More Big Data to Consider: Bioimage Informatics

I’m not sure any more when I signed up for complementary copies of Nature Methods, but just like clockwork my copy arrives each month. If you’d like to get it too, you can apply for a subscription here (Firefox seems to work better than IE, btw). This month’s issue particularly interested me because it contains a focus on Bioimage Informatics. The focus appears to be free to read online.

I found the focus just after having read the Science News article “Blast Injuries Linked to Neurodegeneration in Veterans” by Greg Miller. In Greg’s piece there is a description of a distinctive neuropathology that has been seen in athletes and military veterans who had incurred head injuries. This same distinctive pattern is seen in a mouse model of blast injury & the image of the tangles of tau protein shown in the article struck me as so interesting that I told my husband about it over dinner one night, so I already had bioimages on my mind. I am also always interested in the field of bioinformatics, both personally and as a member of the OpenHelix team.

The commentaries, in the order that they were printed, were what I read initially. The first commentary is by Gene Myers, who was also involved in early genome bioinformatics, and it provided a very interesting perspective on both the current state of bioimage informatics and on the historic use of bioimages in systems genetics.The following quote made me grin:

The field is still in its early days, and there is no such thing as a typical bioimage informatician: they are either computer vision experts looking for new problems, classic sequence-based bioinformaticians looking for the new thing or physicists and molecular biologists whose experiments require them to bite the informatics bullet. … From my perspective, it is very reminiscent of the state of bioinformatics in the early 1980s: the exciting, somewhat chaotic free-for-all that is potentially the birth of something new.”

And the following paragraph stressing the importance of “due diligence of pilot studies” and “optimized protocols” reminded me of my days setting up a Biocore facility without enough funding for either sufficient pilot studies or optimization, which ultimately doomed the utility of the machine to my advisor and department alike. This commentary set the stage well for the rest of the articles. The other commentaries included a description of the difference in goals of the computer vision field and the bioimage informatics field, a plea for usability to be built into bioimaging software, and a historical commentary on the 25 years of NIH Image, now ImageJ.

The usability article sounded many many of the same cries that we make here at OpenHelix – if you want to have usable bioscience software that IS in fact USED, at a minimum you must 1) have funding and a mandate to maintain it over the long run, 2) have motivated developers that are responsive to their users needs and feedback, including fixing bugs and 3) (last but absolutely not least) you must provide awareness and training on your software. And in my opinion, any old training WON”T due – it has to be high quality, up-to-date, and easier to use & absorb than your average dry documentation on programming your VCR clock (OK, I’m dating myself there, but you KNOW what I mean…) I like their suggestion that funding agencies request descriptions of how the software be maintained and documented, and to be prepared to provide funding not just for development, but also for maintenance. (Why reinvent the wheel over & over, just to let each one go flat with disrepair?)

There were also reports on specific software, such as OMERO.searcher, SimuCell, PhenoRipper, Fiji, BioImageXD, and Icy, as well as on the Broad Bioimage Benchmark Collection (BBBC), a collection of microscopy image sets available for the testing and validation of new image-analysis algorithms.

The focus then concludes with a great review of bioimaging software tools, with the goal of providing a “how to” summary of using open-source imaging software for every stage of bioimage informatics. It begins with a discussion of data aquisition & continues through data storage and workflow systems. I might tweek figure one just a bit, but it does visualize that today software is required at every stage of image analysis – from automated image attainment to image retrieval and analysis. The authors also touch on the importance of image annotation and controlled vocabularies, or ontologies. Table 1 provides a nice resource listing including software names, primary function and URL – I have some new resources to check out now! :)

Overall, I’d suggest this focus on bioimage informatics to any life scientist, whether you are analyzing images today or not – I think it is provides a glimpse into an up&coming, exciting field.

Quick Links:

Broad  Broad Bioimage Benchmark Collection (BBBC):







Reference List:
Greg Miller (2012). Blast Injuries Linked to Neurodegeneration in Veterans Science, 336 (6083), 790-791 DOI: 10.1126/science.336.6083.790

Gene Myers (2012). Why bioimage informatics matters Nature Methods, 9, 659-660 DOI: 10.1038/nmeth.2024

Anne E Carpenter, Lee Kamentsky, & Kevin W Eliceiri (2012). A call for bioimaging software usability Nature Methods 9, 9, 666-670 DOI: 10.1038/nmeth.2073

Kevin W Eliceiri, Michael R Berthold, Ilya G Goldberg, Luis Ibáñez, B S Manjunath, Maryann E Martone, Robert F Murphy, Hanchuan Peng, Anne L Plant, Badrinath Roysam, Nico Stuurmann, Jason R Swedlow, Pavel Tomancak, & & Anne E Carpenter (2012). Biological imaging software tools Nature Methods, 9, 697-710 DOI: 10.1038/nmeth.2084