I/O

See I/O and format support for the reader contract, the full format-support table, and exactly what ISMRMRDReader does and does not handle yet.

Dispatch

Read a raw-data file into :class:MRIData, choosing the reader automatically.

Raises:

Type	Description
FileNotFoundError	If path does not exist.
UnsupportedFormatError	If no registered reader recognises the file.

Bases: ABC

Base class for all raw-data readers.

can_read(path) abstractmethod

Return True if this reader recognises path.

Implementations should be cheap: check the extension and, if needed, sniff a few magic bytes. Do not parse the whole file here.

read(path, **options) abstractmethod

Parse path and return a validated :class:MRIData.

Implementations must call .validate() on the result and record a provenance step describing the read.

Register a :class:Reader instance for use by :func:read.

Parameters:

Name	Type	Description	Default
reader	Reader	The reader instance.	required
prepend	bool	If True, give this reader priority over already-registered ones.	False

Return the registered readers, in priority order.

Return the first registered reader whose :meth:Reader.can_read accepts path.

Readers

ISMRMRDReader is the only reader with a real read() today (wraps the ismrmrd package; needs pip install "unimri[ismrmrd]"). HDF5Reader and TwixReader are registered (so can_read/format-detection works) but their read() still raises NotImplementedError — see the roadmap.

Bases: Reader

Read a single Cartesian encoding space from an ISMRMRD file.

Wraps the reference ismrmrd package (pip install "unimri[ismrmrd]"). Scope (Milestone 1, see docs/roadmap.md): exactly one slice / average / contrast / repetition / set / segment. Multi-dimensional acquisitions and non-Cartesian ISMRMRD trajectories raise a clear :class:~unimri.exceptions.ReaderError naming what isn't supported, rather than silently mishandling the data -- see docs/io-format-support.md for the full scope. Noise-measurement acquisitions (ACQ_IS_NOISE_MEASUREMENT) are skipped; pre-whitening is not yet applied. ISMRMRD's header has no standard nucleus field, so :attr:~unimri.data.AcquisitionInfo.nucleus stays at its "1H" default -- set it manually for multinuclear files.

read(path, **options)

Read an ISMRMRD file.

Options

dataset_name: HDF5 group holding the dataset (default "dataset", the ISMRMRD convention).