Hardware, image acquisition
Leica DMLA microscope with a set of PlanApo lenses (HC Fluotar 5/0.15, HC PlanApo 10/0.30, 20/0.50, 40/0.70, 100/1.35 and a Plan 2/0.07 lenses) equipped with the Nikon DMX-1200 digital camera is used to obtain image parts at the resolution at 1200 × 1020 pixels, 3 × 8 bit colour. Motorized stage (Merzhäuser) is automatically moved from one image to another. The system is controlled by Lucia DI (LIM, Prague). Home made software is used to create composed, very large pictures and to prepare the virtual microscope image stacks.
Source texts of the atlas
The atlas source texts are in XML data format. Programs (written mostly in Perl 5.8 programming language) are used to parsing and checking the document structure and to generate the final HTML files, which are uploaded on the server. The overall size of the atlases is about 110 GB of data.
Image post-processing, virtual microscope
Each image part can be taken in one or more focusing levels. This image stack can be processed by pan-focusing function, which selects sharp areas of each image from the stack to create one image tile and overcoming the problem of image artefacts caused by uneven slides. This feature is especially useful if only slides of suboptimal quality are available (slides of rare cases, from old slide collections etc.). We use usually 3 levels, their distance of focusing planes varies according to the objective used.
Alternatively the whole image stacks are taken and saved. The resulting images are created from stitching tiles from each plane separately. This approach allows creation of multilevel images, which can be focused. This approach we use especially in images taken in high resolution using 100× immersion lens, as in bone marrow biopsies. We use usually 5 or 7 focusing planes, sometimes more (up to 13).
After stitching each image is digitally manipulated (colour correction, sharpening), archived and cut into pieces, side of which are of multiples 256. Larger pieces 512 and more) are converted into 256 parts. All these image parts are saved into structured directories.
The browser loads proper parts of image according to current viewport, magnification and focusing level. It reacts to user's actions (image dragging, magnification changing, change in the size of the virtual microscope window, focusing) through catching events, calculates the names of corresponding new image tiles, which are loaded and added into the DOM of the image being displayed. The image parts, which got out of he viewport, are released from memory.
The individual parts can be stored on the server or locally on the disc. No special server application is needed. The magnification can be changed, images can be dragged and the images saved in stacks can be focused. This approach is used especially in images t Users can change the size of the virtual microscope window up to the full screen.