abusayemnur/neuroscan-ai-main-2026 ? reverse-engineered prompt

Build me a brain tumor MRI web app called NEUROSCAN AI that lets a user upload or open sample MRI scans and get both a 2D tumor classification result and a 3D segmentation view.

I want a clean radiology style interface with a dark canvas, a main viewer for axial slices, and controls to scrub through slices, play them like a cine loop, adjust opacity and contrast, and switch between the raw image, prediction overlay, ground truth mask, and split compare view. The app should classify scans into Glioma, Meningioma, Pituitary Adenoma, or No Tumor, and show a Grad CAM style heatmap for explanation.

It should also support NIfTI files for 3D tumor segmentation, with synchronized axial, coronal, and sagittal views, plus an interactive 3D surface model of the tumor. Show volume stats and basic validation metrics like Dice and IoU when masks are available.

Use the existing backend and frontend structure, and look up current docs online if you need to.