Boundary U-Net
Predicts background, nuclear interior and boundary information in evaluation mode.
Private analysis workspace
Upload once. Analyze. Inspect. Download.
Use BioNuclei with an account or as a guest. The released workflow performs Boundary U-Net segmentation, instance measurements and evidence review. Uploads are transient and the analysis job can be deleted immediately after the report is received.
Choose one access mode
Account or guest
Create an account, sign in to an existing account, or continue without registration. These modes are mutually exclusive on this screen.
New account
Create your BioNuclei account
Supabase Auth handles the password securely. BioNuclei does not receive or store your password.
Existing account
Sign in
Only the authenticated session is used to authorize your analysis job.
Guest
Continue without an account
Guest mode creates only an ephemeral job token in this browser session. No account is created.
Analysis pipeline
From image to inspectable evidence
The current public workflow runs a validated nuclei segmentation pipeline, deterministic instance separation, per-nucleus measurements, QC and evidence-constrained specialist review.
Individual nuclei
Converts semantic predictions into explicit nuclear instances for quantitative analysis.
Measurements + QC
Reports morphology, fluorescence intensity, input quality, uncertainty context and provenance.
Upload microscopy image
TIFF or Nikon ND2. The selected file remains in browser memory until it is sent. This page does not write image data to IndexedDB or local storage.
Interactive evidence review
Inspect the detected nuclei
Review the segmentation visually, explore measured objects, color the nuclei by a quantitative variable, and draw a temporary ROI. The ROI is calculated in your browser from the returned measurements and is not saved as a separate server artifact.
NucleiNot available
InputNot available
ModulesNot available
Analyzed overlay
Color detected nuclei by:
Draw a rectangle on the viewer to inspect a spatial subset.
Segmentation mask
The mask contains explicit instance labels used for downstream measurements.
Object measurements
Per-nucleus measurement table
Waiting for measurement data.
| ID | Area | Circularity | Eccentricity | Mean intensity | Centroid |
|---|
Evidence constrained
Expert analysis layer
Specialist agents review measured evidence and the scientific report. They do not change model weights during a user analysis and do not invent unsupported biological conclusions.
Data lifecycle
Designed for disposable analysis
Account mode: Supabase Auth manages the credential. The analyzer uses the authenticated session to authorize the current job.
Guest mode: no account is created. A high entropy job token is held in browser memory for the current session.
Image: the original upload is deleted by the backend after processing. Results expire after a short retention window or can be deleted immediately.
Interactive review: measurements and ROI summaries are rendered in the browser from the returned analysis artifacts. They are not silently added to model training.
Research training: the released Boundary U-Net weights are not updated during user analysis.