signal_model

signal_model ¶

DSC perfusion forward model and binding adapter.

Implements C(t) = CBF * AIF(t) * R(t) (convolution) as a BaseSignalModel with pre-computed SVD via BoundDSCModel.

DSCConvolutionModel ¶

Bases: BaseSignalModel

DSC perfusion forward model: C(t) = CBF * AIF(t) * R(t).

Parameters are CBF (mL/100g/min), MTT (s), and Ta (s).

References

.. [1] OSIPI CAPLEX, https://osipi.github.io/OSIPI_CAPLEX/ .. [2] Ostergaard L et al. MRM 1996;36(5):715-725.

name `property` ¶

name

Return human-readable component name.

parameters `property` ¶

parameters

Return list of parameter names: CBF, MTT, Ta.

parameter_units `property` ¶

parameter_units

Return mapping of parameter names to OSIPI-compliant units.

reference `property` ¶

reference

Return primary literature citation.

get_bounds ¶

get_bounds()

Return default parameter bounds as {name: (lower, upper)}.

BoundDSCModel ¶

BoundDSCModel(
    model, aif, time, matrix_type="circulant", fixed=None
)

Bases: BaseBoundModel

DSC model with AIF and time bound, SVD pre-computed.

Pre-computes the SVD of the AIF convolution matrix once. The fitter uses the pre-computed SVD components to recover R(t) via regularized inversion.

PARAMETER	DESCRIPTION
`model`	The signal model. TYPE: `DSCConvolutionModel`
`aif`	Arterial input function (delta-R2), shape `(n_timepoints,)`. TYPE:* `NDArray`
`time`	Time points in seconds, shape `(n_timepoints,)`. TYPE: `NDArray`
`matrix_type`	`'circulant'` (default, delay-insensitive) or `'toeplitz'` (causal). TYPE: `str` DEFAULT: `'circulant'`
`fixed`	Parameters to fix. TYPE: `dict[str, float] \| None` DEFAULT: `None`

ensure_device ¶

ensure_device(xp)

Transfer arrays to the target device.

predict_array_batch ¶

predict_array_batch(free_params_batch, xp)

Forward model: reconstruct C(t) from CBF, MTT, Ta.

PARAMETER	DESCRIPTION
`free_params_batch`	Free parameter values, shape `(n_free, n_voxels)`. TYPE: `NDArray`
`xp`	Array module. TYPE: `module`

RETURNS	DESCRIPTION
`NDArray`	Predicted concentration, shape `(n_timepoints, n_voxels)`.

get_initial_guess_batch ¶

get_initial_guess_batch(observed_batch, xp)

Compute initial guesses from signal shape.

PARAMETER	DESCRIPTION
`observed_batch`	Observed data, shape `(n_timepoints, n_voxels)`. TYPE: `NDArray`
`xp`	Array module. TYPE: `module`

RETURNS	DESCRIPTION
`NDArray`	Initial guesses, shape `(n_free, n_voxels)`.