Struct MotionCorrectionResult 

pub struct MotionCorrectionResult {
    pub phases: Vec<Vec<f64>>,
    pub magnitudes: Vec<Vec<f64>>,
    pub mask: Vec<u8>,
    pub bdirs: Vec<(f64, f64, f64)>,
    pub motion: MotionSeries,
}

A motion-corrected echo series, ready for super::run_field_mapping.

Fields

phases: Vec<Vec<f64>>

Per-echo wrapped phase on the reference echo’s grid, moved through the complex domain so the wraps survived.

magnitudes: Vec<Vec<f64>>

Per-echo magnitude on the reference echo’s grid.

mask: Vec<u8>

The input mask restricted to voxels every echo still reaches after being moved.

A rotated echo cannot fill the whole reference grid, and a resampled volume is zero where it does not reach. Those zeros are absence of data, not a zero measurement, and feeding them to an unwrapper as if they were brain is how a rim of nonsense gets into a field map. Pass this mask downstream, not the one handed in.

bdirs: Vec<(f64, f64, f64)>

Each echo’s B0 direction in the reference echo’s voxel frame. See the module docs on why these differ and what that costs.

motion: MotionSeries

The motion that was estimated, for reporting and for the suspect flags.

Implementations

impl MotionCorrectionResult

pub fn rotation_spread_deg(&self) -> f64

Angle between the extreme B0 directions in the series, in degrees — how far from a single dipole-kernel orientation this series actually is.

pub fn max_displacement_mm(&self) -> f64

Largest distance any brain tissue moved during the series, in mm. Compare it against the voxel size: below one voxel the correction had nothing to do.