How to load MRI images using monai

How to load .nii using monai

To load a .nii or .nii.gz file using MONAI, you typically use LoadImaged (for dict-style data) or LoadImage (for raw image loading).

---

✅ Option 1: Quick Load with LoadImage (for single image)

from monai.transforms import LoadImage

loader = LoadImage(image_only=True)
image = loader("path/to/your_image.nii.gz")
print(image.shape)

• image_only=True: returns just the image as a NumPy array or PyTorch tensor.
• You can also get metadata (e.g., spacing, affine) by setting image_only=False.

---

✅ Option 2: Pipeline Load with LoadImaged (for dictionary-based workflows)

This is ideal if you’re working with datasets or training pipelines:

from monai.transforms import LoadImaged, Compose
from monai.data import Dataset

data_dicts = [{"image": "path/to/image.nii.gz"}]

transforms = Compose([
    LoadImaged(keys=["image"]),
])

dataset = Dataset(data=data_dicts, transform=transforms)
loaded = dataset[0]["image"]
print(loaded.shape)

---

🔄 Notes

• MONAI supports both .nii and .nii.gz.
• Images are returned as PyTorch tensors, unless image_only=False.

---

🧪 Optional: Get Affine & Metadata

loader = LoadImage(image_only=False)
image, meta = loader("your_image.nii.gz")
print(meta["affine"])

---

How to read in T1w.mgz using monai

To read a .mgz file like T1w.mgz using MONAI, you’ll need to use nibabel as an intermediate step, because MONAI’s built-in image loaders do not natively support .mgz files out of the box.

---

✅ Option 1: Load .mgz using nibabel and convert to a PyTorch tensor

import nibabel as nib
import torch
import numpy as np

# Load the .mgz file
img = nib.load("path/to/T1w.mgz")
data = img.get_fdata()

# Convert to torch tensor and add channel dimension (MONAI expects [C, H, W, D])
tensor = torch.tensor(data).unsqueeze(0).float()
print(tensor.shape)  # should be [1, H, W, D]

---

✅ Option 2: Wrap into MONAI Transform Pipeline

You can create a custom transform to use .mgz files inside a MONAI pipeline:

from monai.transforms import Transform

class LoadMGZ(Transform):
    def __call__(self, filename):
        img = nib.load(filename)
        data = img.get_fdata()
        return torch.tensor(data).unsqueeze(0).float()

Then use it like this:

from monai.transforms import Compose
from monai.data import Dataset

data_dicts = [{"image": "path/to/T1w.mgz"}]
transforms = Compose([
    LoadMGZ(),
])

dataset = Dataset(data=data_dicts, transform=transforms)
image = dataset[0]["image"]

---

🧠 Why .mgz is special

.mgz is a compressed format used by FreeSurfer, and isn’t natively supported by MONAI’s loaders like LoadImage. However, nibabel supports it fully and integrates well with MONAI if you manually handle the conversion.