I am moving from MGH to Northeastern University starting September 1st. I will become a faculty member in the newly founded Bioengineering department in the College of Engineering, and start a new computational imaging lab there. I am very excited because of the opportunities for expanding my research and mentoring/teaching students. The rapid growth of the University provides a growing interdisciplinary environment for me to establish new collaborations and attract bright students. I am sure it will be an exciting new adventure and a lot of fun!
As part of this new opportunity, I am looking for strong candidates to fill two postdoc openings in my new lab - one for candidates with strong optics/data analysis background, the other for candidates with computational physics and GPU-programming background. The full job descriptions can be read online at
In addition to working on NIH/USAID funded projects, this two postdoc associates will help me creating my lab, as well as mentoring PhD students. If you are interested, please apply online or directly email me (q.fang at neu.edu) your cover letter+CV+3 references. Feel free to share this link to anyone who might be interested.
Our new paper entitled "Characterizing breast lesions through robust multimodal data fusion using independent diffuse optical and x-ray breast imaging" was published online today on JBO Letters (with Open Access). In this paper, we proposed and clinically validated a flexible structure-function data fusion approach that can combines independent optical and (2D) mammographic scans of breasts. We show that this approach can achieve comparable performance to our co-registered optical/x-ray system, despite the anticipated negative impact from the registration errors.
I was also extremely excited after receiving an email from PATH:
"We’re pleased to inform you that the non-contact mobile oximeter has been nominated by independent experts as a leading innovation in the maternal, child, and newborn health category as part of Reimagining Global Health, the inaugural report of the Innovation Countdown 2030 initiative. Innovation Countdown 2030 is a PATH-led initiative to identify and showcase lifesaving innovations with great promise to transform global health by 2030 and to help accelerate progress toward the new health targets proposed in the United Nations Sustainable Development Goals."
"PATH’s Innovation Countdown 2030 received over 500 lifesaving ideas from 50 countries. Experts say these were the top 30, with remarkable potential to save lives faster. http://ic2030.org/ #IC2030"
The full Innovation Countdown 2030 report can be downloaded here. It is certainly a pleasure to be mentioned in this report, on the hand, will definitely work hard to get the idea to work. In the meantime, I wish more people to join this endeavor, giving those kids a chance to survive, and make our world a better place!
As part of our recent BOE paper, we'd like to share a 3D digital breast phantom to the community to allow others to test and validate new imaging algorithms and optimize measurement configurations for a compressed breast. We call it DigiBreast. One unique aspect of the DigiBreast is the clinically derived 3D tissue volume fraction maps (for adipose and fibroglandular tissues). Using a compostional model to represent a breast anatomy is much more realistic compared to the conventional piece-wise-constant segmented anatomies, because the former preserves the fine spatial details in the structural images.
We anticipate this digital phantom to be useful for many breast-imaging related research, especially those involving model-based image reconstructions. Potential utilities include, but not limited to, simulations of breast deformation, 2D and 3D x-ray breast imaging, and functional imaging of a compressed breast using tomographic optical, microwave, thermal and electrical impedance methods.
The digital breast phantom data (in matlab .mat format, as well as in JSON/UBJSON format) can be downloaded after registration.
I'd like to thank Dr. Mats Lundquvist and his team from Philips Healthcare for the contributions to this dataset.
We have a new paper(PDF,HTML) published today on the Biomed. Optics Express. In this paper, we reported a comprehensive characterization of our compositional-prior guided DOT reconstruction algorithm (among a few other things - a realistic complex digital breast phantom, how to use tumor priors, how to convert a structure image into tissue compositions, enhancing tumor contrast using mesh refinement, etc). The main findings are quite interesting. We found that using a compositional-prior can reduce HbT estimation error by about 50% (we knew it was better, but now we know by how much). We also found that using correct tumor location as priors can give 1.5x to 4x more contrast to the tumor than without; this translates to roughly 1.5-4 fold reduction in minimum detectable tumor size. The algorithm is also quite robust on false priors: when you tell the algorithm to look for a tumor in the wrong location, it simply rejects it, and falls back to normal tissue priors (because DOT data does not support this idea). Checkout this paper if you are interested. Love to hear what you think.
Also, the developed digital breast phantom - we'd like to call it "DigiBreast" - will be ready for download in the next week or so. We expect this to be a useful benchmark when testing new DOT algorithms, because it is a more realistic test comparing to uniform or segmented breast models used in literature. So, stay tuned.
Over the past month, three open-source software packages received updates. Here is a combined release summary:
The Saving Lives at Birth Program announced this year's awardees at the DevelopmentXChange forum earlier today. My proposal for developing mobile-phone based oximeter was luckily nominated as one of the 24 seed grants. My collaborator, Pat, and I also won another seed grant for mobile phone based thermal imaging. We were thrilled! I can't wait to start the projects! Congrats to all awardees and finalists! and thanks for USAID, Gates Foundation, GCC and other sponsors.
On the flip-side, I almost missed my 5 PM flight back to Boston. The results were announced around 3:10PM, then group photo. I had to beg everyone at the security line to let me go first; fortunately, they were all very kind.
With insufficient funding support, the further maintenance and development of MCX/MMC software are in jeopardy. Please help me for an NIH grant application by sharing your testimonial and/or writing a support letter for MCX/MMC if they have shown value to your research. If successfully funded, the project will deliver many exciting new features, faster and more robust software, helpful training courses and better support to our biophotonics community. Read more details.
Please upgrade to the latest releases of MCX and MMC. A critical bug, affecting all simulations with g=0, was fixed. There are also a number of important new features added. Find out more details from the MCX release notes and MMC Release Notes.
There are still five and half months to go for year 2013, so it is not too late to hand you the final release of Iso2mesh 2013 :) Peter Varga@Charité had contributed a patch to set mesh density for each label with cgalmesher. Also, you can plot mesh slices directly in plotmesh. See Release notes and ChangeLogs for details. To download the latest files, please follow the instructions in this link. Please post your feedback at our mailing list.
I am helping organize the Posters/Demos session for this year's IEEE ICCP conference. The organizing committee is seeking biomedical applications using optics and computational photography techniques. Please help me distribute this call if you think it is relevant. Thanks!
Happy new year! After 20 months of active development, I am proud to announce that a new beta release, Iso2mesh 2013, has finally arrived! This is the first major release towards the 2.0 milestone. Many new features were added. The toolbox grew by 30% in functions and 100% in size. Please read the Release notes and ChangeLogs for the detailed updates. The download files are available for Linux, Windows and Mac.
A new paper entitled "Accelerating mesh-based Monte Carlo method on modern CPU architectures" by Fang & Kaeli was accepted for publication in BOE. In this work, we compared contemporary ray-tracing techniques, aiming for faster and more accurate Monte Carlo modeling on modern CPU processors. The described techniques have been incorporated in the MMC v0.8 and newer releases. Update: download paper PDF.
A new paper entitled "Mesh-based Monte Carlo method in time-domain widefield fluorescence molecular tomography" by Chen, Fang & Intes was accepted for publication in JBO. In this work, the MMC software was extended to support wide-field illumination. In addition, MPI-based parallel computing for distributed systems was also added. The widefield branch of the MMC code is currently developed in the Git repo. This branch will be heading the next milestone release, version 1.0, of MMC.
The original MMC paper finally moved to the top in the BOE monthly Top Download list. It first entered the top-10 download list exactly a year ago. [Update 01/2013:]: The paper moved to the top position again in December 2012.
The full announcement can be found in the mailing list.
After some long procrastination, I finally uploaded the OpenCL version of MCX from a private repo to github. The code was essentially written over 2 years ago, and I have been trying to keep up with the MCX development with occasional updates. It shares a lot codes from the CUDA version of MCX, although it has not been tested as much. I spend some time over the past two days to update it again, and tested with nvidia and ATI cards. The basic features seem to work robustly. Now I think it is time to benefit from public tests and feedbacks. A complete develop log can be found here.
The full announcement can be found in the mailing list.
A new iso2mesh workflow diagram was uploaded to the Sourceforge website. In particular, the updated diagram summarizes the up-to-date additions to iso2mesh towards v2.0 milestone, and the roadmap to further development towards version 3.0 in the future.
Happy new year! I am excited to let you know that two new releases for MCX and MCXLAB had become available recently. In the stable release v0.5.3, we fixed a forward-scattering-bias issue, discovered by David Giraud from BU, and added JSON-formatted input file support. In the beta release v0.7.9, we added additional support for 3D shape files and run-time rasterization. You are recommended to upgrade to the news releases. Your feedback to the new beta release is welcome! See ChangeLogs for more details. Get them from here.
I would also make a preannouncement to the next milestone release of MMC, i.e. version 1.0. We will include wide-field illumination and MPI supports in MMC, as a result of a collaboration between me and Dr. Xavier Intes and Jin Chen from RPI (Jin did excellent job implementing this). Stay tuned if you are interested.
I am looking for a postdoctoral research fellow as part of a new collaboration with Philips Healthcare (Andover, MA) on developing novel multimodality breast imaging systems. The job description is detailed in this link. If you are interested, please email me your research statement, full CV, at least 2 recommendation letters or list of references.
JSONlab is a by-product of my effort in developing a flexible mesh description file format (.jmesh) for my mesh generator. It contains loadjson.m - a JSON->MATLAB decoder, and savejson.m - a MATLAB->JSON encoder. The loadjson.m script was derived from the previous works by several other people. I tested the script with a number of complex JSON inputs, and it worked just fine! If you want to try out this toolbox, you can download it from its subversion.
The GPU-based Monte Carlo simulator, MCX, also received a new update, v0.5.1, last week. It contains a fix to the shared memory buffer and improved flux/fluence normalization. MCXLAB was also updated to 0.5.1. Find more from the Release Notes and download the updated software from here.
We release a new mouse atlas FEM mesh based on the popular Digimouse atlas dataset. The new mesh contains 30% less nodes but with improved mesh quality and much balanced element sizes. Mesh generation script and optical properties of 21 tissue types are provided.
It is my great pleasure to announce the final release of MCX v0.5. The most important addition in this release is MCXLAB, the native MEX version of MCX for MATLAB and GNU Octave. Now you can do everything, including setting up the problem domain, launching simulations and analyzing the results, entirely inside MATLAB/Octave without involving disk files. Here are some screenshots for MCXLAB in MATLAB and Octave. For more details, please read the Full Release Notes and ChangeLogs. Download the software from this page.