostic procedures. Proscia’s AI applications are available for research use only.
SOURCE: Proscia
