Data

The data module provides comprehensive tools for loading, processing, and visualizing medical images in MIPCandy.

Overview

Datasets:

• PyTorch-compatible dataset classes for medical imaging

• Built-in K-fold cross validation

• nnU-Net format support with multimodal handling

• Dataset inspection and patch-based training

See Datasets for detailed documentation.

Visualization:

• 2D and 3D rendering with Matplotlib and PyVista

• Overlay segmentation masks on images

• Automatic value normalization

See Visualization for detailed documentation.

Quick Start

from mipcandy import NNUNetDataset, visualize2d, overlay
from torch.utils.data import DataLoader

# Load dataset with K-fold support
dataset = NNUNetDataset("dataset/", device="cuda")
train, val = dataset.fold(fold=0)

# Create data loader
loader = DataLoader(train, batch_size=4, shuffle=True)

# Visualize sample
image, label = train[0]
overlaid = overlay(image, label)
visualize2d(overlaid)