PuzzleVision
Next-Gen Pathology Platform
Unified workspace, flexible plugin ecosystem. Seamlessly connecting cloud, hospital, and local — reshaping pathology workflows.
PuzzleVision is a digital pathology image viewing platform centered on intelligence, all-scenario coverage, and high security. It serves pathologists, medical researchers, and healthcare administrators by addressing core pain points in pathology image viewing, annotation, collaboration, and deployment.
The system adopts a modular platform design, providing AI-assisted diagnostic interfaces and flexible integration with multiple annotation tools to support efficient pathology image annotation and management. It also supports multi-terminal access and three deployment modes — cloud, hospital/on-premises, and local — to meet the needs of efficient clinical workflows and flexible research collaboration while ensuring data security and ease of use.
Unified Slide-Review Experience
Connects to the Pathology Information System (PIS) or imports local data, adapts flexibly to different work scenarios, and provides an advanced interactive experience for diagnosis and annotation.
Extensible Plugin Ecosystem
Modular access to the AI ecosystem, covering multi-parameter prediction with small models, in-slide analysis with multimodal foundation models, interactive region segmentation, protein/spatial transcriptomics prediction, agent-based pathology report generation, and more.
Multi-Scenario Deployment
Supports multi-terminal access and multi-platform deployment, offering cloud, hospital, and local modes to meet different data privacy and compliance requirements.
Core Highlights
End-to-end digital pathology analysis capabilities that empower precision medicine.
Synergistic Diagnostic Analysis with Large and Small Models
Foundation models capture slide-level context and perform risk stratification, while domain-tuned models focus on lesion boundaries and biomarker interpretation; together they generate explainable, reviewable reference results for quality control and longitudinal follow-up.
Adaptive Scanning Assistant
Multi-scale analysis powered by the Triplet foundation model.
Agent-Based Report Generation
Specialized agents link slide review, evidence fusion, and quality-control checks into structured narratives and pathology highlights, then export review-ready, editable report drafts in one step.
Multi-agent Evidence-based Analysis
Human-AI collaboration enhances diagnostic report generation.
Low-level Vision Prediction
Pixel- and region-level modeling of tissue texture, boundaries, and staining response supports front-end segmentation, weak-stain enhancement, and pseudo-color mapping. It provides a stable, aligned visual substrate for downstream semantic reasoning and quantitative analysis.
Interactive Precision Reasoning
Multiple tools, from freehand drawing to magic-wand auto-contouring, enhance the annotation interaction experience.
App Marketplace
On-demand plugin subscriptions, including AI models, segmentation, report generation, and more.
Structured Annotation
Structured annotation: label selection, template filling, and category management
AI Segmentation
One-click precise tissue segmentation with multi-model support
Liver Cancer Grading / mcp
SubscribedGrades liver cancer into well-differentiated (G1), moderately differentiated (G2), poorly differentiated (G3), and undifferentiated (G4) levels.
Breast Histology Dx / mcp
SubscribedClassifies breast tissue by histological features, including invasive carcinoma, DCIS, LCIS, and more.
Colorectal Cancer Tumor Detection / mcp
SubscribedDetects the presence of colorectal cancer tumor tissue.
Gastric Cancer Grading / mcp
SubscribedGrade gastric cancer into G1–G4 differentiation levels
Three Deployment Modes
Flexibly adapts to diverse data security and usage needs, with seamless switching between the three modes.
Cloud Mode
Zero deployment and ready to use. Data is encrypted in the cloud and accessible across regions, making it ideal for research teams and cross-regional collaboration.
Hospital Mode
All raw data remains on premises; only authorized, de-identified metadata is synced to the cloud, meeting the security and compliance requirements of large hospitals.
Local Mode
Standalone offline viewing and annotation, with one-click cloud sync when online; ideal for mobile consultations and portable research.
Technology and Data Architecture
Demonstrates system robustness for technical decision-makers.
Smart Tile Gateway
Securely retrieves JPEG tiles through the cloud gateway, enabling full front-end/back-end decoupling and fine-grained permission control.
Flexible Data Node Storage
A single node can seamlessly mount multiple data stores, including NAS, local disks, and cloud object storage.
Strict Permission Isolation
The workspace serves as the minimum isolation unit; cases, WSIs, and annotations are all bound to a workspace, ensuring clear security boundaries.
Custom AI Plugin Development & Deployment
Models built through UnPuzzle and similar workflows can connect to standardized APIs for one-click deployment and invocation. It shortens the path from experiment to production and accelerates the translation of research outputs into platform-based applications.
FAQ
Get Started with PuzzleVision
Whether you need cloud collaboration, on-premises compliance, or offline portability — we have you covered.