Center for Imaging Research

Medical College of Wisconsin Center for Imaging Research Software

Center for Imaging Research software capabilities

Unique Software Capabilities

  • Access to advanced software, vendor, and local prototypes
  • Ability to custom develop software using established off-label advanced care operating procedure
  • Ability to deploy third party AI tools or post-processing

Available Software

Listed are descriptions and links to several pieces of software that have been developed in the CIR and which our team has made available to the public.
Image Viewing

Python Image Viewer
The NIFTI format is ubiquitous in medical imaging research. This program, written in Python, is an exceptionally light weight viewer for NIFTI image volumes.

Python Image Compare
It is often useful to compare two volumes of images, be they of the same anatomy with different contrasts or the same underlying data with different reconstructions or processing. This program, written in Python, is an exceptionally light weight viewer to compare two different NIFTI image volumes.

Image Segmentation

Vessel Cast
In angiographic imaging volumes, the vasculature is generally significantly brighter than other tissue. It can be useful to segment this vasculature and use the resulting binarized cast for vessel rendering or measurement. This program displays angiographic images and can generate a segmentation of the vasculature through a region growing algorithm.

Training Data Curation

ROI Defining GUI
Image classification algorithms often classify individual smaller image regions, or patches. To develop a training dataset for this work, patches with the classification target and patches without the target, or controls, are needed. This program displays three orthogonal planes through a NIFTI image and writes image patches to numpy files. By centering the cursor at a given location and pressing a key to write a patch centered at that location to file, the software can also immediately write additional “augmented” patches, which include flips along all three dimensions and offsets in the centering of the patches, as well as non-overlapping “control” patches, to disk.

Image Reconstruction

Split-Slice RAKI Image Reconstruction
This deep learning based algorithm for the reconstruction of images acquired with simultaneous multi-slice imaging achieves more robust image separation compared to the originally described RAKI algorithm through a novel method of training data parameterization. Details are available in this arXiv manuscript.

DICOM Management

SCP Forward
Most clinically acquired images are stored in the DICOM format. This standards-based format is well integrated with clinical workflows and research systems. There are DICOM standards for transferring images from one system to another, and there are DICOM standards for the de-identification of medical images. This software serves as a DICOM “node” which can receive DICOM images from another DICOM node, modify DICOM image tag metadata, and send DICOM images to another DICOM node. The default tag modification configuration available with this tool complies with a DICOM standard for image de-identification.

Physiologic Files Encoded into DICOM
A C++ utility that allows the user to encode a file containing an array of text (like the physiologic waveform and trigger files of the GE MRI system) into private tags of a DICOM image, and the matching Python script to re-create the text files from the private DICOM tags. This allows physiologic monitoring files to be saved and transferred in the same manner as the DICOM images with which they are associated. The source code is provided and can be easily adapted to handle similar use cases.”