bayesian_ivim

bayesian_ivim ¶

Two-stage Bayesian IVIM fitting with empirical priors.

Implements the Barbieri et al. (MRM 2020) approach: 1. Stage 1: Least-squares fit on all voxels 2. Build empirical per-parameter priors from the population of Stage 1 fits 3. Stage 2: MAP fitting with empirical priors and per-voxel initial guesses

References

.. [1] Barbieri S et al. MRM 2020;83(6):2160-2172. doi:10.1002/mrm.28060

TwoStageBayesianIVIMFitter ¶

TwoStageBayesianIVIMFitter(
    noise_std=None,
    compute_uncertainty=False,
    prior_scale=1.5,
    stage1_max_iter=100,
)

Two-stage Bayesian IVIM fitter with empirical priors.

Stage 1 runs a standard Levenberg-Marquardt fit to obtain point estimates. Those estimates are used to build per-parameter empirical priors (mean + std). Stage 2 runs a Bayesian MAP fit using those priors as regularization.

PARAMETER	DESCRIPTION
`noise_std`	Assumed noise standard deviation. If None, estimated from data. TYPE: `float \| None` DEFAULT: `None`
`compute_uncertainty`	If True, compute posterior uncertainty maps via Laplace approximation after Stage 2. TYPE: `bool` DEFAULT: `False`
`prior_scale`	Scale factor applied to empirical prior standard deviations. Larger values create weaker priors. Default 1.5. TYPE: `float` DEFAULT: `1.5`
`stage1_max_iter`	Maximum iterations for Stage 1 LM fitting. TYPE: `int` DEFAULT: `100`

fit_image ¶

fit_image(
    model,
    data,
    mask=None,
    bounds_override=None,
    progress_callback=None,
)

Fit IVIM model using two-stage Bayesian approach.

PARAMETER	DESCRIPTION
`model`	Bound IVIM model. TYPE: `FittableModel`
`data`	Image data, shape `(x, y, z, n_observations)`. TYPE: `NDArray`
`mask`	Boolean mask. TYPE: `NDArray[bool_] \| None` DEFAULT: `None`
`bounds_override`	Per-parameter bound overrides. TYPE: `dict` DEFAULT: `None`
`progress_callback`	Progress callback (0.0 to 1.0). TYPE: `Callable` DEFAULT: `None`

RETURNS	DESCRIPTION
`dict[str, ParameterMap]`	Parameter maps.