Training history
Accuracy and loss over the training run.
Checking model status...
Case intake
MRI Analysis Workbench
Upload one brain MRI image, inspect the selected scan, then review confidence and class distribution after local inference.
Scan input
Upload scan
Start with a single MRI image. Results stay hidden until a case is active.
Drop brain MRI scan
PNG or JPG under 16 MB
Selected image
Ready for analysis
MRI image only
224 x 224 preprocessing
Local inference
01
Upload
Select one scan from the case workspace.
02
Analyze
Run the local TensorFlow model.
03
Review
Compare confidence, spread, and reference classes.
Model evidence
Evaluation artifacts
Training curves, confusion matrix, and Grad-CAM outputs are loaded from the project artifacts.
Confusion matrix
Class-level evaluation distribution.
Grad-CAM view
Representative model attention overlay.
Reference
Supported classes
The output order matches the trained Keras model.
Clinical guardrails
Research support only
Use the output as a research signal alongside full imaging context and professional medical review.
Validate scan type
MRI-only inputs produce the most reliable model behavior.
Review confidence
Low or close probabilities need extra scrutiny.
Confirm clinically
Do not use this output for care decisions.