Struct RobustFitParams 

pub struct RobustFitParams {
    pub loss: RobustLoss,
    pub tuning: Option<f64>,
    pub iterations: usize,
    pub downweight_level: f64,
    pub flag_ratio: f64,
    pub estimate_offset: bool,
}

Parameters for multi_echo_robust_fit.

Fields

loss: RobustLoss

Which robust loss turns a residual into a weight.

tuning: Option<f64>

Tuning constant, in units of the robust scale; smaller is more aggressive.

None uses RobustLoss::conventional_tuning for whichever loss is selected, which is what you want unless you have measured otherwise. It is Option rather than a number because the two losses’ natural constants differ by about 3.5x: a fixed default would silently mean something quite different after changing Self::loss.

iterations: usize

IRLS reweighting passes. Three is plenty: the weights are a bounded function of a residual that is already close after the first pass, and more iterations mostly re-confirm the same decision.

0 is allowed and means report only: every robust weight stays at 1, so the field map is exactly multi_echo_linear_fit’s, and EchoQuality (including EchoQuality::flagged, which reads the final fit’s residuals) is still computed.

downweight_level: f64

Robust weight below which an echo counts as downweighted at a voxel, for EchoQuality::downweighted_fraction. Purely a reporting threshold — it changes no weight and no fit.

flag_ratio: f64

How far the worst EchoQuality::outlier_score in the series has to sit above the best before EchoQuality::flagged names anything. Reporting only; nothing is dropped.

Relative, not absolute, because an absolute residual has no fixed meaning across field strengths, echo trains and SNR. Measured on the four-echo 7 T phantom, where the ratio is max(outlier_score) / min(outlier_score):

series	ratio	flagged at 3.0
clean	2.5	no
one echo, 35% slab dropout + phase scramble	40.1	yes
one echo, +1 rad everywhere	28.2	yes
one echo, +2.5 rad everywhere	35.6	yes
one echo, 12% slab dropout	9.8	yes

3.0 sits in the gap between the clean series and the smallest true positive.

estimate_offset: bool

Estimate and remove a constant phase offset, as in multi_echo_linear_fit. Costs one degree of freedom, which matters here — see multi_echo_robust_fit.

Implementations

impl RobustFitParams

pub fn min_echoes(&self) -> usize

Fewest echoes multi_echo_robust_fit can act on: the model’s parameter count plus two, so 4 with Self::estimate_offset and 3 without. Below this it has no residual degrees of freedom to judge an echo with and returns the plain fit.

pub fn validate(&self) -> Result<(), String>

Check every field is in range, naming the first one that is not.

field	valid
tuning	None, or finite and > 0
iterations	any (0 = report only)
downweight_level	0 < level ≤ 1
flag_ratio	≥ 1 (+∞ never flags)

Trait Implementations

impl Clone for RobustFitParams

fn clone(&self) -> RobustFitParams

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for RobustFitParams

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for RobustFitParams

fn default() -> Self

Returns the “default value” for a type.

impl PartialEq for RobustFitParams

fn eq(&self, other: &RobustFitParams) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl Serialize for RobustFitParams

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for RobustFitParams