Hello all,
I collected marine diatoms from algae in natural open beach ponds. It was easy, and the samples are being processed. However, what about the sand? How problematic would it be?
So, having consulted some sources and the recently described procedures by forum members @Charles and @MicroBob, I dug some handfuls of wet sand from the open seawater pond, added fresh water, swirled the mixture in a jar, waited a few seconds until the coarse sand settled, decanted the top layer to another jar, 2x repetitions.
The mixture (even before it further settling) shows several more types of alive diatoms, not too much debris and slit around (single images with 40X0.75 Neofluar phase contast, resized & cropped, below). Possibly, these diatoms are not benthic, since recent strong sea side winds caused waves that could cover epiphytic diatoms with sand, and that is what I sampled. Yet the population distribution seems to differ somewhat from that of the algae-adherent diatoms.
Bottom line - a worthwhile experiment. Cleaning won't be easy, but the alive diatoms from sand are beautiful.
Now an intriguing question: Does anyone know about a software that can automatically sort diatom images (alive, frustules, valves, anything obtained with OPTICAL microscope) and identify at least their genus (a species would be too much to ask)? Several years ago there was some development going on, but is there a product?
Easier to find them (diatoms) - what about automatic identification?
Easier to find them (diatoms) - what about automatic identification?
- Attachments
-
- Sand diatom.jpg (113.08 KiB) Viewed 3589 times
-
- IMG_1191.JPG (109.56 KiB) Viewed 3608 times
-
- IMG_1190.JPG (122.24 KiB) Viewed 3608 times
Re: Easier to find them (diatoms) - what about automatic identification?
Hi Doron,
nice images and diatom species!
The question of the automated identification is interesting. I have no knowledge of such a software, but that doesn't mean much. My guess is that it would be very difficult to get a really perfect identification, and for scientists anything less would not be acceptable. So they sit an count and make the errors on their own.
There is a free scientific image editor an analyse software : FIJII (based on ImageJ)
May be there is a plugin to identify diatoms. I think it is very much work to build the necessary database.
Precise and up to date literature on diatoms is very expensive and takes up quite a bit of room. A member of our group had to have structural calculations done to be able to store his books without collapsing the building he lives in.
There is a quick graphical diatom overview on our website:http://www.mikrohamburg.de/Hustedt_Netz ... Penna.html
Bob
nice images and diatom species!
The question of the automated identification is interesting. I have no knowledge of such a software, but that doesn't mean much. My guess is that it would be very difficult to get a really perfect identification, and for scientists anything less would not be acceptable. So they sit an count and make the errors on their own.
There is a free scientific image editor an analyse software : FIJII (based on ImageJ)
May be there is a plugin to identify diatoms. I think it is very much work to build the necessary database.
Precise and up to date literature on diatoms is very expensive and takes up quite a bit of room. A member of our group had to have structural calculations done to be able to store his books without collapsing the building he lives in.
There is a quick graphical diatom overview on our website:http://www.mikrohamburg.de/Hustedt_Netz ... Penna.html
Bob
Re: Easier to find them (diatoms) - what about automatic identification?
An intriguing question indeed !Hobbyst46 wrote:Now an intriguing question: Does anyone know about a software that can automatically sort diatom images (alive, frustules, valves, anything obtained with OPTICAL microscope) and identify at least their genus (a species would be too much to ask)? Several years ago there was some development going on, but is there a product?
I tried the following phrase [without quotes] in a google search:
" automated system identify diatom "
... and was surprised to find serveral promising results.
MichaelG.
Too many 'projects'
Re: Easier to find them (diatoms) - what about automatic identification?
Thanks Bob for the comments and link.MicroBob wrote: There is a free scientific image editor an analyse software : FIJII (based on ImageJ)
May be there is a plugin to identify diatoms. I think it is very much work to build the necessary database.
Precise and up to date literature on diatoms is very expensive and takes up quite a bit of room. A member of our group had to have structural calculations done to be able to store his books without collapsing the building he lives in.
There is a quick graphical diatom overview on our website:http://www.mikrohamburg.de/Hustedt_Netz ... Penna.html
Bob
The free plugin for ImageJ is named "PlanktoVision" and dates back to 2013. Probably integrates with Fiji as well. Although I have some experience with ImageJ, I did not try to use the plugin, since it seems to be an analysis tool but not an identification tool. But I might be wrong!
Marine Biology experts told me that a single optical microscopy image might help to identify the genus, at best. To identify the species, more info and EM images are needed. I will appreciate knowing the genus anyhow, since now I am clueless about the names of these creatures.
p.s. I got capsuled digestion enzymes from an organic market. It they do the job, I will report.
Re: Easier to find them (diatoms) - what about automatic identification?
Thanks MichaelG.MichaelG. wrote:Hobbyst46 wrote: I tried the following phrase [without quotes] in a google search:
" automated system identify diatom "
... and was surprised to find serveral promising results.
MichaelG.
I searched with the same phrase, and there are indeed a few R&D projects on diatom automatic identification. Modern works rely on image analysis, and I agree that they show some promising results. They need a large (tens of samples) "learning set" - known and identified diatom images, that make an initial database, to develop the algorithms. Not as yet user-friendly, push-button automatic identification.