code/m4raw_io.py

#!/usr/bin/env python3
"""Portable I/O helpers for M4Raw Brain H5 files."""

from __future__ import annotations

import csv
from dataclasses import dataclass
from pathlib import Path
from typing import Any

import h5py
import numpy as np


NATIVE_KSPACE_AXES = ("slice", "coil", "frequency_encode", "phase_encode")
NATIVE_TARGET_AXES = ("slice", "frequency_encode", "phase_encode")


@dataclass(frozen=True)
class M4RawVolume:
    """One complete H5 acquisition in its native stored axis order."""

    kspace: np.ndarray
    reconstruction_rss: np.ndarray
    acquired_pe_mask: np.ndarray
    ismrmrd_header: bytes
    attrs: dict[str, Any]


def _plain_attr(value: Any) -> Any:
    if isinstance(value, np.generic):
        return value.item()
    if isinstance(value, bytes):
        return value.decode("utf-8")
    return value


def derive_acquired_pe_mask(kspace: np.ndarray) -> np.ndarray:
    """Return one for every acquired PE sample on the final stored axis."""
    array = np.asarray(kspace)
    if array.ndim != 4:
        raise ValueError(f"Expected [slice,coil,FE,PE], got {array.shape}")
    return np.any(array != 0, axis=(0, 1, 2)).astype(np.uint8)


def read_h5(path: str | Path) -> M4RawVolume:
    """Read one H5 file without changing its native axis order."""
    with h5py.File(path, "r") as handle:
        header = handle["ismrmrd_header"][()]
        if isinstance(header, str):
            header = header.encode("utf-8")
        kspace = np.asarray(handle["kspace"])
        if "acquired_pe_mask" in handle:
            pe_mask = np.asarray(handle["acquired_pe_mask"], dtype=np.uint8)
        else:
            pe_mask = derive_acquired_pe_mask(kspace)
        return M4RawVolume(
            kspace=kspace,
            reconstruction_rss=np.asarray(
                handle["reconstruction_rss"], dtype=np.float32
            ),
            acquired_pe_mask=pe_mask,
            ismrmrd_header=header,
            attrs={key: _plain_attr(value) for key, value in handle.attrs.items()},
        )


def read_target(path: str | Path) -> np.ndarray:
    """Read only the stored RSS target as float32."""
    with h5py.File(path, "r") as handle:
        return np.asarray(handle["reconstruction_rss"], dtype=np.float32)


def centered_ifft_rss(kspace: np.ndarray) -> np.ndarray:
    """Reconstruct `[slice,coil,FE,PE]` k-space with orthonormal IFFT/RSS."""
    array = np.asarray(kspace)
    if array.ndim != 4:
        raise ValueError(f"Expected [slice,coil,FE,PE], got {array.shape}")
    coil_images = np.fft.fftshift(
        np.fft.ifft2(
            np.fft.ifftshift(array, axes=(-2, -1)),
            axes=(-2, -1),
            norm="ortho",
        ),
        axes=(-2, -1),
    )
    return np.sqrt(np.sum(np.abs(coil_images) ** 2, axis=1)).astype(np.float32)


def read_manifest(path: str | Path) -> list[dict[str, str]]:
    """Read a packaged tab-separated manifest."""
    with Path(path).open(newline="", encoding="utf-8") as handle:
        return list(csv.DictReader(handle, delimiter="\t"))