Medaset

A collection of pre-built dataset classes for MEDical datASETs.

Objectives

• Compatible to PyTorch and MONAI
• Providing some off-the-shelf features to the Dataset class, including but not limited to, dataset extracting, loading and visualization.

Getting started

git clone this repo and install it using

pip install -e .

Caution

Naming convention changed!! For the sake of readibility, only the first letters in each acronym are now capitalized.

Currently Supported Datasets

• AMOS (Abdominal Multi-Organ Segmentation)
  [Grand Challenge] [Arxiv]
  • AMOS provides 500 CT and 100 MRI scans collected from multi-center, multi-vendor, multi-modality, multi-phase, multi-disease patients, each with voxel-level annotations of 15 abdominal organs.
  • Structure: spleen, right kidney, left kidney, gallbladder, esophagus, liver, stomach, aorta, inferior vena cava, pancreas, right adrenal gland, left adrenal gland, duodenum, bladder, prostate/uterus.
  • Dataset classes: AmosDataset, SimpleAmosDataset

  ---

• CHAOS (Combined (CT-MR) Healthy Abdominal Organ Segmentation)
  [Grand Challenge] [Arxiv]
  • The CHAOS challenge data contains 40 CT scans and 40 MR scans of upper abdomen area.
  • Structure (Modality): Liver (CT/MR), Kidneys (MR), Spleen (MR)
  • Dataset classes: ChaosCtDataset, ChaosT2spirDataset (The T1-DUAL sequence is not supported currently)

  ---

• SMAT (Skeletal Muscle and Adipose Tissue)
  • Private dataset
  • Dataset classes: SmatCtDataset, SmatMrDataset, SmatDataset

Useful Links

Metadata Desciptions

• NIfTI file header discriptions
• DICOM Standard Browser

Medical Image Coordinate Systems

• An Introduction from 3D Slicer
• An Introduction of coordinate systems and affines from nipy.org
• Defining the DICOM orientation
• Geometry in Medical Imaging: DICOM and NIfTI formats
• The First Step for Neuroimaging Data Analysis: DICOM to NIfTI conversion