Deep-vision IHC engine · live on Railway

See the stain.
Not the noise.

ImageSL analyzes heavy .tif immunohistochemistry slides pixel-by-pixel — automatically separating true chromogenic staining from background for any color, quantifying it, and letting you re-render the image exactly how you want to see it.

Free desktop client · your slides are analyzed securely on our servers.

Built for pathology

A microscope for your microscope

Everything ImageSL does is grounded in real digital-pathology math — then made effortless with AI.

True stain separation

Color deconvolution with automatic Macenko stain-vector estimation isolates the chromogen from the counterstain and glass — for brown DAB or any custom color.

Deep vision reasoning

A Claude vision model inspects each slide, identifies the staining, judges tissue quality, and tunes the analysis — reasoning layered on top of exact pixel math.

Customizable renders

Generate multiple versions on demand — recolor the background to any code, and make the target staining darker or lighter with a slider.

Built-in AI assistant

Ask the ImageSL Assistant anything about your results — it knows your slide's positive-area percentage, thresholds, and optical density, and explains them in plain language.

Heavy .tif, no sweat

Pyramidal and multi-page TIFFs are decoded server-side and intelligently downsampled, so multi-gigabyte slides analyze in seconds.

Source stays server-side

The desktop app is a thin, signed shell. All proprietary logic and keys run remotely — nothing sensitive ever ships inside the download.

Workflow

From slide to answer in four steps

Load your slide

Drop a .tif, .tiff, PNG, or JPEG into the app or the web console.

AI reads it

Vision reasoning identifies the stain and the engine separates it pixel-by-pixel.

Quantify

Get positive-area %, optical density, and an overlay of exactly what was counted.

Render & discuss

Recolor, brighten or darken the stain, and ask the assistant to interpret.

Your logic, protected by design

ImageSL uses a thin-client architecture. The Windows app you download is a small native shell — it uploads a slide and displays results. Every algorithm and every API key lives on the backend.

  • No analysis source code ships in the executable.
  • Deconvolution, AI, and secrets run only on Railway.
  • Access is gated by per-user license keys.
ImageSL.exe (thin shell)
  │ upload slide + license key
  ▼
Railway backend
  ├─ IHC deconvolution engine
  ├─ Claude vision + chat
  └─ your API keys 🔒
  │ results + renders
  ▼
ImageSL.exe (display)