MIScnn

MIScnn implements a Python framework for developing and evaluating convolutional neural network (CNN) and deep learning-based medical image segmentation pipelines.


Key Features:

  • Comprehensive Pipeline Components: Implements data input/output (I/O), preprocessing, data augmentation, patch-wise analysis, and evaluation metrics for medical image segmentation.
  • Models: Includes a curated collection of deep learning architectures, including 3D U-Net, with demonstrated use in the Kidney Tumor Segmentation Challenge 2019.
  • Full Pipeline Customization: Provides configurable pipeline components and open interfaces to tailor preprocessing, model training, and inference parameters.
  • Automated Evaluation: Supports automated evaluation workflows, including cross-validation, for rigorous model assessment across datasets.

Scientific Applications:

  • Multi-class semantic segmentation: Enables multi-class semantic segmentation of volumetric medical imaging datasets.
  • Kidney tumor segmentation: Has been applied to kidney tumor segmentation as demonstrated in the Kidney Tumor Segmentation Challenge 2019.

Methodology:

Integrates data preprocessing, data augmentation, patch-wise analysis, convolutional neural networks (e.g., 3D U-Net), and automated evaluation including cross-validation.

Topics

Details

License:
GPL-3.0
Tool Type:
command-line tool
Programming Languages:
Python
Added:
3/19/2021
Last Updated:
4/11/2021

Operations

Publications

Müller D, Kramer F. MIScnn: a framework for medical image segmentation with convolutional neural networks and deep learning. BMC Medical Imaging. 2021;21(1). doi:10.1186/s12880-020-00543-7. PMID:33461500. PMCID:PMC7814713.

PMID: 33461500
PMCID: PMC7814713
Funding: - Bundesministerium für Bildung und Forschung: 01ZZ1804E