Quickstart

Your first analysis

Every workflow follows the same three beats: load → decompose → describe (then analyze or render).

import spectralbrain as sb

# 1. Load — any supported format; the loader auto-detects it.
vertices, faces = sb.load_freesurfer_surface("lh.pial")

# 2. Decompose — solve the Laplace–Beltrami eigenproblem once.
mesh = sb.BrainMesh(vertices, faces)
decomp = mesh.decompose(k=300)            # -> SpectralDecomposition

# 3. Describe — every descriptor reads the same decomposition.
shapedna = sb.compute_shapedna(decomp)                 # global fingerprint
hks = sb.compute_hks(decomp, t_values=[1, 10, 100])    # multiscale, per-vertex
wks = sb.compute_wks(decomp, n_energies=100)           # band-pass, per-vertex

# 4. Render.
sb.plot_brain(data=hks[:, 0], atlas="schaefer_400")

The single SpectralDecomposition is the hub: compute it once, then derive as many descriptors as you like from it without re-solving the eigenproblem.

Pick your path

New to spectral geometry?

Start with the concepts. The Learn section explains the Laplace–Beltrami operator and what each descriptor measures, before you wire it into a study.

Go to Learn →

Already know the methods?

Jump to worked, end-to-end examples on real data, or straight to the function signatures.

Tutorials → · API →