Data Access

Reading Data

After loading a twix file, access scan data through the RawData objects stored on the TwixObj:

twixObj = read_twix("meas.dat")

# Full data read
data = getdata(twixObj.image)

# Or equivalently, using indexing syntax
data = twixObj.image[:, :, :]

Scan Types

A twix file can contain multiple scan types. Use MDH_flags to see which ones have data:

MDH_flags(twixObj)
# => ["image", "noise", "refscan", "phasecor"]

Common scan types:

Type	Description
`image`	Primary image acquisitions
`noise`	Noise adjustment scans
`refscan`	GRAPPA/ACS reference lines
`refscanPC`	Phase correction reference scans
`phasecor`	EPI phase correction navigators
`phasestab`	Phase stabilization scans

Access each via dot syntax: twixObj.image, twixObj.noise, twixObj.refscan, etc.

The 16-Dimension Data Model

Siemens raw data is organized into 16 dimensions:

Index	Name	Description
1	`Col`	Readout (columns / samples)
2	`Cha`	Receive channels (coils)
3	`Lin`	Phase-encoding lines
4	`Par`	3D partition encodes
5	`Sli`	Slices
6	`Ave`	Averages
7	`Phs`	Phases (cardiac, etc.)
8	`Eco`	Echoes
9	`Rep`	Repetitions
10	`Set`	Sets
11	`Seg`	Segments (EPI interleaves)
12–16	`Ida`–`Ide`	Free indices

Inspect dimensions with:

fullSize(twixObj.image)   # all 16 dimensions
# => [4096, 32, 128, 1, 5, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]

dataSize(twixObj.image)   # accounts for removeOS and averaging
sqzSize(twixObj.image)    # non-singleton dimensions only
# => [4096, 32, 128, 5]

sqzDims(twixObj.image)    # names of non-singleton dimensions
# => ["Col", "Cha", "Lin", "Sli"]

Slicing / Indexing

Use standard Julia indexing on RawData objects:

# Read all data
data = twixObj.image[:, :, :]

# Read specific ranges (1-based indexing)
data = twixObj.image[1:128, :, :]

# Single index
data = twixObj.image[1, 2, :]

Interaction with `squeeze`

When squeeze = false (default), indices map directly to the 16 dimensions:

twixObj.image.squeeze = false
data = twixObj.image[1:128, :, 1:64, :, 1:3, :, :, :, :, :, :, :, :, :, :, :]
#                    Col    Cha Lin    Par Sli ...

When squeeze = true, indices map to non-singleton dimensions only:

twixObj.image.squeeze = true
sqzDims(twixObj.image)  # => ["Col", "Cha", "Lin", "Sli"]

data = twixObj.image[1:128, :, 1:64, 1:3]
#                    Col    Cha Lin    Sli

This is much more convenient — you don't need to specify all 16 dimensions.

Trailing Colon Behavior

A trailing : captures all remaining dimensions (similar to MATLAB):

twixObj.image.squeeze = true
# These are equivalent:
data = twixObj.image[1:128, :]     # Col=1:128, everything else
data = twixObj.image[1:128, :, :, :]  # Col=1:128, Cha=all, Lin=all, Sli=all

Processing Flags

All processing flags default to false — MRITwixTools.jl returns raw, unprocessed data by default. Enable processing either at load time or afterwards:

# At load time
twixObj = read_twix("meas.dat", removeOS=true, squeeze=true)

# Or afterwards (triggers recomputation on next data read)
twixObj.image.removeOS = true
twixObj.image.squeeze = true
data = getdata(twixObj.image)

Available Flags

Flag	Default	Effect
`removeOS`	`false`	Remove 2× readout oversampling via FFT crop (halves `Col` dimension)
`regrid`	`false`	Ramp-sample regridding for non-Cartesian readouts (requires trajectory)
`doAverage`	`false`	Average across the `Ave` dimension
`averageReps`	`false`	Average across the `Rep` dimension
`averageSets`	`false`	Average across the `Set` dimension
`ignoreSeg`	`false`	Collapse the `Seg` dimension
`squeeze`	`false`	Drop singleton dimensions from returned arrays
`disableReflect`	`false`	Skip readout reflection (bipolar gradient) correction
`skipToFirstLine`	varies	Skip to first acquired k-space line (see below)
`ignoreROoffcenter`	`false`	Ignore readout off-center phase correction during regridding

`skipToFirstLine` Default Behavior

This flag has a scan-type-dependent default:

image and phasestab: defaults to false (preserve full k-space extent)
All other scan types (noise, refscan, etc.): defaults to true (skip unacquired leading lines)

This matters for partial-Fourier or GRAPPA reference scans where acquisition doesn't start at line 0.

Backward-Compatible Flag Names

The old flag-prefixed names still work as aliases:

twixObj.image.flagRemoveOS = true     # same as twixObj.image.removeOS = true
twixObj.image.flagDoAverage = true    # same as twixObj.image.doAverage = true
twixObj.image.flagIgnoreSeg = true    # same as twixObj.image.ignoreSeg = true

The setter functions also remain available:

set_flagRemoveOS!(twixObj.image, true)
set_flagRampSampRegrid!(twixObj.image, true)  # errors if no trajectory available

Processing Pipeline

When data is read (via getdata or indexing), the following processing steps are applied in order:

Read raw data from file at stored byte positions
Reflection correction — reverse readout for bipolar gradient lines (unless disableReflect)
Regridding — interpolate ramp-sampled readouts onto Cartesian grid (if regrid)
Oversampling removal — FFT → crop center half → IFFT (if removeOS)
Range selection — extract requested slice/index ranges
Averaging — accumulate and divide for averaged dimensions
Squeeze — drop singleton dimensions (if squeeze)

Unsorted Data Access

For debugging or custom processing, read raw acquisitions without sorting into the k-space grid:

# All acquisitions as [NCol, NCha, NAcq]
raw = unsorted(twixObj.image)

# Single acquisition (by index)
acq = unsorted(twixObj.image, 42)