Concentriq Embeddings

Concentriq Embeddings, developed by Proscia, is a foundation model API designed to accelerate AI development in computational pathology for life sciences organizations. By connecting a suite of leading pathology and vision foundation models directly to the Concentriq enterprise pathology platform, it enables data science and AI teams to rapidly build, iterate, and refine high-performing AI algorithms at scale. The solution is purpose-built for biopharma and life sciences R&D teams seeking to advance drug discovery, biomarker identification, safety endpoint assessments, clinical trial stratification, and precision medicine breakthroughs.

Foundation models transform complex whole slide images into embeddings — numerical representations of an image's essential features — that serve as an ideal starting point for building downstream AI models. Concentriq Embeddings makes this powerful technology accessible by eliminating the infrastructure and pipeline overhead typically associated with managing foundation models, allowing data scientists to focus on high-value AI development work.

Key Capabilities

• Seamless integration of multiple leading pathology and vision foundation models directly within the Concentriq platform where enterprise pathology data is stored, enriched, and analyzed.
• Support for rapidly building, testing, and swapping foundation models to find the best fit for specific datasets and downstream applications.
• Ensemble approaches that combine features from multiple foundation models to improve downstream model performance.
• Elimination of manual data transfers, external processing, pipeline maintenance, and image format standardization, resulting in significant computational and cost savings.
• Ability to build AI models at scale — demonstrated by building 80 breast cancer biomarker prediction models in under 24 hours, 13x faster than traditional approaches.
• Reduction of AI development time from weeks to hours, enabling faster iteration across the full R&D lifecycle from target identification to clinical trial optimization.

Integrated Foundation Models

• H-optimus-0: A 1.1 billion parameter vision transformer trained on more than 500,000 H&E stained whole slide histology images.
• PLIP: A visual-language foundation model for pathology image analysis leveraging medical Twitter data.
• Virchow: A self-supervised vision transformer pretrained on 1.5 million whole slide histopathology images.
• DINOv2: A vision transformer model pre-trained on a large collection of images in a self-supervised fashion.
• CTransPath: The first transformer-based unsupervised feature extractor for histopathological images.
• ConvNext: A pure convolutional model inspired by the design of vision transformer architectures.
• KEEP: A vision-language model linking pathology images with disease knowledge for zero-shot cancer insights.

Workflow and Platform Integration

• Concentriq centralizes pathology data and R&D workflows, creating an enriched, accessible data foundation for AI development that can be augmented with Proscia's high-quality real-world data.
• Proprietary AI models built with Concentriq Embeddings can be incorporated directly into image analysis workflows on the Concentriq platform to accelerate the R&D lifecycle.
• Data generated through everyday workflows on Concentriq can be used to continuously iterate and improve AI models over time.
• The open-source Proscia AI Toolkit provides a Python client for API use, Jupyter Notebook tutorials, and a library of Python helper functions for tasks such as image tiling and organizing API outputs.

Deployment and Commercial Reach

• Proscia supports the pathway to regulatory approval for precision therapy AI-based biomarkers and diagnostics through a platform with regulated components already on the market for diagnostic use.
• Organizations can leverage Proscia's global network of diagnostic labs and health systems to access key markets and expand commercial reach for AI-driven diagnostics and therapies.
• The platform is designed to support late-stage clinical trial pathology workflows, bridging R&D and commercial deployment.