A Real-Time platform-agnostic Pathology Companion.

Bring AI Insights Directly To Your Workflow.

OnSight is a standalone, vendor-agnostic software that bridges the gap between powerful AI models and the daily workflow of pathologists, offering real-time analysis without compromising data privacy.

Download (GPU) Download (CPU)

The Adoption Barrier

Artificial intelligence holds immense potential to enhance objectivity and efficiency in histopathology. Yet, its integration into clinical practice is critically stalled. Pathologists face a fragmented landscape of proprietary WSI formats, incompatible viewers, and data security hurdles that make deploying AI tools a complex, expensive, and often prohibitive endeavor. The best AI model is useless if it can't be used.

A Radically Simple Approach

Vendor-Agnostic

Works with any slide viewer on any OS. If it's on your screen, OnSight can analyze it. No APIs, no plugins needed.

100% Secure & Private

All processing happens locally on your machine. Patient data and proprietary images are never transmitted or uploaded.

Real-Time Feedback

Leverages local GPU/CPU power to deliver live AI inferences as you navigate slides, enabling an interactive workflow.

The OnSight Workflow

Pathologist viewing a slide on a computer. — The Onsight companion offers a solution by enabling real-time, local AI inference directly from the screen capture of any standard WSI viewer. The OnSight workflow is initiated when the user captures a Region of Interest within any standard slide viewer. The raw image data from this ROI is then processed by OnSight's AI algorithms, and the real-time output is displayed as an overlay on the graphical user interface. The inference results overload in real time and the software runs locally to ensure privacy.

See It In Action

Example Use Cases

OnSight comes with a suite of pre-trained models to assist with common histopathological tasks. Users can easily switch between models to fit their needs.

Brain Tumor Classification

Utilizes a powerful Vision Transformer (ViT) to provide real-time classification of common brain tumor patterns (glial, meningiothelial, etc.) on H&E slides, guiding initial diagnosis.

Automated Mitosis Counting

Objectify tumor grading by automatically detecting and counting mitotic figures in real-time. Calibrate results to physical area (mm²) for reproducible, report-ready metrics.

IHC Quantification

A YOLO-based model segments and counts Ki-67 positive and negative nuclei, providing a rapid and objective proliferation index. Includes controls for overlay transparency to aid interpretation.

Interactive AI Chat

Integrates a Vision-Language Model (HuaTuo) that supports natural-language queries about the current image. Ask questions like, "What feature makes this glial?" to get deeper, interpretable insights from the AI.

The Team

OnSight is proudly developed by the Diamandis Lab, a multidisciplinary team of pathologists, scientists, and engineers.

Meet The Researchers

Common Questions

What are the system requirements?

For optimal real-time performance, a Windows PC with a dedicated NVIDIA GPU (RTX 20-series or newer) is recommended. The software remains functional on CPU-only systems with a slight delay in inference.

Can I use my own AI models?

The current version is distributed with the pre-packaged, validated models. We are actively working on a framework to allow users to easily integrate their own custom models in future releases.

What is the current status of OnSight regarding clinical diagnostic use?

OnSight is an experimental and research tool that is intended to support future clinical applications but has not yet been validated for clinical diagnostic use. At this stage, it is provided strictly for research, educational, and investigational purposes.

Ready to Augment Your Workflow?

Download the latest version of OnSight for Windows. It’s free, secure, and ready to use in minutes.

Download (GPU) Download (CPU)