Function apply_mask_ops 

pub fn apply_mask_ops(
    mask: Vec<u8>,
    ops: &[MaskOp],
    input_data: &[f64],
    magnitude: Option<&[f64]>,
    metadata: &ScanMetadata,
) -> Result<Vec<u8>, PipelineError>

Apply mask operations to an existing mask, in order.

This is the one implementation of what every mask op means; build_mask_section is this function starting from an all-ones mask, and hosts that drive masking themselves (e.g. an interactive UI applying one refinement at a time) should call it rather than reimplement the operations, so a mask built step-by-step matches the one a --mask section would produce.

Two different images. input_data is what a generator looks at — thresholding may use a phase-quality map, for instance. magnitude is the magnitude image, and is what the ops that need real signal use: MaskOp::Bet, MaskOp::HdBet and MaskOp::SignalErode. Passing the section input as magnitude is a bug: signal-gated erosion divides out a receive-coil bias estimate and gates on the in-mask median, which only means anything for a magnitude image. Those ops error when magnitude is None.

Arguments

• mask - Starting mask (0/1), length nx*ny*nz
• ops - Operations to apply, in order
• input_data - Image the generators threshold (may be a phase-quality map)
• magnitude - Magnitude image, for BET / HD-BET / signal-gated erosion
• metadata - Scan metadata (dims + voxel size)