nifti-mrs-plus 0.1.0

Efficient, backend-agnostic batching of NIfTI-MRS spectra (NumPy / PyTorch / JAX / TensorFlow / Keras)

Efficient, backend-agnostic batching of NIfTI-MRS spectra

---

nifti-mrs-plus wraps a list of NIfTI-MRS objects into a single batch-aware tensor, keeping the header, dimension tags, and provenance intact. It materializes in whatever backend your pipeline needs (NumPy, PyTorch, JAX, TensorFlow, or Keras) with no code changes.

---

Install

pip install nifti-mrs-plus                   # core (numpy + nifti-mrs)
pip install "nifti-mrs-plus[torch]"          # + PyTorch
pip install "nifti-mrs-plus[viz]"            # + plotting (matplotlib + fsl-mrs)
pip install "nifti-mrs-plus[all]"            # everything

Extras: torch · jax · tensorflow · keras · viz · hdf5 · all · dev

---

Quick start

from nifti_mrs_plus import NIfTI_MRS_Plus, Backend

batch = NIfTI_MRS_Plus(nifti_list)

batch.shape                          # (8, 1, 1, 1, 2048)
batch.numpy()                        # np.ndarray, cached
batch.get_data(Backend.PYTORCH)      # torch.Tensor  (if torch installed)
batch[0]                             # NIFTI_MRS  (single subject)
batch[0:4]                           # NIfTI_MRS_Plus  (sub-batch)

---

Backends

Extra	Backend	Returns
(none)	NIFTI_LIST	list[NIFTI_MRS]
(none)	NUMPY	numpy.ndarray
torch	PYTORCH	torch.Tensor
jax	JAX	jax.Array
tensorflow	TENSORFLOW	tf.Tensor
keras	KERAS	Keras tensor

---

_{Built with ❤️ for the MRS community}