About

EverSlidePath

An open educational pathology platform built for the next generation of pathologists.

Mission

Pathology education has a tooling problem. Medical students spend weeks rotating through glass slide libraries, residents rely on static image atlases, and most digital pathology software is locked behind institutional licenses. EverSlidePath is an attempt to close that gap — a free, open, browser-based whole-slide viewer with AI detection built in from the start.

The goal is simple: let anyone with a browser open a real pathology case, annotate it, run AI-assisted analysis, and learn — without creating an account, installing software, or waiting for IT approval.

The Platform

EverSlidePath is built around four tightly integrated components:

WSI Viewer

OpenSeadragon-powered tile renderer with smooth pan/zoom, multi-resolution pyramid support, and a fullscreen study mode.

AI Inference

TensorFlow.js StarDist model for nuclear instance segmentation — runs fully client-side in the browser, no server required.

Annotation System

Shape drawing, per-slide tissue labels, adjustable thresholds, and per-annotation metadata for structured learning.

Cloud Sync

Annotations are auto-saved to Cloudflare D1 with a debounced sync loop. Slide metadata is indexed via D1 for fast lookups.

Technology

The stack is chosen for performance at the edge and zero cold-start latency in the browser:

  • OpenSeadragonBattle-tested whole-slide image viewer with DZI tile support and smooth interpolation.
  • TensorFlow.js + StarDistClient-side nuclear instance segmentation. The model is loaded once and cached in the browser.
  • Cloudflare Workers + D1Edge-deployed API with SQLite (D1) for annotation persistence and slide metadata indexing.
  • TanStack Start + React 19File-based routing, type-safe server functions, and RSC-ready architecture.
  • Tailwind CSSUtility-first styling with a custom dark navy/cyan design language.
Get In Touch

EverSlidePath is open source. Feedback, bug reports, and pull requests are welcome via GitHub.

View on GitHub