Build A Brain
A current research project at the museum is rebuilding the brain structures from reverse casts of the interiors of fossil skulls. This is done manually by visually checking greyscale values in many hundreds of CT scan TIFF image files. Expensive proprietary software is currently being used to build these 3D models based on the image stacks. Automation of this process could save incredible amounts of time and rapidly advance the state of the research. Is it possible to use machine learning, machine vision, medical imaging advances to address this?
- Software or scripts which may enhance local contrast of images in stack to improve edge detection (say between bone and deposited rock)
- Making a 3D model of just the skull without air or sedimentary rock surrounding the skull (turning the TIFF image stack into .stl files)
- Based on skull reconstruction model, create the reverse cast and output an STL
- EXTRA CREDIT - understand VGL file format (XML?) and output a "rough draft" in VGL file
- Example video of brain reconstruction
- Information about the AMNH CT scanner
- GE Phoenix CT scanner
- CT image stacks of TIFF files (Please ask Hack the Dinos organizers for the USB stick containing images)
- .pca files with metadata - need translation of file format / parameters (Please ask Hack the Dinos organizers for the USB stick containing .pca files)
- ImageJ2 - used to enhance contrast in grayscale images, could be useful for many aspects of solutions for this challenge
- myVGL viewer - freely available software to view vg project files
- myVGL docs
- Video depicting manual process an example of how it is currently done
- Screen capture below of VG studio showing outline of bone manually created by human - the light blue outlines indicate the skull ROI (region of interest) and the dark blue outline is the endocast ROI.
