Open Source Software Contributions
I have developed several open-source algorithms and software tools that have been widely adopted in the medical imaging community. My contributions include:
- Grow From Seed Interactive Segmentation: Core module in 3D Slicer, the world's largest open-source medical image computing platform
- Robust Statistics Segmenter: Over 39,000 accesses, integrated into 3D Slicer
- Spherical Wavelet ITK Filter: Over 5,000 downloads, included in InsightToolkit (ITK)
- Conformal Flattening ITK Filter: Over 1,700 downloads, included in InsightToolkit (ITK)
- Statistical Shape Analysis (SPoM): Quantitative shape analysis framework with Docker image available
- Multi-scale Shape Representation: Wavelet-based shape decomposition for 3D shapes
These tools have been used by researchers and clinicians worldwide and have contributed to advancing medical image computing research.
Statistical shape analysis --- A quantitatively validated approach Source code
Docker Image: docker pull gaoyi/spom
In our project:
``Gao Y, Bouix S. Statistical Shape Analysis using 3D Poisson Equation---A Quantitatively Validated Approach. Medical Image Analysis. , vol. 30, pp. 72-84, 2016.''
- We proposed a new statistical shape analysis/morphometry technique based on the 3D Poisson Equation.
- In addition, we proposed a quantitative evaluation framework for morphometry/shape analysis methods.
- To the best of our knowledge, this is the first quantitative morphometry evaluation framework and the proposed method is the morphometry technique that is systematically and quantitatively validated and evaluated.
- The source code is hosted on github at here.
Multi-scale shape representation & 3D wavelet transformation Source code
In our project:
``Y. Gao, B. Corn, D. Schifter, and A. Tannenbaum, Multiscale 3D shape representation and segmentation with applications to hippocampal/caudate extraction from brain MRI, Medical Image Analysis, vol. 16, no. 2, pp. 374-385, 2012''
we proposed a multi-scale shape decomosition and reconstruction for 3D shapes of arbitrary topology. As a bi-product, we release the code for performing 2D and 3D wavelet transform on ITK images. The source code is hosted on github at here.
Conformal map a VTK surface to a sphere Source code
Given a genus-0 surface, that is, a closed surface with no holes and no handles, this filter maps the surface to a sphere. The angle will be preserved during the mapping. It's based on the paper On the Laplace-Beltrami operator and brain surface flattening by Angenent, Haker, Tannenbaum and Kikinis, published on IEEE TMI. The implementation is now shipped with the InsightToolkit. The implementation and the usage is detailed in the
Gao Y, Melonakos J, Tannenbaum AR. Conformal Flattening ITK Filter. MICCAI, Open Science Workshop/Insight Journal, 2006
For example, the surface below on the left will be mapped to the sphere on the right.
The sources code here, however, is a further improvement over the previous publication: First, after we submitted it to the Insight Journal, thanks to the contribution of Dr. Alexandre Gouaillard and the open science philosophy of the journal, the code has been improved to be more coherent with the ITK framework. Later, we utilized the preconditioned conjugate gradient solver in the computation which significantly accelerated the algorithm. The last improvement was a to make sure that the points on the final sphere is more evenly distributed.
Comparing with the code now in ITK 3.20, the code here has tighter convergence criteria and therefore gives better result. These changes will be adopted in ITKv4.
Grow From Seed: Interactive 3D segmentation software based on local statistics Source code
This algorithm/software is an end-user oriented, interactive 3D image segmentation tool fully integrated in the 3D Slicer software. Since 2014, it has become a core interactive segmentation component of 3D Slicer, the world's largest open-source medical image computing software platform.
``Gao Y*, Chen X, Yang Q, Lasso A, Kolesov I, Pieper S, Kikinis R, Tannenbaum A, Zhu L*. An effective and open source interactive 3D medical image segmentation solution. Scientific Reports. 2024; 14:29878.''
Key Features:
- User-friendly interactive segmentation with minimal user input (just a few strokes)
- Real-time 3D segmentation feedback
- Integrated into 3D Slicer as a core module
- Used by researchers from multiple countries in their research and clinical experiments
- Supports various medical imaging modalities (MRI, CT, ultrasound)
Impact: This algorithm has been widely adopted in the medical imaging research community and is used in clinical research applications worldwide. 3D Slicer is jointly developed by universities and research institutes from the US, Germany, UK, Japan, Canada and other countries over ten years, with NIH funding exceeding $30 million.
The source code is available on GitHub: Grow From Seed Repository
A bank of shapes for validating statistical shape analysis methods
In our project:
``Gao Y, Riklin-Raviv T, Bouix S. Shape analysis, a field in need of careful validation. Human brain mapping. 2014 Oct 1;35(10):4965-78.''
We proposed a framework for validating statistical shape anlaysis methods.
TODO: I'll add the shapes here soon.
Wavelet transform on sphere
- How to do wavelet analysis on a 2-sphere?
- How that can be applied to localized analysis of a shape?
TODO: I'll add code here soon.