We are a research pathology core facility and we plan to acquire a slide scanner that performs brightfield and fluorescence whole slide imaging on tissues sections. It seems like this topic has not been discussed in depth recently, and some of the instruments we are considering are relatively new and have not been discussed much here since the last post I could find on this topic (Looking for feedback on fluorescence slide scanners).
We need a scanner that can do both brightfield (around 90-95% of our usage; brightfield should include capacity for polarized light) and fluorescence (5-10%; DAPI, FITC, TRITC, Cy5, Cy7), is high throughput (planning to scan up to 200 or 300 slides a day), read barcodes on slide labels, and is reasonably automated and easy to use. It should accept regular (1x3â) and oversized (2x3â) slides. We need to image slides with 20x or 40x dry objective. Z stack capability may be used rarely but is not a high priority. The scanner should generate images in a format that works well with our main image analysis software: Halo and Qupath.
We will also want to run an image management system/server to store images and associate metadata and provide efficient access to the images for our pathologists and users. We are considering third party image management systems, but if the vendor of the scanner has a good image management system we will consider it.
At this time we have narrowed down our options to the Zeiss Axioscan 7 and Olympus VS200, both released in the last 2 years.
I would like to hear from users from both systems, especially in a core facility environment, what you experience had been. Also I would be interested to know if anyone would recommend other systems given our requirements.
Hello and welcome to the forum.
I am a clinical pathologist and have used several commercial slide scanners - and I have built my own - high throughput and roboticised standard scopes. However I have not used the particular models you mentioned so I can only give some general pointers for you to consider and perhaps look out for in the particular scanners you mentioned.
The first is that all the scanners I have used can do all the bar code, archiving etc. - that is just a simple matter of software and is irrelevant to the optical performance. The company reps can tell you whatâs available.
Regarding optics, all the âhigh throughputâ scanners I have used - no matter how expensive - give awful quality images at the highest magnifications due to lack of proper condenser optics and Kohler illumination. They may be useful for some purposes but I wouldnât use them for making diagnoses or for high resolution 3D deconvolution, etc. Also, I have not yet come across a high throughput scanner with any useful polarisation features (âusefulâ here means you get so see how the birefringence changes with angle instead of just a single 90 degree crossed polars snapshot, and also have the option of a retarder for positive/negative birefringence assessment).
The roboticised standard scopes are much better at this with higher quality images at high magnification and very good for 3D deconvolution. The image quality of a full 3D scan at x40 or higher may be suitable for diagnostic work but would take ages to acquire and many GB of storage per slide to avoid corrupting the spectral quality with lossy compression.
In both cases (both types of machine) the colour gamut will be lower than simply observing by eye. Others have commented on the problems of restricted spectral resolution for certain fluorescence experiments but I have no personal experience of fluorescence with these machines (although I can fully believe it).
The other âtrapâ to watch out for is restrictions in what the software allows you to do - donât assume that if a feature is not there it will be added for you, at all let alone free of charge.
The best thing to do is âtry before you buyâ from both systems - if the output is OK for your needs then the specifics are irrelevant on a practical level - just make sure when you assess âOK for your needsâ it is the actual people who will be using the images who are the ones that will be assessing the quality because they all give very, very pretty pictures.
In case anyone else is reading up on this, the above comment is VERY important when avoiding 3DHISTECH scanners. The scanners were great, the MRXS format is and was awful and functions to prevent use of the slides outside of very limited software options.
In case you havenât already, the confocal listserv doesnât exclusively discuss confocal topics, and there are a great many core facilities present there that might be able to give firsthand opinions about how the systems have performed over time and what issues they have run into.
Thank you @P_Tadrous. I have worked with images from scanners from a few of the major vendors for the past 15 years, and occasionally some from roboticised standard scopes, but the bulk of my work has been and still remains looking at glass slides, so I am familiar with some of the pitfalls of whole slide imaging, and you provide good advice.
We will demo the systems to see how they work in our hands, with our slides, and our analysis workflow, but trying them for a short period of time is of course not the same as long term use, and that is why Iâm also seeking input from users of these particular systems.
@Research_Associate MRXS is one of the most common image format we are given to work with, and we are well aware of this issue. Brightfield MRXS images generally work fine in Qupath, but not fluorescence. And they are not supported at all by OMERO.
We recently purchased a Zeiss Axioscan 7 system to replace a 13-year-old Hamamatsu Nanozoomer HT (300 slide capacity, multi-band fluorescence capable). I am still learning how best to utilize the Axioscan and was disappointed in its inability to easily create brightfield/fluorescence overlay scans. This is due to the use of different dedicated cams for BF and FL which wasnât an issue in the older Hamamatsu. We have several neuro groups that like to stain their fluorescent reporter mice (GFP/RFP) for c-Fos using DAB necessitating the scanning of slides in two modes. On the older Hamamatsu, it was just a box you checked. On the Axioscan, you have to use the FL cam and use a dedicated blue, green, and red filter cube (they can be your FITC, DAPI, and RFP cubes) with the transmitted light to create a pseudo-RGB BF image that is perfectly overlaid with the fluorescence signal. This adds considerable time to the image capture (three filter cube changes for RGB plus whatever you need for your fluorescence reporters).
Iâm quite satisfied with the image quality @ 20x but make sure you are using #1.5 coverslips. Higher magnification imaging of BF slides stained for H&E, trichrome, etc. looks good but if you want to image fluorescently stained material that is mounted in lower RI media, you might have to open up the box to adjust the correction collar on the 40x to get the best image quality. I also must do a lot more testing of the spectral bleed-through that is possible with their 110 and 112 filter sets. I had to resort to imaging some slides stained with CF555 with a single band filter cube instead of the multi-band due to bleedthrough and, again, capturing a large ROI with multiple filter changes can take a long time. I have always charged clients on a âper-slideâ basis but now see why some/many cores charge users by the hour. It seems, to me, that once you set up a list of imaging protocols for particular tissue types/stains, the imaging process becomes much more streamlined and âhands-offâ. The problems arise when you are trying to automatically detect tissue such as adipose that has virtually no contrast. If the tissue isnât easily segmented by the overview image, you must resort to more manual ROI identification or an additional pre-scan using either Zeissâ TIE contrast or a fluorescence channel. Of course, this adds to the time required to scan the slide.
I have found the Zeiss support team to be very knowledgeable and accessible and with remote assistance, most issues can be resolved quickly. It is my understanding that the .czi format is compatible with several software packages such as Visopharm or QuPath but I have not tested that myself.
I know this is an old post, but I wanted to add a few more tidbits on the Zeiss Axioscan 7 in case someone finds this forum post while looking for information on the scanner.
If you will be doing brightfield scanning and need to overlay those scans with fluorescence - consider another scanner. As I mention in my June 30 reply, it IS possible to collect this data using the mono-camera and shuffling filter cubes around BUT it takes an inordinate amount of time to collect the data. A slide containing an array of mouse brain coronal sections imaged in this way can take over 3 hours to scan - for one slide. The Hamamatsu NanoZoomer we have can do that in 30-40 min. Zeiss has a tool called ZEN Connect that should allow you to manually create an overlay of a true BF scan and the FL scan but I have not had a chance to evaluate how well it works when trying to overlay very large slide scan data.
The automatic tissue recognition ability of the system leaves room for improvement. Very weakly stained tissue such as adipose or fluorescently-stained tissue require either manual involvement or âpre-scanâ imaging to identify ROIs for subsequent high-resolution scans. For tissue such as adipose, the smaller overview camera and illumination system results (on our system) in a brightness gradient from one end of the slide image to the other. Trying to adjust the threshold and offset of the overview image so the software can automatically detect faint tissue and not background is difficult. Note: it is possible to draw a line around the tissue using a dark colored pen and let the software find that line but this requires you doing some additional work. Pre-scan imaging of fluorescently labeled tissue is useful but bear in mind, many pap-pens used have strong autofluorescence properties and will confuse the tissue-recognition step ensuring you will need to manually draw ROIs around tissue. Some of these issues could probably be avoided if Zeiss added an oblique illumination method similar to their new AI Sample Finder that can be equipped on their inverted platforms.
Converted 4-channel image files open as multichannel fluorescence with proper channel assignment in QuPath, can be manipulated using imagecombiner/Warpy, etc. I hope to use these tools to liberate us from the mrxs straightjacket so we can use our existing 3DHistech scanner for several current projects.
I do not have much experience using mrxs outside the vendorâs proprietary viewer but the ability to generate ome-tiff seems like a promising route to make use of legacy hardware. I am, however, apprehensive that future changes in the mrxs format may break this workflow. This would be relevant for anyone considering acquiring a ânewâ 3DHistech scanner.
Very important I forgot to mention this only works if the original mrxs file is created as png or bmp. Does not work if the mrxs file is jpg or if the mrxs jpg is converted to png or bmp. I wonder if this is due to ânon-standardâ jpg compression used by 3DH. In any case, I am able to use png for this purpose.