Monday, August 18, 2008

An interactive poll on the adoption of digital pathology

I thought that one of the ways we can estimate the adoption of digital pathology is to run an admittedly unscientific poll, asking respondents to describe changes they observe in their work environment. With that in mind, I have put together an introductory poll whose results you will be able to see on the blog as respondents reply.

The first question is an attempt to evaluate the internal-influence or "word of mouth" driver of the famous "S" curve [dating back to Bass in 1969. For details, see, e.g., Mahajan's monograph]. The question asks you to specify how many of your colleagues you know have worked with whole-slide images on a computer, have transformed glass slides into digital image collections, have used image analysis to grade specimens, etc.

We don't suggest a strict definition of "digital pathology." Instead we are interested in finding out how many of your peers you see doing what you would consider to be a digital form of pathology.

A couple of quick instructions: (1) Note the slider and move it to see all answer choices as well as the accumulating statistics; and (2) every respondent can only vote once so please don't overlook any of your colleagues, you won't have a chance to update your answer. Finally, we will keep the poll open for two (2) weeks. The results are displayed live at all times.

Thursday, June 5, 2008

How to interpret GE's and UPMC's entry into the nascent digital pathology market?

The long-anticipated move by GE Healthcare has finally taken place. GE has partnered with UPMC to develop a digital pathology total solution. The venture is called Omnyx ( and is funded to the tune of $40MM. That amount of money is almost as much as the funding that Aperio Technologies has accumulated since its founding.

This is exciting news, especially as it may affect the current players and the market. Gene Cartwright, Omnyx's CEO, compared GE and Aperio Technologies in the Pittsburgh Tribune-Review as follows: "Aperio is a nice little company, and it's the current market leader," Cartwright said. "But GE is a software company, and Aperio just can't keep up with the high scan through-put customers need."

DMetrix's perspective on this has to do with the last part of that quotation. We have long focused on high-throughput imaging. Our current products can already scan a glass slide in less than 60 seconds and sustain throughputs on the order of 40 slides/hour. GE's and UPMC's entry into this market validates the promise of digital pathology further but it also confirms that emphasis on speed of image capture will be a key differentiating feature among slide scanners. GE's prototype scanner, assuming that it is similar to that presented at APIII 2006 in Vancouver, relies on a specialized design, rather than an adaptation of a conventional microscope. Here, too, DMetrix's innovative approach centered on the array-microscope concept is validated.

We will be blogging further from the upcoming CAP Futurescape meeting taking place this weekend in Chicago. With talks by Omnyx senior management, this promises to be a very interesting meeting.

Tuesday, April 8, 2008

Topographic focusing

There are two basic ways of scanning a glass slide with a specimen. The one way involves using a single microscope objective. The other way involves using an array of microscope objectives in order to get the job done a whole lot faster.
Part of the speed advantage of the array microscope comes from being able to focus much more rapidly on the material on the glass slide with the array microscope than with a single microscope objective. It's a matter of searching out tissue, for instance, in many places at once instead of having to plumb the depth of the slide one spot at a time.
An array microscope searches for biological material on a glass slide in at least 80 places simultaneously. For most slides, the number of points at which tissue is detected reaches 240. It may be as high as 480. An array microscope measures topography of the specimen every 1.6 mm or so. That means that tissue topography whose knowledge is needed for accurate, focused imaging, leaves little to the imagination.
Compare 240 focus points to the 24 focus points or so that a single-objective scanner measures, one at a time. An array microscope is well along in its accurate scanning of a slide by the time that a single-objective scanner gets done with the 24th focus point.
Finally, it turns out that you can connect dense focus measurements of the array microscope to the operations of a histopathology laboratory. We will return to how that's accomplished in a later post.

Monday, April 7, 2008

What connects 30" computer displays, eye-tracking, and pathologists' levels of experience?

An interesting threshold has been crossed by computer displays relative to modern light microscopes.
A well corrected light microscope uses so-called plano or "Plan" objectives (as in "PlanApo"). Such objectives are made to have a particularly large, flat field of view. The diameter of a 20X PlanApo objective's field of view is approximately 1 mm.
In a different context, 30" computer displays have become relatively common today. They are offered by Apple, Dell, and HP, for instance. A 30" computer display has a format of 2,560 pixels by 1,600 pixels.
A slide scanner that captures images at an equivalent magnification of 20X, creates images with pixels whose size on the specimen is typically 0.5 microns/pixel. The area of the specimen that can be displayed on a 30" computer display therefore measures 1.25 mm by 0.8 mm. A 30" display therefore shows a specimen area that is 27% larger than the area visible through eyepieces on a modern analog microscope (1 sq-mm vs. 0.78 sq. mm). Any smaller format display (e.g., a 24" display with 1,920 by 1,200 pixels) does not exceed the "PlanApo" threshold. Viewing images on a smaller computer display is equivalent to turning down the field stop in an analog microscope.

This is interesting news in its own right, but what is its real impact on digitizing pathology? A recent human-factors study by Krupinski, et al., compared eye movements by medical students, residents, and fully trained pathologists when reviewing breast core biopsy cases:

“…Our objective in this study was to take advantage of virtual slide technology to compare the eye movements of virtual slide readers with different levels of experience. …”
“…Unlike either the medical students or the residents, the [fully trained] pathologists frequently choose areas for viewing at higher magnification outside of areas of foveal (central) vision. …” [emphasis added by me]

(The Krupinski paper [Krupinski, et al., Human Pathology (2006) 37, 1543–1556] includes several amazing figures (especially Figures 7 and 8) which show the scan-paths of fully trained pathologists, residents, and medical students, respectively.)

The conclusion is that 30” computer displays are the minimum size that allow experienced pathologists to work naturally instead of being constrained to working as they had at earlier stages in their training, if paired with a smaller-format computer display. It's very hard to use one's peripheral vision when there is nothing to see with it!
This conclusion is further supported by anecdotal evidence. DMetrix has been demonstrating its scanners with 30" displays since 2005.

Sunday, April 6, 2008

Inaugural posting

The compound microscope is a device that was invented to extend the function of the human eye. In this role, the microscope has served many uses and has become synonymous with science unlike any other instrument. It still serves as a ubiquitous instrument in research and medicine, mounted on many a desk or benchtop. However, the emergence of digital imaging, novel photophysics, broadband communications, search engines, and analytical software have created a new context for the glass-and-metal instrument familiar to so many. In this blog, we will span the range from presenting large scale trends and applications to the fine details of implementation of all these elements in the particular context of digital pathology.