steinbock.io

img_dtype

mask_dtype

list_data_files(data_dir, base_files=None)

Source code in steinbock/io.py

def list_data_files(
    data_dir: Union[str, PathLike],
    base_files: Optional[Sequence[Union[str, PathLike]]] = None,
) -> List[Path]:
    if base_files is not None:
        return _list_related_files(base_files, data_dir, ".csv")
    return sorted(Path(data_dir).rglob("*.csv"))

list_image_files(img_dir, base_files=None)

Source code in steinbock/io.py

def list_image_files(
    img_dir: Union[str, PathLike],
    base_files: Optional[Sequence[Union[str, PathLike]]] = None,
) -> List[Path]:
    if base_files is not None:
        return _list_related_files(base_files, img_dir, ".tiff")
    return sorted(Path(img_dir).rglob("*.tiff"))

list_mask_files(mask_dir, base_files=None)

Source code in steinbock/io.py

def list_mask_files(
    mask_dir: Union[str, PathLike],
    base_files: Optional[Sequence[Union[str, PathLike]]] = None,
) -> List[Path]:
    if base_files is not None:
        return _list_related_files(base_files, mask_dir, ".tiff")
    return sorted(Path(mask_dir).rglob("*.tiff"))

list_neighbors_files(neighbors_dir, base_files=None)

Source code in steinbock/io.py

def list_neighbors_files(
    neighbors_dir: Union[str, PathLike],
    base_files: Optional[Sequence[Union[str, PathLike]]] = None,
) -> List[Path]:
    if base_files is not None:
        return _list_related_files(base_files, neighbors_dir, ".csv")
    return sorted(Path(neighbors_dir).rglob("*.csv"))

read_data(data_file)

Source code in steinbock/io.py

def read_data(data_file: Union[str, PathLike]) -> pd.DataFrame:
    return pd.read_csv(
        data_file, sep=",|;", index_col="Object", engine="python"
    )

read_image(img_file, use_imageio=False, native_dtype=False)

Source code in steinbock/io.py

def read_image(
    img_file: Union[str, PathLike],
    use_imageio: bool = False,
    native_dtype: bool = False,
) -> np.ndarray:
    if use_imageio:
        img = imageio.volread(img_file)
        orig_img_shape = img.shape
        while img.ndim > 3 and img.shape[0] == 1:
            img = img.sqeeze(axis=0)
        while img.ndim > 3 and img.shape[-1] == 1:
            img = img.sqeeze(axis=img.ndim - 1)
        if img.ndim == 2:
            img = img[np.newaxis, :, :]
        elif img.ndim != 3:
            raise ValueError(
                f"Unsupported shape {orig_img_shape} for image {img_file}"
            )
    else:
        img = tifffile.imread(img_file, squeeze=False)
        if img.ndim == 2:
            img = img[np.newaxis, :, :]
        elif img.ndim == 5:
            size_t, size_z, size_c, size_y, size_x = img.shape
            if size_t != 1 or size_z != 1:
                raise ValueError(
                    f"{img_file}: unsupported TZCYX shape {img.shape}"
                )
            img = img[0, 0, :, :, :]
        elif img.ndim == 6:
            size_t, size_z, size_c, size_y, size_x, size_s = img.shape
            if size_t != 1 or size_z != 1 or size_s != 1:
                raise ValueError(
                    f"{img_file}: unsupported TZCYXS shape {img.shape}"
                )
            img = img[0, 0, :, :, :, 0]
        elif img.ndim != 3:
            raise ValueError(
                f"{img_file}: unsupported number of dimensions ({img.ndim})"
            )
    if not native_dtype:
        img = _to_dtype(img, img_dtype)
    return img

read_image_info(image_info_file)

Source code in steinbock/io.py

def read_image_info(image_info_file: Union[str, PathLike]) -> pd.DataFrame:
    image_info = pd.read_csv(
        image_info_file,
        sep=",|;",
        dtype={
            "image": pd.StringDtype(),
            "width_px": pd.UInt16Dtype(),
            "height_px": pd.UInt16Dtype(),
            "num_channels": pd.UInt8Dtype(),
        },
        engine="python",
    )
    for required_col in ("image", "width_px", "height_px", "num_channels"):
        if required_col not in image_info:
            raise ValueError(
                f"Missing '{required_col}' column in {image_info_file}"
            )
    for notnan_col in ("image", "width_px", "height_px", "num_channels"):
        if notnan_col in image_info and image_info[notnan_col].isna().any():
            raise ValueError(
                f"Missing values for '{notnan_col}' in {image_info_file}"
            )
    for unique_col in ("image",):
        if unique_col in image_info:
            if image_info[unique_col].dropna().duplicated().any():
                raise ValueError(
                    f"Duplicated values for '{unique_col}'"
                    f" in {image_info_file}"
                )
    return image_info

read_mask(mask_file, native_dtype=False)

Source code in steinbock/io.py

def read_mask(
    mask_file: Union[str, PathLike], native_dtype: bool = False
) -> np.ndarray:
    mask = tifffile.imread(mask_file, squeeze=False)
    if mask.ndim == 5:
        size_t, size_z, size_c, size_y, size_x = mask.shape
        if size_t != 1 or size_z != 1 or size_c != 1:
            raise ValueError(
                f"{mask_file}: unsupported TZCYX shape {mask.shape}"
            )
        mask = mask[0, 0, 0, :, :]
    elif mask.ndim == 6:
        size_t, size_z, size_c, size_y, size_x, size_s = mask.shape
        if size_t != 1 or size_z != 1 or size_c != 1 or size_s != 1:
            raise ValueError(
                f"{mask_file}: unsupported TZCYXS shape {mask.shape}"
            )
        mask = mask[0, 0, 0, :, :, 0]
    elif mask.ndim != 2:
        raise ValueError(
            f"{mask_file}: unsupported number of dimensions ({mask.ndim})"
        )
    if not native_dtype:
        mask = _to_dtype(mask, mask_dtype)
    return mask

read_neighbors(neighbors_file)

Source code in steinbock/io.py

def read_neighbors(neighbors_file: Union[str, PathLike]) -> pd.DataFrame:
    return pd.read_csv(
        neighbors_file,
        sep=",|;",
        usecols=["Object", "Neighbor", "Distance"],
        dtype={
            "Object": mask_dtype,
            "Neighbor": mask_dtype,
            "Distance": np.float32,
        },
        engine="python",
    )

read_panel(panel_file, unfiltered=False)

Source code in steinbock/io.py

def read_panel(
    panel_file: Union[str, PathLike], unfiltered: bool = False
) -> pd.DataFrame:
    panel = pd.read_csv(
        panel_file,
        sep=",|;",
        dtype={
            "channel": pd.StringDtype(),
            "name": pd.StringDtype(),
            "keep": pd.BooleanDtype(),
        },
        engine="python",
        true_values=["1"],
        false_values=["0"],
    )
    for required_col in ("channel", "name"):
        if required_col not in panel:
            raise ValueError(
                f"Missing '{required_col}' column in {panel_file}"
            )
    for notnan_col in ("channel", "keep"):
        if notnan_col in panel and panel[notnan_col].isna().any():
            raise ValueError(
                f"Missing values for '{notnan_col}' in {panel_file}"
            )
    for unique_col in ("channel", "name"):
        if unique_col in panel:
            if panel[unique_col].dropna().duplicated().any():
                raise ValueError(
                    f"Duplicated values for '{unique_col}' in {panel_file}"
                )
    if not unfiltered and "keep" in panel:
        panel = panel.loc[panel["keep"].astype(bool), :]
    return panel

write_data(data, data_file)

Source code in steinbock/io.py

def write_data(data: pd.DataFrame, data_file: Union[str, PathLike]) -> None:
    data = data.reset_index()
    data.to_csv(data_file, index=False)

write_image(img, img_file, ignore_dtype=False)

Source code in steinbock/io.py

def write_image(
    img: np.ndarray,
    img_file: Union[str, PathLike],
    ignore_dtype: bool = False,
) -> None:
    if not ignore_dtype:
        img = _to_dtype(img, img_dtype)
    tifffile.imwrite(
        img_file,
        data=img[np.newaxis, np.newaxis, :, :, :, np.newaxis],
        imagej=True,
    )

write_image_info(image_info, image_info_file)

Source code in steinbock/io.py

def write_image_info(
    image_info: pd.DataFrame, image_info_file: Union[str, PathLike]
) -> None:
    image_info.to_csv(image_info_file, index=False)

write_mask(mask, mask_file)

Source code in steinbock/io.py

def write_mask(mask: np.ndarray, mask_file: Union[str, PathLike]) -> None:
    mask = _to_dtype(mask, mask_dtype)
    tifffile.imwrite(
        mask_file,
        data=mask[np.newaxis, np.newaxis, np.newaxis, :, :, np.newaxis],
        imagej=True,
    )

write_neighbors(neighbors, neighbors_file)

Source code in steinbock/io.py

def write_neighbors(
    neighbors: pd.DataFrame, neighbors_file: Union[str, PathLike]
) -> None:
    neighbors = neighbors.loc[:, ["Object", "Neighbor", "Distance"]].astype(
        {
            "Object": mask_dtype,
            "Neighbor": mask_dtype,
            "Distance": np.float32,
        }
    )
    neighbors.to_csv(neighbors_file, index=False)

write_panel(panel, panel_file)

Source code in steinbock/io.py

def write_panel(panel: pd.DataFrame, panel_file: Union[str, PathLike]) -> None:
    panel = panel.copy()
    for col in panel.columns:
        if panel[col].convert_dtypes().dtype == pd.BooleanDtype():
            panel[col] = panel[col].astype(pd.UInt8Dtype())
    panel.to_csv(panel_file, index=False)