The Daily Scan

IT questions: storage + bandwidth

There are two big IT questions people ask about Digital Pathology: how much storage will I need, and how much bandwidth?  And how much will they cost?  Let's figure this out.

(In the discussion below, the key assumptions are highlighted in gold.)

Storage

The main concerns about storage are capacity, cost, and backup.

raiders of the lost slide

Capacity + Cost

• Typical slide = 250MB
  • Digital slide = 20mm x 15mm @ 20X (.5 micron/pixel)
  • = 40,000 x 30,000 pixels = 1.2 Gpixel
  • x 3 bytes/pixel = 3.6GB
  • / 20:1 compression (JPEG2000) = 180MB

• Typical lab: One year = 5TB = $5,000
  • (100 slides / day) x (200 days / year) = 20,000 slides
  • = 5TB (1TB can hold 4,000 slides)
  • x $1,000 / TB ($500 disk + $500 server, etc) = $5,000

Note that number of slides per day is the key variable.

• Typical server: Up to 80TB = 16 years
  • 2 controllers, each with 4 racks, each with up to 12 drives
  • With RAID 5, capacity = 2 x 4 x 11 (12-1) drives = 88TB
  • With RAID 6, capacity = 2 x 4 x 10 (12-2) drives = 80TB

This example drawn from current HP models, Dell and others are similar.  It is a bad idea to buy too much storage too soon; the capacity of disk drives keeps going up and the cost keeps coming down.  I usually recommend buying a year's worth of storage each year.

Backup

As can be seen from the capacity and cost data above, there is no reason not to keep all needed data online as long as it is needed.  Any kind of archive solution which keeps some data online and takes others offline to tape, etc. will be more complicated and more expensive, with no benefit.

There are two reasons to backup data, to protect against device failure (e.g. disk drive fails), and to protect against catastrophic failure (e.g. computer room flooded).  With RAID systems can easily recover from device failure with no further backup provisions.  (Given the relatively small incremental cost, I recommend RAID 6 for most situations.)  Offsite backup to protect against catastrophe is often best accomplished with disk-to-disk copies to removeable drives.  Because of the typical data volumes network backup solutions are usually not practical.

Bandwidth

The main concerns about bandwidth are capacity and cost.

high bandwidth (high capacity, lower speed)

low bandwidth (low capacity, higher speed)

Capacity + Cost

• Typical user = 20KB/s
  • Computer screen = 1280 x 1024 pixels = 1.3Mpixel
  • x 3 bytes / pixel = 4MB
  • / 20:1 compression (JPEG2000) = 80KB / screen
  • / screen every 4 seconds = 20KB/s


• Typical lab: 3 T1 lines = $18,000 / year
  • 20 concurrent users x 20KB/s = 400KB/s
  • / 200KB/s (T1 line = 1.5Mb/s = 200KB/s) = 2 T1 lines
  • + 50% (safety utilization) = 3 T1 lines
  • x $500 / month = $1,500 / month

Note that number of current users is the key variable.

• Additional capacity is less expensive
  • DS3 (aka OC3 aka T3) = 24 x T1 = 5MB/s
  • = 125 concurrent users = $6,000 / month

Each individual situation is different, but hopefully with these examples you could compute storage and bandwidth requirements for your circumstances, and approximate the cost.

Your questions and comments are welcome!

Cancer superstar Judah Folkman dies

The Boston Business Journal reports on the passing of Judah Folkman, the "father" of angiogenesis.  Dr. Folkman was the first to discover that tumors depend on blood vessels for survival and growth, and that tumors can be treated by cutting off their blood supply.

Additionally, analysis of blood vessel density can detect cancers; as a tumor forms the blood vessels in the vicinity will grow, providing clues to the tumor's existence.

Dr. Folkman had authored 389 original peer-reviewed papers.  He will be missed.

Aperio Receives FDA Clearance for Manual Read of HER2 Digital Slides

When I resumed blogging I promised I wasn't going to recap Aperio press releases, but this one is pretty important: Aperio Receives FDA Clearance for Manual Read of Digital Her2 Slides.

As the release explains, Aperio is the first company to receive FDA clearance for the manual read of digital slides for clinical use in pathology.  This is an important benchmark; as adoption of digital pathology for clinical applications continues labs and clinicians want the assurance of safety and efficacy implied by FDA clearances.  (Labs are regulated by CLIA, not by FDA, but even so FDA clearances are very important.)  Manual reads from computer monitors are becoming increasingly common as Pathologists use digital pathology systems for frozen sections, tumor boards, secondary consults, virtual IHC, and other key applications.

There are four key technologies responsible for delivering digital slides with the quality and speed needed for manual reads.  First is the scanner, the device which digitizes entire microscope slides in minutes and produces a seamless high-quality digital image.  Second is JPEG2000 compression technology, which enables 5GB of image data to be stored in a 250MB file, and makes network access to digital images 20 times faster.  Third is the incremental transmission technology which allows digital slides to be opened instantly and panned and zoomed rapidly, even across the Internet.  And fourth is monitor technology; without today's big high-quality LCD monitors manual reads from digital slides would not be possible.

I plan to post about these technologies in more detail; please stay tuned!

Histology, circa 1885

From the back page of The Scientist, January 2008:

(click to enlarge)

Today they'd use a ScanScope, and store the digital slide.  Of course the wooden box is pretty cool :)

View digital slides on your iPhone!

If you have an iPhone - or a spiffy new iPod Touch - you can now view digital slides on it!

When you view digital slides on a website using Aperio software, a little program called WebViewer is displayed right inside your browser.  You have probably seen this on many different sites, it looks like this:

This WebViewer program is written in a language called Flash.  And the iPhone does not support Flash.  Boo.  Maybe someday the iPhone will support Flash, and everything will just work. (That day could be as soon as next week; Apple has a lot of announcements planned at Macworld.)  However, that day could be never, and who wants to wait that long?

So we made it possible to display digital slides without Flash, using a different language called JavaScript.  The iPhone does support JavaScript.  Yay.  The JavaScript viewer works almost like the Flash viewer, but there are some differences.  Mostly these differences are to make using the viewer easier on an iPhone.

When you’re viewing on an iPhone, you can zoom in.  I’m sure you’ve tried this (or seen it), you pinch your fingers apart, and poof, the screen zooms.  Very cool.  Once you’ve zoomed in, you can pan around.  You just brush your finger in the direction you want the screen to go, and poof, it pans.  Very cool.  The result of this zooming and panning is that at any moment you might be looking at only a portion of the viewer window.  Something like this:

You can see, the controls for zooming and panning are at the bottom of the window, and are not visible on the screen.  Not good.

So with the iPhone viewer, the controls are only displayed when you want to use them.  You click on the screen with your finger, and the controls are displayed right there in on the screen, where you clicked. Like this:

The red circle shows where you clicked, and the controls are displayed so you can see them.  Now you can click on one of the controls to zoom in (plus), zoom out (minus), pan (arrows), or “show the whole slide” (the circle with an arrow).  If you click outside the controls, they are hidden again.  When you click on a control, it turns red to confirm that you clicked it.  For example here’s what it looks like after you zoom in:

The plus stays red while the image is being updated; then the controls are hidden again.

That’s really it; hopefully you will find this pretty easy to use.  This is now installed this on the Aperio slide gallery, which you can get to from our website under the Applications menu.  You can also go directly to http://aperio.com/gallery.

Oh, and you can view slides in landscape format just by turning your iPhone sideways.  Like this:

Have fun!

Please let me know if you have suggestions.  Or bugs!  This is only v0.1 :)

Breast Cancer Metastasis Suppressor MicroRNAs Identified

This is excellent: Reuters reports [ via Medscape ]:

NEW YORK (Reuters Health) Jan 09 - A team of scientists at Memorial Sloan-Kettering Cancer Center in New York has identified several endogenous human microRNAs that suppress breast cancer metastasis to lung and bone -- the two main sites of metastatic spread of breast carcinoma.

In the January 10 issue of Nature, Dr. Sohail F. Tavazoie and colleagues report that their search for "general regulators of cancer metastasis has yielded a set of microRNAs for which expression is specifically lost as human breast cancer cells develop metastatic potential."

Restoring the expression of these microRNAs, by means of plasmid vectors, in malignant cells suppresses lung and bone metastasis by human cancer cells in vivo, they found.

In mice with human breast cancer tumors, restoration of one microRNA, miR-126, reduced overall tumor growth and proliferation, whereas restoration of miR-335 inhibited metastatic cell invasion.

Dr. Tavazoie and colleagues also observed that expression of miR-126 and miR-335 was lost in the majority of primary breast tumors from women who relapsed, and loss of expression of either of these two microRNAs was associated with poor survival.

In further studies, the investigators found that miR-335 suppresses metastasis by targeting a transcription factor that regulates the SOX4 group of genes responsible for cell development and migration, as well as an extracellular matrix component implicated in cell migration called tenascin C.

"The strong association of the loss of miR-335 and miR-126 expression with metastatic relapse suggests the potential for the use of these molecules in prognostic stratification of breast cancer patients in addition to conventional clinical and pathological staging markers," the researchers write.

This research may also "enhance our ability to come up with more effective drugs to prevent or treat cancer metastasis," Dr. Joan Massagu, the study's senior author, added in a written statement.

"A team of scientists at Memorial Sloan-Kettering Cancer Center in New York has identified several endogenous human microRNAs that suppress breast cancer metastasis to lung and bone -- the two main sites of metastatic spread of breast carcinoma." 

This could lead to more effective drugs to treat tumors to reduce or prevent metastasis.

Milestone ScanScopes

Here's a cool Aperio milestone, we recently completed manufacturing our 100th ScanScope XT and our 200th ScanScope CS!  The picture below shows these two units in the manufacturing test area, along with a few of the many people responsible for this achievement. 

Not pictured but very much "in the picture" is Marc Cohn, Aperio's Director of Manufacturing, who leads this team.  Congratulations to all Aperions for their hard work and dedication which made this accomplishment possible!

Measuring Physician Performance

Wow, this is a pretty hot topic.  MedScape has an article by Dr. Carolyn Clancy: Physician Performance Measurement: Potential Benefits Outweigh the Challenges.  There are certainly some challenges!  Still, if you want to improve something, measure it. 

I wonder how one might measure the performance of Pathologists?

Something to talk about

When I resumed posting to this blog, I had given quite a bit of thought to why I stopped.  After all, I didn't want to start again, only to stop again :)  I love blogs and I like companies which have bloggers, so it really seemed like I should be blogging!

Posting blog items isn't really that time consuming, so that wasn't the reason.  I concluded that the reason I stopped was that I didn't know what to write about.  I can't really post about work in progress at Aperio – trade secrets, competitive intelligence, and all that – and I didn’t think merely recapitulating press releases and stuff from our website was worthwhile.  Blogs need a personal voice to be effective. There is quite a bit of stuff going on in Pathology which can be shared, but I didn’t think a news clipping service was exactly helpful either, although important developments could be noted, with our / my view of them.  (There are a lot of great resources for Digital Pathology news online, including some linked in the sidebar.)

The angle I’ve hit on for blog content is explaining digital pathology.  Especially some of the more technical things, after all, I’m a techie :)  One afternoon while watching a random college bowl game I made a list of “topics to be explained”.  It was a l o n g  list. There is certainly enough there to keep me posting for a while. We’ll see how it goes :)

Oh, and you can help!  First, if you have any suggestions for "things that should be explained", please let me know, either by commenting on the blog or sending me email.  A simple "hey, you should explain X" would be fine.  Next, please keep reading!  I'd love to get your opinions and feedback, and who knows, you might learn something :)  Or see something which is wrong, or isn't right, or isn't complete.  Many eyes have much power.

Web Site reveals the Mystery of Pathology

From Medscape (free registration required), this introduction to pathtalk.org, and a brief interview with the chief blogger, Dr. Trent McBride...

---

Nicholas Genes, MD, PhDMedscape Med Students.  2008; ©2008 Medscape Posted 01/08/2008 Of all the specialties in medicine, pathology might be the most mysterious. Although other physicians are dependent on their work, pathologists are rarely seen. They don't spend much time with (living) patients. Although you can find a great deal of pathology information on the Internet, the thoughts and opinions of pathologists are rarely heard. But that may change with PathTalk.org, a new site led by blogger Trent McBride, MD. I recently corresponded with Dr. McBride about his site and his perceptions on pathology. Dr. Genes: Pathology seems well suited for Web teaching. What was the motivation for PathTalk? Do you intend for it to be an educational tool or more like a forum? Dr. McBride: There are tons of pathology educational materials on the Web, but it's very fragmented, and in some sense it is backwards. (This applies to most pathology texts as well.) If you want to find some good pictures of hemangiopericytoma, you can find them in the length of time it takes to type "hemangiopericytoma" into Google Images. But if you want to look for a differential for spindle cell lesions that may have some vascular features, it's much harder to do, and most resources haven't gotten there yet. I would like to help develop resources that go from the problem to the answer, and not the other way around. I hope PathTalk can get a foot in the door and contribute to that. We can archive a set of problems and talk about how to arrive at the answer. But the blog can be so much more than that. I'd like it to be a clearinghouse of ideas within pathology specifically and medicine in general -- not just practical problems, but technology, research, policy, etc. Right now, we are just starting to get off the ground, and it has been really difficult; we have 3 main contributors who have healthy workloads in the real world. PathTalk hosts Grand Rounds January 8, 2008 Dr. Genes: Who are your colleagues at PathTalk, and how did you begin to work together? Dr. McBride: I met one of them (Gretchen Galliano) at a technology conference, and we decided to try to work together. From there I essentially emailed every path resident in the country to see who wanted to help. A third (Kenneth Youens) joined up, and thankfully so, as he did all the heavy lifting to get the Web site up. We need more people to generate the content at a healthy frequency, so there is pretty much an open invitation. And all the contributors so far being residents is just a starting point. I would like people at all levels of pathology training to be a part -- as both the audience and as authors. Dr. Genes: What's residency like for pathologists? Until I saw that movie trailer for Pathology that you previewed, I didn't really know why people chose pathology as a specialty. Are you in it for the glamour, the crime-solving satisfaction, or some other reason? Dr. McBride: If Pathology is half as good as Flatliners, I'll be impressed. Pathology residency is about what I expected: shorter hours at the hospital than other residencies, more time at home with a book open. You do have to know a little bit about everything, and if you are an academic personality, then it's a good residency. I could be way off base here, but I have to think we have more month-to-month variation in what we do than any other discipline. It's very diverse; many have this idea that what we do is boring, but we do so many different things that it is hard to get bored. Dr. Genes: What are some of the provocative pieces on PathTalk that have engendered good discussion? Dr. McBride: Gretchen has done a very good series on lab testing that brings up a host of difficult issues and has generated good discussion in the comments. With much of the new types of tests that are coming out, there is a hierarchy of questions to answer. First, does the test work scientifically? Next, clinically, what do you do with the information? Next, economically, is this thing we can do worth doing? Finally, politically, who pays for it? These aren't new questions, but the more important questions are shifting from the first to the last as these tests get more complex. When we first learned how to measure sodium, the clinical and economic benefits were pretty clear. With gene testing for polymorphisms that lead to differences in drug responses, those questions are really hard. Pathologists should have a lot to say in answering these questions. Dr. Genes: Pathologists will have a lot to say when PathTalk.org hosts Grand Rounds on January 8, 2008. Join Trent McBride and company as they review the best in online medical writing, with links to doctors, nurses, students, healthcare professionals, and patients around the world. Nicholas Genes, MD, PhD, second-year resident in emergency medicine, Mount Sinai Hospital, New York, NY Disclosure: Nicholas Genes, MD, PhD, has disclosed no relevant financial relationships.