Metadata in BioImaging: Management, Organization, and Sharing BioImaging Data • iBiology

00:00:13.11 Hello.
00:00:14.02 My name is Jason Swedlow.
00:00:15.08 I'm from the University of Dundee.
00:00:16.26 I founded and ran the Open Microscopy Environment project,
00:00:19.24 or OME.
00:00:21.10 And I'm here today to tell you about
00:00:23.19 using metadata in bioimaging,
00:00:25.29 how to access that metadata,
00:00:27.23 and also hopefully why it's important.
00:00:30.18 So, in this image,
00:00:33.16 we have lots of what apparently are cells.
00:00:36.07 And, you know,
00:00:38.08 it may be that you can look at this image
00:00:40.08 and start to understand...
00:00:42.07 guess at what is in fact in this image.
00:00:45.00 But to really be able to understand
00:00:46.26 what's going on in this image,
00:00:48.19 there's a large amount of metadata that we need to have access to,
00:00:51.04 that describes actually how the image was recorded;
00:00:54.10 what the different fluorescence channels mean;
00:00:56.25 the various dimensions and scales of the image;
00:01:00.17 how the image is displayed, and this is very important;
00:01:04.14 any associated measurements;
00:01:06.29 and, in fact, if we're making this image publicly available,
00:01:09.05 how to... what is the resource and how to find that image.
00:01:12.16 So, while there's a... you know,
00:01:15.18 this image has a certain amount of aesthetic value
00:01:17.18 -- it might impress you with some kind of view
00:01:21.29 of what's going on in a biological system --
00:01:24.00 it's this metadata that provides
00:01:26.16 the essential background that really...
00:01:29.14 that allows this image to be so powerful,
00:01:32.04 and to be used for a scientific purpose.
00:01:35.02 So, in the next few slides,
00:01:37.00 what we're going to do is to look at different representations of that metadata
00:01:40.06 to try to get some understanding of,
00:01:42.24 really, what this image is trying to tell us.
00:01:46.17 So, this is a screenshot of software that's displaying this image.
00:01:52.02 And you can see several different representations of the metadata
00:01:55.02 over here.
00:01:56.10 So, first of all, you'll recall from
00:01:58.29 Nico Stuurman's talk
00:02:00.26 the power of an image histogram.
00:02:03.02 So, this is a histogram
00:02:04.28 that's displaying the range of intensities in one of the channels,
00:02:07.24 in this case the cyan channel.
00:02:09.18 Note that there's actually four channels in this image.
00:02:12.16 We only have two of them activated.
00:02:15.00 So, we only have the cyan and magenta.
00:02:16.26 But there's two other channels in this image.
00:02:19.11 And that's obviously very important to know,
00:02:20.22 because in fact that tells us that
00:02:23.01 there's more data in this image than we can see.
00:02:25.12 Note that there's different display parameters here.
00:02:28.04 And Nico told you about lookup tables
00:02:30.04 and how the data that's stored on disk
00:02:34.05 is transferred into a represen...
00:02:37.01 a representation for display.
00:02:39.23 So, these are the... how...
00:02:41.23 these sliders tell us how that data
00:02:44.18 is being transformed to this display.
00:02:46.20 Note that that's very important,
00:02:48.12 because in some of the cases
00:02:49.29 -- for example, the cyan channel, here,
00:02:51.22 as well as, you know, well over here --
00:02:53.23 there are examples where the display parameters I've chosen
00:02:56.20 are saturating the DAPI channel.
00:02:59.23 That's a deliberate decision,
00:03:01.26 so that in fact some of these dimmer areas are easily visible.
00:03:04.22 But it's very important to be able to see that,
00:03:06.26 and be able to understand how the display...
00:03:10.08 how the raw data is being transformed for display.
00:03:15.20 Several other metadata concepts.
00:03:17.12 Over on the far side, you can see there's a...
00:03:21.24 there's a slider there.
00:03:22.28 This is actually a Z slider.
00:03:25.08 And so this is in fact...
00:03:27.04 this single two-dimensional plane that you see here
00:03:29.26 is actually part of a three-dimensional data stack.
00:03:31.23 So, this is actually a 3D data cube
00:03:33.23 that just shows us some volume in this biological system.
00:03:36.19 There are display characteristics here.
00:03:39.04 So, I've chosen one set of display characteristics.
00:03:42.08 My colleague Michael has a different set of display characteristics.
00:03:45.13 So, we each are looking at different things in this...
00:03:47.29 in this image.
00:03:49.11 And finally, on the bottom corner there,
00:03:51.22 there's the all-important scale bar that gives us
00:03:57.08 some sense of the physical dimensions of this image.
00:04:00.28 Okay.
00:04:02.10 So, all of this... these are graphical representations of the...
00:04:05.26 of a lot of metadata associated with this image.
00:04:08.11 Now we're going to look at some more numerical values.
00:04:11.08 And again, these are screenshots from software
00:04:14.06 that are showing us the different values.
00:04:15.23 So, note, over here, we have an indication of the size of the image.
00:04:19.26 This is 512 by 512 pixels... in the image.
00:04:26.04 This... each of the pixels is a 16-bit measurement, which is very important.
00:04:29.25 Nico, again, told you about that.
00:04:32.24 The pixel sizes, here,
00:04:35.18 that are mapped into the sample space are 0.42 by 0.42 microns
00:04:39.16 in the image plane,
00:04:41.05 and then 1 micron between each of the optical sections.
00:04:43.23 Note, again, there are several Z-sections.
00:04:46.05 There's only one time point in this image,
00:04:48.19 so this is not a time lapse image.
00:04:51.05 And we have several different fluorescence channels,
00:04:53.28 and the markers that were used to pick up each of those fluorescence channels
00:04:56.24 are displayed.
00:05:00.01 Over on this side, we have
00:05:02.23 an indication of the objective lens
00:05:04.29 that was used to record this data.
00:05:06.17 So, we have the magnification
00:05:08.06 and we have the numerical aperture.
00:05:09.19 That's very important for us to be able to
00:05:12.10 understand the maximum...
00:05:14.08 the maximal resolution that we can achieve with this image.
00:05:17.01 It's an immersion lens,
00:05:19.08 so we have some sense of the...
00:05:22.05 of the immersion medium that's used.
00:05:24.04 In this case, it's an oil lens.
00:05:25.25 Note that there's a mistake here.
00:05:27.20 The metadata is saying that the...
00:05:30.01 that the medium that the biological system is in
00:05:32.24 is actually oil.
00:05:34.05 That's probably unlikely,
00:05:35.18 given the fact that it's a biological system,
00:05:37.07 so it's probably aqueous.
00:05:38.29 There's the refractive index of the... of the...
00:05:41.19 of the immersion medium that's used.
00:05:43.25 For each of the channels,
00:05:45.25 we have metadata around...
00:05:47.23 for example, the excitation and emission wavelengths
00:05:50.08 that were used to image...
00:05:55.18 to collect the data.
00:05:57.07 This is a laser scanning confocal microscope that was used.
00:05:59.25 It has a photomultiplier tube, or a PMT,
00:06:02.19 and it has various settings that are...
00:06:05.24 that are recorded,
00:06:08.10 and important to understand what's going on in the image.
00:06:11.14 Now, each of the representations I've been showing you
00:06:13.17 are actually screenshots from software
00:06:15.18 that's reading and then... and then displaying metadata.
00:06:18.07 In the next slide, I'm gonna show you
00:06:21.02 the raw metadata itself.
00:06:22.10 And so these are values that are
00:06:24.04 taken out of the incoming file,
00:06:26.08 the incoming data file from the microscope.
00:06:29.14 And you'll see that there's a long list of them here.
00:06:31.10 In fact, though, the full list is much, much larger.
00:06:34.06 It goes way... way through the top of the screen
00:06:36.13 and way, way, way through the bottom of the screen.
00:06:38.17 These are just a few of the values that are important.
00:06:40.17 I'm gonna focus actually on three of them,
00:06:43.05 and look at these three values here.
00:06:44.16 Remember, on the previous slide
00:06:46.06 I said that the pixel sizes were
00:06:48.26 0.42 microns by 0.42 microns by 1 micron.
00:06:52.08 And you'll notice that in the raw data,
00:06:54.29 the way this is written out...
00:06:56.18 and you can read this whole line...
00:06:58.00 there's experiment, acquisition block,
00:07:00.10 acquisition mode setup, scaling X #1 equals,
00:07:02.29 and then some number here.
00:07:05.23 And if you do the mental math, you realize that this number
00:07:07.16 is recorded in meters.
00:07:08.25 So, this is in scientific notation.
00:07:11.29 4.15 times 10^-7 meters,
00:07:14.17 or, rounded up, 0.42 microns.
00:07:17.23 We have that for X and Y and Z.
00:07:20.04 Now, obviously what's happening, then,
00:07:22.14 is that the incoming metadata from the...
00:07:24.17 from the data file is being converted or transmitted
00:07:28.20 via the software
00:07:30.07 into a representation that we see on the screen.
00:07:32.16 And so that is very important because
00:07:34.20 there's lots of different metadata here.
00:07:36.02 We saw the pixel sizes in the software...
00:07:38.23 in the screenshot...
00:07:41.12 in the screenshots I showed you.
00:07:42.18 There's other metadata that's equally important.
00:07:44.11 Look down here
00:07:46.04 -- these are the different lasers that are used
00:07:48.12 for the imaging on the confocal microscope.
00:07:51.11 So, you know, metadata
00:07:54.07 is incredibly important to be able to understand
00:07:56.10 what that image was, okay?
00:07:59.05 Now... so, there's lots of different types of metadata.
00:08:01.17 We've been focusing on imaging metadata,
00:08:03.10 but in fact to really understand
00:08:05.25 what that image actually was,
00:08:07.22 and what it's showing us,
00:08:09.10 and what we can do with the scientific result,
00:08:11.13 we probably... there's many different types of metadata we have to handle.
00:08:14.14 So, we have to handle the experimental metadata
00:08:17.02 -- this is the protocols, the probes that we used,
00:08:20.11 the various... any treatments,
00:08:23.00 any genes that were targeted, and so on.
00:08:25.14 And this metadata often sits,
00:08:27.24 you know, in various places in the lab.
00:08:29.19 Maybe in your lab notebook,
00:08:31.24 spreadsheets that you've recorded on your laptop computer.
00:08:34.24 Some labs will have laboratory information systems,
00:08:37.05 so-called LIMS systems,
00:08:39.13 that are recording all this metadata.
00:08:41.13 So, all of this is incredibly important
00:08:43.05 to be able to understand what actually was done to the sample
00:08:47.06 as it was brought to the microscope.
00:08:50.00 As we've seen,
00:08:51.14 there's lots of different parameters associated with the imaging itself.
00:08:53.24 We've kind of gone through that
00:08:56.16 -- the various settings on the microscope parameters, etc.
00:08:59.07 Those are usually written down into the data...
00:09:02.20 into the output data file,
00:09:04.07 but they are usually in proprietary file format.
00:09:07.04 So, if you're using a microscope
00:09:08.24 that's supplied by a commercial vendor,
00:09:10.11 the file formats are usually
00:09:14.28 in some sort of proprietary file format.
00:09:16.25 And finally, there's the analytics
00:09:18.23 that we will perform,
00:09:21.05 the outputs of, for example, finding the cells,
00:09:23.06 and measuring fluorescence intensities,
00:09:24.28 and so on.
00:09:25.26 Now, these values are often written into spreadsheets.
00:09:28.20 You'll... you may have heard of terms like
00:09:32.11 CSV or TSV.
00:09:33.26 CSV stands for comma-separated value,
00:09:37.01 so the individual values in the tables
00:09:39.07 are separated by commas.
00:09:40.17 That's just how they're... they're delimited so it's easy to read.
00:09:45.16 Alternatively, TSV, tab-separated values.
00:09:49.10 And so these are just ways of writing tabular data down.
00:09:52.17 And, you know, these tables
00:09:54.29 might be quite large.
00:09:56.05 There might be 20 or 100 different parameters
00:09:58.14 that we're measuring,
00:10:00.10 represented in columns,
00:10:01.28 and the... and all of the individual measurements,
00:10:06.00 that may be individual cells,
00:10:07.22 are represented as row.
00:10:08.25 So, there could be hundreds or sometimes thousands of rows, or more,
00:10:12.04 in these tables.
00:10:14.01 So, all of these different types of metadata, combined,
00:10:19.02 start to represent the full experiment
00:10:21.04 that we do in our imaging experiment.
00:10:23.11 And you'll probably know that...
00:10:24.28 and you can sort of see here...
00:10:26.12 they're not all in one place.
00:10:28.02 They're in many different places,
00:10:29.20 yet it's absolutely important that we keep track of all this if we're going to,
00:10:33.05 you know, be able achieve the scientific result we want with the...
00:10:37.26 with the imaging experiment we're doing.
00:10:39.28 So, how to do that?
00:10:41.10 So, I'm gonna show you some examples from the project I lead,
00:10:43.20 called OME.
00:10:45.06 That's just one example of many,
00:10:47.05 but it serves to illustrate how to handle the metadata
00:10:51.11 in these experiments.
00:10:53.04 So, we have data...
00:10:54.25 we have to think about the file formats that we're using.
00:10:56.12 We have to think about ways of reading those
00:10:58.20 proprietary file formats.
00:11:00.00 And then we have... need some sort of tool to put everything together,
00:11:02.20 to manage all of that metadata.
00:11:05.25 So, OME-TIFF is an open file format
00:11:08.24 for writing down metadata and binary data.
00:11:11.12 It uses open specifications.
00:11:14.12 So, TIFF is an open specification.
00:11:16.26 And many of you probably know that,
00:11:18.24 you know, most imaging software will open a TIFF file.
00:11:21.18 That's because the TIFF specification is open and available,
00:11:24.16 and anyone can use it.
00:11:26.17 But the OME...
00:11:28.08 so that's a way to write down the pixel data, or the binary data.
00:11:32.04 The OME model presents a way to write down the imaging metadata.
00:11:36.07 And so we can cast that in so-called XML
00:11:39.01 and put that in the header of the TIFF file.
00:11:41.02 It's very flexible.
00:11:42.19 It can handle lots of different imaging modalities
00:11:45.22 and lots of different imaging experiments.
00:11:47.28 We can... we can make public specifications.
00:11:50.17 We can make examples, reference implementations, etc.
00:11:53.22 And we've been running this project, now,
00:11:56.02 for over a decade.
00:11:57.12 So, in fact, we're dedicated...
00:11:58.28 every time there's a new release of OME-TIFF,
00:12:02.22 we make sure that we support all of the previous versions,
00:12:07.22 and make it possible to transfer the different versions
00:12:09.22 and convert between them.
00:12:11.13 So, that's a way to write down the data in a file format.
00:12:14.14 With respect to proprietary file formats,
00:12:18.11 Bio-Formats is a...
00:12:20.05 is a software library that's a plug-in to lots of different software
00:12:23.13 that can read the binary data
00:12:25.18 as well as the binary metadata,
00:12:27.24 and make all of that available to other software.
00:12:30.14 So, we reverse engineer and create a reader
00:12:33.17 for every one of those proprietary file formats.
00:12:36.05 And then... then soft...
00:12:38.13 then other software can use those readers
00:12:41.20 to access each of those different proprietary file formats.
00:12:44.14 There's over 150 different file formats supported,
00:12:46.26 and in fact many different variants of each of the formats.
00:12:49.19 So, in fact, the number of file formats is much larger than that.
00:12:53.14 Lots of open source software tools use Bio-Formats.
00:12:56.16 There are commercial tools that use Bio-Formats as well.
00:12:59.19 Bio-Formats is obviously open source
00:13:02.08 and available for anyone to use.
00:13:04.10 So, this is a way to make the proprietary binary data
00:13:08.08 and the proprietary metadata accessible.
00:13:11.18 And then the question is,
00:13:13.25 how do you combine all that?
00:13:15.01 And that's what we use OMERO for. So, we have the incoming images,
00:13:19.17 we have Bio-Formats to read those images,
00:13:22.11 and OMERO, effectively,
00:13:24.07 is the data management system that puts all of that metadata together,
00:13:26.10 makes it possible to access those data sets,
00:13:29.24 for example through a remote client,
00:13:32.26 for example a web client,
00:13:34.20 or a cluster processing system.
00:13:37.01 And so then you have, effectively,
00:13:39.08 the full solution to having a full data management system.
00:13:42.22 And that brings up the question of
00:13:44.11 why you would want to manage data in the first place.
00:13:46.08 So, you know,
00:13:47.25 what do you get,
00:13:49.14 and what is the scientific value for managing data?
00:13:51.16 So, first of all, there's a question of,
00:13:53.25 you know, just keeping things organized.
00:13:55.14 And, you know, different...
00:13:57.18 you know, different people have different views on organizing stuff.
00:13:59.26 You know, you might think...
00:14:02.13 you might want to be able to find an experiment
00:14:04.07 you did a year ago in the lab,
00:14:05.18 or a colleague in the lab might want to find an experiment
00:14:08.18 that you did,
00:14:10.02 or maybe your PI or your group leader
00:14:13.01 wants to find something you did a couple years ago.
00:14:15.16 So, there's sort of an argument for making this data organized
00:14:18.06 in some sort of way
00:14:19.28 -- for annotating it, for example with the date or the probes you used
00:14:23.23 or some sort of a way of expressing what you've done --
00:14:27.23 that other people can find.
00:14:30.10 Now, as an example of that,
00:14:32.17 here's a screenshot from OMERO.
00:14:34.11 And here we have data that...
00:14:36.14 where, you know, we have the data organized
00:14:38.08 in terms of the different cell cycle stages that we were looking at.
00:14:40.14 Obviously, we have the images.
00:14:42.11 We have ways that you, you know...
00:14:44.17 that you can view the...
00:14:47.14 view the data in a web browser,
00:14:49.25 change the display settings, etc.
00:14:52.02 You can obviously also see all of the...
00:14:54.06 all of the acquisition metadata,
00:14:56.17 analytic metadata, and so on.
00:14:58.22 So, that's putting data under management.
00:15:01.00 That can be nice.
00:15:02.06 It's not, frankly, all that compelling.
00:15:05.05 Keeping things organized and clean might not be
00:15:08.19 the strongest priority for everyone.
00:15:10.18 Probably a stronger argument is that there's...
00:15:13.14 if you have data under management,
00:15:15.13 it becomes much easier to run the data
00:15:18.05 through large processing workflows or batch processing,
00:15:24.01 thus saving yourself some time.
00:15:25.14 So, rather than processing a single image at once,
00:15:28.01 you can process a large number of images at once,
00:15:30.14 and on these kinds of manage...
00:15:32.17 on these data management systems.
00:15:34.12 An example of that, again with OMERO,
00:15:36.10 is that we can take many different third-party tools
00:15:39.04 -- some of them open source, some of the commercial --
00:15:41.17 bolt them onto OMERO,
00:15:43.11 and then make them...
00:15:45.05 basically run them against large collections of images,
00:15:47.20 and thus effectively automate the image processing tools.
00:15:52.14 We can also make that data available for...
00:15:54.28 through a web browser,
00:15:56.02 applications like ImageJ, and so on.
00:15:59.00 So, that might be another reason to make these...
00:16:01.23 with these data sets...
00:16:03.12 to put these data sets under management.
00:16:05.01 And to, you know, go through the effort
00:16:08.04 of keeping track of your data...
00:16:09.17 then it becomes a lot more...
00:16:11.08 a lot easier to automate these systems.
00:16:12.23 An example of that.
00:16:13.29 So, this is... these are... this is an experiment that we've been doing at Dundee
00:16:17.05 using the cell painting protocol
00:16:19.29 that was developed by Anne Carpenter's group at the Broad Institute.
00:16:22.16 These are human iPS cells,
00:16:24.08 where we're going through a cell painting protocol.
00:16:26.26 Then we run...
00:16:28.12 basically, put all of them in OMERO,
00:16:30.01 run CellProfiler through them
00:16:32.02 -- and CellProfiler is an application
00:16:34.04 that can identify the boundaries of the nuclei,
00:16:36.19 the cytoplasm of each of the cells --
00:16:38.26 calculate a whole list of features for every one of those cells,
00:16:41.13 every one of those images... all of that...
00:16:44.25 all of the imaging data and all of the analytic outputs
00:16:47.18 are all in one system,
00:16:48.26 which we could then use for further processing and calculations.
00:16:53.16 So, by keeping things together,
00:16:55.26 by keeping them well managed,
00:16:57.11 we can run this in an automated fashion.
00:16:59.05 And also it makes it easier to do downstream calculations.
00:17:03.07 So, the third reason to think about [managing] data
00:17:05.07 is actually to share data
00:17:07.14 and to make data available in various ways.
00:17:09.29 And so, an example of that is
00:17:11.29 an application in OMERO that we call OMERO.figure,
00:17:14.06 and this is for building figures from your...
00:17:17.07 from your data sets.
00:17:19.22 And so, many of you will make figures for your papers,
00:17:21.20 and you usually take screenshots
00:17:24.04 and put them into some sort of graphical tool.
00:17:25.17 And that's fine,
00:17:27.28 but then the data that you're actually sending for publication
00:17:30.02 is completely detached from the original data.
00:17:32.14 In this case, we have a web application
00:17:34.20 called OMERO.figure,
00:17:36.11 where every one of these viewers
00:17:38.11 is a portal into an OMERO database.
00:17:40.17 And we can actually, then,
00:17:42.18 make it so that the figure that we're generating,
00:17:44.22 and ultimately submitting, is always...
00:17:47.24 is always tied to the original data.
00:17:49.14 So, that the idea of sharing and presenting data...
00:17:54.23 you know, if you do that,
00:17:56.02 you're maintaining the connection to the original data.
00:17:59.24 Another example is...
00:18:01.24 here's a paper that we published from my lab
00:18:04.13 in the last couple of years.
00:18:05.18 And, you know, there's the paper.
00:18:06.28 The DOI is down there at the bottom.
00:18:08.18 If you scroll down to the HTML representation
00:18:11.07 of that paper online,
00:18:12.18 you'll come to a data availability section
00:18:14.24 that has a DOI.
00:18:16.17 And that DOI resolves to this page.
00:18:19.08 And this is the data I was showing you before.
00:18:21.18 Here are all of the data sets
00:18:23.20 associated with the different figures in that paper.
00:18:25.20 We have all the imaging data.
00:18:27.14 We have all the representations...
00:18:30.05 the thresholding and the display parameters that we used.
00:18:32.19 We have the metadata that was associated with the acquisition,
00:18:36.05 the analytics, and so on.
00:18:37.07 We also have all the figures associated with the...
00:18:39.14 coming from OMERO.figure all available.
00:18:41.28 And so that makes... that makes it possible, effectively,
00:18:44.19 to publish these data sets online,
00:18:46.16 make them available for others to examine,
00:18:50.05 to check whether what we were doing was reasonable,
00:18:52.24 as well as to reuse.
00:18:55.01 And that brings up this issue of...
00:18:57.00 this idea of making data FAIR.
00:18:59.04 So, the acronym is FAIR.
00:19:01.23 And the words behind the acronym are
00:19:04.01 findable, accessible, interoperable, and reusable.
00:19:07.08 And so the idea -- and this... this idea...
00:19:09.14 this concept was published by Wilkinson and colleagues a couple years ago --
00:19:13.25 is that these... that data sets associated with scientific publications
00:19:18.02 -- the underlying data sets --
00:19:20.17 should be publicly available.
00:19:22.20 And what... and they actually define what these different words mean.
00:19:25.05 That they should be findable,
00:19:27.02 and so there are...
00:19:28.28 there are ways of effectively using DOIs or other identifiers,
00:19:31.23 being able to find the data.
00:19:34.22 The metadata associated with those data sets
00:19:38.10 is accessible through a UI,
00:19:40.06 in some of the ways that we've been seeing,
00:19:42.01 before but also programmatically,
00:19:43.20 so there's an application program programming interface, or API.
00:19:46.29 What that means is if someone wants to
00:19:49.28 write some sort of analysis tool against this data set,
00:19:51.23 it's possible.
00:19:54.06 The data sets are
00:19:57.13 annotated with known and defined vocabularies.
00:20:00.05 And most importantly,
00:20:01.26 the reuse of that data is actually known and defined.
00:20:04.10 So, the data is sitting under a license.
00:20:06.23 In the example I showed you before,
00:20:08.16 we're using the Creator Commons CC-BY license.
00:20:12.28 And what that means is that anybody can take the data
00:20:14.29 and use it for whatever they want.
00:20:16.20 If they publish a result, they just have to cite us.
00:20:19.10 And that's... that's kind of a common concept in science.
00:20:22.05 So, you know, using the tools,
00:20:25.02 and working through the metadata,
00:20:26.17 and going through the management process, etc,
00:20:29.01 it becomes quite possible -- in fact, even routine --
00:20:31.26 to be able to make these image data sets that we're collecting in our lab
00:20:35.07 truly FAIR.
00:20:37.13 And so that's an important reason
00:20:39.18 to run these management systems.
00:20:44.01 And then you can sort of take that idea to the next level.
00:20:47.00 And so we've been doing that
00:20:48.20 using Bio-Formats and OMERO
00:20:50.01 to build an application
00:20:52.06 called the Image Data Resource, or IDR.
00:20:54.17 And so this is publishing imaging data
00:20:56.24 from many different studies online,
00:20:59.14 and collecting them into a single system,
00:21:02.08 with all of the experimental metadata,
00:21:04.12 the annotations, the analytic results
00:21:05.29 -- all the different metadata that we talked about before --
00:21:08.14 and making all of that data browsable
00:21:11.13 in, for example, a web browser,
00:21:13.01 but also downloadable
00:21:15.18 and available for reanalysis in some sort of cloud environment.
00:21:18.26 And so the URL is there,
00:21:20.05 if you want to see more.
00:21:21.22 But, you know, that's another...
00:21:23.05 kind of the next level of the idea of making...
00:21:25.16 the value of considering and working
00:21:28.28 with all of this metadata
00:21:30.20 to make it possible to...
00:21:32.14 to generate an awful lot of value from these published data sets.
00:21:36.12 So, just to sum up,
00:21:38.11 I hope I've convinced you... you know... you know...
00:21:41.12 shown you examples of how you can access this metadata
00:21:43.06 that's coming in on the... on the data...
00:21:45.28 on the imaging data sets that you're collecting.
00:21:48.04 There are various open tools and formats
00:21:51.11 for accessing that data.
00:21:53.01 And there are also, you know,
00:21:54.26 ways that you can organize and use that metadata,
00:21:57.02 hopefully in your lab,
00:21:59.08 for your benefit.
00:22:01.00 But probably most importantly,
00:22:03.00 why should you worry about this metadata,
00:22:04.28 and why should you actually think about managing this metadata?
00:22:09.02 So, it makes it possible, now,
00:22:11.08 to routinely make imaging data FAIR,
00:22:13.06 and that's prob...
00:22:14.20 that's very important, I think, also, just, you know,
00:22:17.06 for science,
00:22:18.18 but also our funding agencies and our journals certainly want to see that,
00:22:21.00 so that's a very important issue.
00:22:22.12 And then it allows... it makes it possible
00:22:24.18 to publish these imaging data sets routinely
00:22:27.24 for reuse by the scientific community.
00:22:31.16 So, I hope that's useful.
00:22:33.01 I really appreciate the time.
00:22:34.10 And thank you very much for listening.

This Talk

Speaker: Jason Swedlow

Audience:

Educators of Adv. Undergrad / Grad
Researcher

Recorded: November 2018

Talk Overview

In order to understand an image of a biological sample and what it represents, one needs to understand its metadata. Metadata is the information behind the image that shows the experimental procedure, image acquisition settings, and the analysis performed on the data in order to obtain the represented image. Dr. Jason Swedlow explains what constitutes image metadata, and provides examples on how to catalog, organize, analyze, and share the metadata of biological images.

Concepts:
Types of metadata, OME, BioFormats, Omero, Data management, CellProfiler

Questions

What is image metadata and why is it important?
What’s the difference between proprietary and open file formats?
What’s the value of using data management tools?

Answers

Speaker Bio

Jason Swedlow

Jason Swedlow

Dr. Jason Swedlow is a professor at the University of Dundee in Scotland, and co-founder of the Open Microscopy Environment (OME) project. Swedlow obtained his bachelors in Chemistry from Brandeis University (1982), and completed his PhD in Biophysics at the University of California, San Francisco (1994). He continued his scientific training as a postdoctoral fellow… Continue Reading

Leave a Reply Cancel reply