MONAI — Medical Open Network for AI Imaging

MONAI — Medical Open Network for AI (Imaging)

MONAI is the dominant open-source framework for deep learning in medical imaging. Built on PyTorch, it provides production-grade components for 2D/3D image segmentation, classification, detection, and registration across CT, MRI, pathology, ultrasound, and X-ray modalities.

Key Features

• Domain-specific transforms: intensity normalisation, random cropping, spacing resampling, affine augmentation in 3D
• Pre-trained model zoo: auto-segmentation for 104 anatomical structures, whole-body CT, brain MRI
• MONAI Label: active-learning annotation server (integrates with 3D Slicer, OHIF)
• Distributed training, AMP, and gradient checkpointing for large 3D volumes
• Federated learning support via FLARE
• NIfTI, DICOM, and MetaImage I/O out of the box

Quick Start

pip install monai

from monai.networks.nets import UNet
from monai.losses import DiceLoss
from monai.transforms import Compose, LoadImaged, ScaleIntensityd

model = UNet(spatial_dims=3, in_channels=1, out_channels=2,
             channels=(16,32,64,128,256), strides=(2,2,2,2))
loss_fn = DiceLoss(to_onehot_y=True, softmax=True)

npx ai-supply add monai-medical-imaging-framework

Curated mirror of the open-source MONAI (Apache-2.0). Get it from the source.