PRODUCT OVERVIEW
DovaVision UC
An introduction to DovaVision UC for GI researchers
Introduction
DovaVision UC is an AI-powered software solution designed to enhance the evaluation of Ulcerative Colitis (UC) from colonoscopy videos. It delivers frame-level, quantitative assessments and standardized scoring outputs aligned with the Mayo Endoscopic Score (MES).
The system provides validated, objective metrics that support:
- Consistent disease severity assessment for monitoring and treatment surveillance
- Quantitative endpoints for clinical trials and research
- Scalable automation for reducing variability across readers and sites
Metrics Included
- Frame-Level Disease Severity Score: 5–10 second frame-level scoring across the entire procedure, enabling precise tracking of disease activity.
- Frame-Level Sub-Score Components: Drill-down detail on key disease features (vascular pattern, bleeding, erosions, biopsy process, and non-scorable segments).
- Video-Level MES Score: AI-generated MES score for the full procedure, benchmarked against human reader consensus.
- Video-Level Interpretability: Transparency layer providing interpretability of the MES score for the entire video.
How is the MES score generated?
The MES score is derived from DovaVision UC’s proprietary AI models trained on thousands of expertly annotated colonoscopy frames. Each frame is scored in near real-time, producing objective, quantitative, and reproducible metrics that eliminate subjectivity common in human reads.
The system applies deep learning algorithms to assess disease features frame by frame, then aggregates results to produce an overall MES score with full interpretability. By generating structured outputs in both visual outputs and JSON data formats, DovaVision UC supports integration into clinical workflows, EHR systems, and clinical trial platforms.
Clinical Relevance
DovaVision UC is purpose-built to solve the critical challenge of variability in endoscopic scoring. By standardizing UC assessment, it delivers consistent, reproducible results across clinicians and sites. Unlike traditional Mayo Endoscopic Score (MES) readings, DovaVision UC enriches outputs with granular, frame-level data that reveal subtle disease nuances often hidden within the broader MES categories. This enriched data stream enables multi-site analysis, integration with multi-omic datasets, and longitudinal tracking over time—unlocking a deeper understanding of disease activity.
Through this standardized, data-rich approach, DovaVision UC:
- Reduces inter-observer variability
- Improves diagnostic confidence
- Provides reproducible endpoints for clinical studies
- Supports precision treatment decisions and monitoring
Evidence & Validation
Multiple studies are underway to validate the use of DovaVision UC as a regulatory-ready AI tool for clinical practice and research. Early results demonstrate strong alignment with expert consensus reads, while offering superior scalability and consistency across diverse clinical environments.
Evidence & Validation
Built on 108 million expertly annotated labels, DovaVision UC delivers industry-leading performance with 0.867 Quadratic Weighted Kappa (QWK) and 0.869 Intraclass Correlation Coefficient (ICC) for video-level Mayo Endoscopic Score (MES).
Embedded Quality Control
Automated checks on ingestion flag anomalies such as artifacts, lighting issues, and unscorable segments, ensuring only high-quality, consistent data proceeds to analysis.
Secure and Scalable Analysis via API
Robust authentication and authorization safeguards protect patient data while enabling seamless integration into your existing workflows. Designed for scalability, our API delivers high-throughput, reliable analysis.
Optimized for Modern Video Standards
DovaVision UC supports video inputs at a minimum of 720p resolution and 8 Mbps bitrate, to support granular AI-powered evaluation.