daily coverage of the optics & photonics industry and the markets that it serves
Featured Showcases
Photonics West Showcase
Historical Archive

Virtual microscopy gets machine-vision upgrade

12 Dec 2006

The digital imaging of microscope slides is benefiting from machine-vision know-how in the form of high-performance linescan cameras and custom software. James Tyrrell speaks with Hamamatsu’s Paul Cormack to get to grips with the technologies driving virtual microscopy.

Imagine a digital image that can be magnified many times without losing any detail, viewed on a PC and shared over the Internet. It is this concept that lies at the heart of virtual microscopy, a technique that is capturing the imagination of pathologists, pharmaceutical companies and teachers.

Rapid advances in digital-camera technology and computer hardware have pushed virtual microscopy a long way since its beginnings in the late 1990s. Increasingly, the technique is benefiting from expertise in other fields such as machine vision and software engineering. Today, developers are offering systems configured with a linescan camera, which can manipulate gigabytes of data.

OLE spoke with Paul Cormack, sales engineer for Hamamatsu Photonics, UK, to find out more, starting with the basics.

OLE: What is virtual microscopy?

PC: The term virtual microscopy is used in different contexts. In one context it refers to interactive microscope simulators, which are graphical representations of a microscope. Parameters can be altered to simulate what will happen to an image when the settings on a real microscope are changed. This type of virtual microscope can be very useful as a learning tool.

A different type of virtual microscopy is a rapidly developing technology that is set to make a big impact in the laboratory. In this context, a virtual microscope is a means of creating and then viewing a virtual slide (sometimes called a digital slide) over a computer network, or on the screen of a stand-alone computer.

It is the virtual– or digital–slide that is the major enabling component of a virtual-microscopy system. The virtual slide is a very large digitized image file of a glass slide, which can be viewed, panned and zoomed (while still maintaining real resolution rather than increasing empty magnification) on a computer screen.

Key markets for the technique are routine clinical, research and teaching applications in pathology and high-throughput screening, and analysis of slides in the pharmaceutical industry.

Virtual microscopy boasts an extensive list of advantages, many of which are not possible with traditional glass slides. Virtual slides can be easily stored, archived, retrieved, annotated, duplicated, distributed, integrated into electronic patient records and viewed over the Internet or via a private computer network.

With virtual slides, operators can simultaneously view an image of the whole tissue section and zoom into a particular part of the sample for a more detailed view. This can be an invaluable aid to the pathologist who wants to relate a detailed view of one part of the section to the overall geography of the specimen.

How are virtual slides created?

Virtual slides are generated by scanning systems based on area-scan or line-scan CCD technology, which digitize the image of the glass slide. Other technologies are starting to appear, but are still in the minority.

Area-scanning systems generally consist of a microscope, a conventional digital camera, a motorized XY stage and a Z stepper to adjust focus. A PC with integral stage and microscope control boards automates the stage and the stepper, and controls the microscope’s illumination, condenser, top lens and objectives. These systems work by moving the slide, which is mounted on a motorized scanning stage under the microscope objective, and capturing a digital image of the field of view (often referred to as a tile). The scanner then moves to the adjacent field of view and repeats the digitization process until the whole, or a selected region, of the slide has been captured. The tiles produced by this method are then stitched together using sophisticated software that aligns each tile to create a large mosaic image.

Line-scanning systems often comprise an automatic slide-loader, a robotic stage, a single very-high-quality microscope objective and a linear-array camera. The slide is scanned in a continuous manner up and down the region of interest to create the virtual slide.

What are the pros and cons of the different methods?

As area-scan systems are usually microscope-based, the technology is familiar to laboratory staff and there is the opportunity for existing equipment to be used or adapted to create virtual slides.

System calibration is critical and the electromechanical and software components have to be of the highest quality to ensure the accurate alignment of what could be thousands of image tiles.

Using an area-scan system and camera with a resolution of 1280 × 1024 pixels (1.3 Mpixel), approximately 800 tiles would have to be captured for a tissue section area of 15 × 15 mm if a virtual slide resolution of 0.46 μm/pixel is to be achieved. If the desired resolution is 0.23 μm/pixel then four times as many tiles have to be digitized to give this resolution.

Line-scan systems have their origins in machine vision and therefore the technology is unknown to many laboratory staff. There is also very little likelihood that existing laboratory equipment could be used or adapted to create virtual slides.

One of the benefits of line-scan systems is that they produce far fewer tiles than an area-scan system. For example, a linear array sensor with 4096 pixels would only produce approximately 15 tiles for an area of 15 × 15 mm at a resolution of 0.23 μm/pixel. This greatly reduced number of tiles simplifies the process of tile alignment.

Resolutions of 0.46 and 0.23 μm/pixel equate to resolutions obtained using good-quality microscope objectives of 20× and 40× magnification respectively.

In addition to the reduced number of tiles, linear-array systems continuously focus many times along the length of the tiles and may be better at compensating for subtle changes in the topology of a specimen than area-scan systems.

Some of the fastest systems can process a 15 × 15 mm sample in around 1 min.

TDI is one way of enhancing the image – how does it work?

Time delay and integration (TDI) is a specialized sensor-readout mode that integrates the light signal from each pixel in the image and can greatly improve the quality and contrast of images, even under low-light-level conditions. For example, a 4096 ×  64 line-scan array would integrate (build up) the signal from each pixel by a factor of 64. This feature can be particularly beneficial when scanning fluorescence-labelled slides.

The primary advantage of TDI operation is the increased image-integration time that it gives in comparison to standard linear-array sensors.

From a developer’s point of view, what are the technical hurdles?

The challenge is to be able to consistently create a high-fidelity digital representation of the material on the original glass slide. This is particularly important for virtual slides that are to be used as the source for quantitative data analysis software.

Another consideration is that experienced microscope users expect the operation of virtual microscopes to be as intuitive and fast as traditional microscopes. This means being able to handle large amounts of data very rapidly as the user focuses back and forth through the virtual sample.

In fact, some of the most serious short-term challenges are due to the shortcomings of current PC technology. A huge volume of data can be generated by the latest virtual-microscopy systems and the bottleneck for rapidly scanning and viewing slides is often the architecture of the PC.

• This article originally appeared in the December 2006 issue of Optics & Laser Europe magazine.

First Light ImagingAlluxaUniverse Kogaku America Inc.CHROMA TECHNOLOGY CORP.TRIOPTICS GmbHBerkeley Nucleonics CorporationLaCroix Precision Optics
© 2024 SPIE Europe
Top of Page