toothcanal/utils.py

import os, json, random, glob
import numpy as np
import yaml


def load_config(path="configs/default.yaml"):
    with open(path, "r") as f:
        cfg = yaml.safe_load(f)
    return cfg


def set_seed(seed=42):
    random.seed(seed)
    np.random.seed(seed)
    try:
        import torch
        torch.manual_seed(seed)
        torch.cuda.manual_seed_all(seed)
    except Exception:
        pass


def ensure_dir(p):
    os.makedirs(p, exist_ok=True)
    return p


# ----------------------------- NIfTI IO -----------------------------
def load_nii(path):
    """Return (array[z,y,x] -> we keep numpy in (x,y,z)? ) We use SimpleITK and keep
    arrays in numpy index order [x, y, z] together with spacing/origin/direction."""
    import SimpleITK as sitk
    img = sitk.ReadImage(path)
    arr = sitk.GetArrayFromImage(img)          # [z, y, x]
    arr = np.transpose(arr, (2, 1, 0))         # -> [x, y, z]
    meta = dict(
        spacing=np.array(img.GetSpacing(), dtype=np.float64),     # (sx, sy, sz)
        origin=np.array(img.GetOrigin(), dtype=np.float64),
        direction=np.array(img.GetDirection(), dtype=np.float64),
    )
    return arr, meta


def save_nii(arr_xyz, meta, path):
    import SimpleITK as sitk
    arr = np.transpose(arr_xyz, (2, 1, 0))     # back to [z, y, x]
    img = sitk.GetImageFromArray(arr)
    img.SetSpacing(tuple(float(s) for s in meta["spacing"]))
    img.SetOrigin(tuple(float(o) for o in meta["origin"]))
    img.SetDirection(tuple(float(d) for d in meta["direction"]))
    sitk.WriteImage(img, path)


def resample_to_spacing(arr_xyz, meta, new_spacing, is_label=False):
    """Resample a volume (numpy [x,y,z]) to an isotropic-ish new_spacing (list of 3)."""
    import SimpleITK as sitk
    arr = np.transpose(arr_xyz, (2, 1, 0))
    img = sitk.GetImageFromArray(arr)
    img.SetSpacing(tuple(float(s) for s in meta["spacing"]))
    img.SetOrigin(tuple(float(o) for o in meta["origin"]))
    img.SetDirection(tuple(float(d) for d in meta["direction"]))

    old_spacing = np.array(img.GetSpacing())
    old_size = np.array(img.GetSize())
    new_spacing = np.array(new_spacing, dtype=np.float64)
    new_size = np.round(old_size * (old_spacing / new_spacing)).astype(int).tolist()

    rs = sitk.ResampleImageFilter()
    rs.SetOutputSpacing(tuple(float(s) for s in new_spacing))
    rs.SetSize([int(s) for s in new_size])
    rs.SetOutputOrigin(img.GetOrigin())
    rs.SetOutputDirection(img.GetDirection())
    rs.SetInterpolator(sitk.sitkNearestNeighbor if is_label else sitk.sitkLinear)
    out = rs.Execute(img)

    o_arr = sitk.GetArrayFromImage(out)
    o_arr = np.transpose(o_arr, (2, 1, 0))
    o_meta = dict(spacing=new_spacing,
                  origin=np.array(out.GetOrigin()),
                  direction=np.array(out.GetDirection()))
    return o_arr, o_meta


def numeric_id(case_id):
    """Extract leading number from a case folder name like '031-修牙髓...' -> 31."""
    import re
    m = re.match(r"\s*0*(\d+)", str(case_id))
    return int(m.group(1)) if m else -1


def write_json(obj, path):
    with open(path, "w") as f:
        json.dump(obj, f, ensure_ascii=False, indent=2)


def read_json(path):
    with open(path, "r") as f:
        return json.load(f)