π¬
Drop your WSI file here
Whole Slide Image scanned at 0.5 MPP Β· CONCH v1.5 448px patches
.svs.ndpi.mrxs
.tiff.scn.vms
βΉοΈ
Demo mode: click βΆ Run Analysis to simulate the pipeline.
For real WSI analysis, deploy the
backend/ FastAPI server and enter its URL above.
Pipeline Execution
β
π§¬
No analysis yet
Upload a WSI and run the pipeline to see results here.
Overview
Spatial
Tissue
Morphology
Fold Detail
πΊοΈ
No heatmap available
Run an analysis first.
β
β
Attention Heatmap
Classifier relevance (LR coefficient projection) overlaid on tissue
LowHigh
Top 3 Most Relevant Regions
π΅
No clusters available
Run an analysis first.
Dimensionality Reduction & Clustering
UMAP Embedding
768-d CONCH β PCA 64-d β UMAP 2-d
Cluster Statistics
K-Means (k=7) Β· HDBSCAN Β· PCA
Analysis History
| Slide | Date | Patches | Prediction | Confidence | AUC |
|---|
π
No report generated
Complete an analysis to generate the 10-page PDF.
PDF Report β 10 pages
CONCH v1.5 + TITAN Β· bracs_v3_model Β· explainability report
Settings
Pipeline
Patch extraction and analysis parameters
Target MPP
Microns per pixel for scanning
Patch size (px)
CONCH v1.5 native input
Stride factor
Overlap between patches
Max patches
GPU memory cap
Tissue threshold
Minimum tissue fraction per patch
Decision threshold
Sensitivity-optimised (0.30)
Model
Encoding and inference
Half precision (FP16)
Halves GPU memory usage
AMP (autocast)
torch.autocast for speed
TITAN subprocess isolation
Prevents meta tensor pollution
Encode batch size
Patches per GPU batch
Clustering
UMAP Β· t-SNE Β· KMeans Β· HDBSCAN
PCA dimensions
UMAP neighbors
K-Means k
HDBSCAN min cluster
API Reference
FastAPI backend β deploy
backend/main.py
then set its URL in the Upload page.
Connection
Configure and test your backend URL
Backend URL
Where your FastAPI server is running
Status
Not checked
Endpoints
Full REST API provided by
backend/main.pyPOST/api/analyzeUpload WSI + start pipeline
GET/api/status/{job_id}Real-time step progress
GET/api/logs/{job_id}Terminal log stream
GET/api/results/{job_id}Full results JSON
GET/api/heatmap/{job_id}attention_heatmap.png
GET/api/report/{job_id}wsi_analysis_report.pdf
GET/api/patch-data/{job_id}patch_data.csv
GET/api/embedding/{job_id}/{kind}.npy arrays (umap/tsne/conch/titan/relevance)
GET/api/historyPast analyses list
DELETE/api/job/{job_id}Remove job + files
GET/healthServer health check
Deploy Backend
Quick start commands
# Local (dev) cd backend pip install -r requirements.txt uvicorn main:app --host 0.0.0.0 --port 8000 --reload # With Docker (GPU) docker compose -f docker/docker-compose.yml up --build # Kaggle: run backend/main.py in a Kaggle notebook, # then use ngrok or Cloudflare Tunnel to expose it: # cloudflared tunnel --url http://localhost:8000