Globally Optimal Stitching of Tiled 3D Microscopic Image Acquisitions

Max Planck Institute of Molecular Cell Biology and Genetics, Dresden, Germany.
Bioinformatics (Impact Factor: 4.98). 05/2009; 25(11):1463-5. DOI: 10.1093/bioinformatics/btp184
Source: PubMed


Modern anatomical and developmental studies often require high-resolution imaging of large specimens in three dimensions (3D). Confocal microscopy produces high-resolution 3D images, but is limited by a relatively small field of view compared with the size of large biological specimens. Therefore, motorized stages that move the sample are used to create a tiled scan of the whole specimen. The physical coordinates provided by the microscope stage are not precise enough to allow direct reconstruction (Stitching) of the whole image from individual image stacks.
To optimally stitch a large collection of 3D confocal images, we developed a method that, based on the Fourier Shift Theorem, computes all possible translations between pairs of 3D images, yielding the best overlap in terms of the cross-correlation measure and subsequently finds the globally optimal configuration of the whole group of 3D images. This method avoids the propagation of errors by consecutive registration steps. Additionally, to compensate the brightness differences between tiles, we apply a smooth, non-linear intensity transition between the overlapping images. Our stitching approach is fast, works on 2D and 3D images, and for small image sets does not require prior knowledge about the tile configuration.
The implementation of this method is available as an ImageJ plugin distributed as a part of the Fiji project (Fiji is just ImageJ:

Download full-text


Available from: Stephan Preibisch,
76 Reads
  • Source
    • "datasets were processed by ImageJ software. When a region of interest was larger than the viewing fi eld of the microscope, confocal sections of adjacent viewing fi elds were stitched using an ImageJ plug-in (Preibisch et al., 2009). To enhance some magnifi ed images (Figures 2C4, 2C5, 3E, 3F), confocal datasets were deconvolved by Huygens Pro software (Scientifi c Volume Imaging, Hilversum, the Netherlands). "
    [Show abstract] [Hide abstract]
    ABSTRACT: Although the gustatory system provides animals with sensory cues important for food choice and other critical behaviors, little is known about neural circuitry immediately following gustatory sensory neurons (GSNs). Here, we identify and characterize a bilateral pair of gustatory second-order neurons in Drosophila. Previous studies identified GSNs that relay taste information to distinct subregions of the primary gustatory center (PGC) in the gnathal ganglia (GNG). To identify candidate gustatory second-order neurons (G2Ns) we screened ∼5,000 GAL4 driver strains for lines that label neural fibers innervating the PGC. We then combined GRASP (GFP reconstitution across synaptic partners) with presynaptic labeling to visualize potential synaptic contacts between the dendrites of the candidate G2Ns and the axonal terminals of Gr5a-expressing GSNs, which are known to respond to sucrose. Results of the GRASP analysis, followed by a single cell analysis by FLPout recombination, revealed a pair of neurons that contact Gr5a axon terminals in both brain hemispheres, and send axonal arborizations to a distinct region outside the PGC but within the GNG. To characterize the input and output branches, respectively, we expressed fluorescence-tagged acetylcholine receptor subunit (Dα7) and active-zone marker (Brp) in the G2Ns. We found that G2N input sites overlaid GRASP-labeled synaptic contacts to Gr5a neurons, while presynaptic sites were broadly distributed throughout the neurons' arborizations. GRASP analysis and further tests with the Syb-GRASP method suggested that the identified G2Ns receive synaptic inputs from Gr5a-expressing GSNs, but not Gr66a-expressing GSNs, which respond to caffeine. The identified G2Ns relay information from Gr5a-expressing GSNs to distinct regions in the GNG, and are distinct from other, recently identified gustatory projection neurons, which relay information about sugars to a brain region called the antennal mechanosensory and motor center (AMMC). Our findings suggest unexpected complexity for taste information processing in the first relay of the gustatory system.
    Journal of neurogenetics 05/2015; DOI:10.3109/01677063.2015.1054993 · 1.27 Impact Factor
  • Source
    • "To evaluate whether our device is suitable for scanning of biological probes larger than the field-of-view of the microscope, we cultured HeLa cervix carcinoma cells in 34-mm dishes and stained the cells using haematoxylin and eosin (H&E) as described in the methods section. In a next step, cells were imaged in a meander-like pattern at 900 (30x30) overlapping locations and were subsequently stitched using FIJI software (Schindelin et al. 2012) and the plugin Grid/Collection stitching (Preibisch et al. 2009)(see Methods). Figure 3A shows the result of the reconstructed image. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Time-resolved visualization and analysis of slow dynamic processes in living cells has revolutionized many aspects of in vitro cellular studies. However, existing technology applied to time-resolved live-cell microscopy is often immobile, costly and requires a high level of skill to use and maintain. These factors limit its utility to field research and educational purposes. The recent availability of rapid prototyping technology makes it possible to quickly and easily engineer purpose-built alternatives to conventional research infrastructure which are low-cost and user-friendly. In this paper we describe the prototype of a fully automated low-cost, portable live-cell imaging system for time-resolved label-free visualization of dynamic processes in living cells. The device is light-weight (3.6 kg), small (22×22×22 cm) and extremely low-cost (<€ 1,250.-). We demonstrate its potential for biomedical use by long-term imaging of recombinant HEK293 cells at varying culture conditions and validate its ability to generate time-resolved data of high quality allowing for analysis of time-dependent processes in living cells. While this work focusses on long-term imaging of mammalian cells, the presented technology could also be adapted for use with other biological specimen and provides a general example of rapidly prototyped low-cost biosensor technology for application in life sciences and education.
    Biosensors & Bioelectronics 12/2014; 64. DOI:10.1016/j.bios.2014.09.061 · 6.41 Impact Factor
  • Source
    • "Orientation of cell sheets was determined from actin images using fast-fourier transform (FFT) analysis (FIJI Directionality tool), with pixel intensities summed at 2 increments over the power spectrum (À90 eþ90 ) to generate alignment histograms [40] [41]. In cases where overall pattern dimensions were larger than a single image field, tiled images were stitched (FIJI Grid Stitching [42]) prior to FFT analysis. An image's degree of alignment was quantified by calculating the Alignment Index (AI) [40] using Equation (1), where q m is the mode of the FFT histogram and I is pixel intensity. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Tissue and biomaterial microenvironments provide architectural cues that direct important cell behaviors including cell shape, alignment, migration, and resulting tissue formation. These architectural features may be presented to cells across multiple length scales, from nanometers to millimeters in size. In this study, we examined how architectural cues at two distinctly different length scales, "micro-scale" cues on the order of ∼1-2 μm, and "meso-scale" cues several orders of magnitude larger (>100 μm), interact to direct aligned neo-tissue formation. Utilizing a micro-photopatterning (μPP) model system to precisely arrange cell-adhesive patterns, we examined the effects of substrate architecture at these length scales on human mesenchymal stem cell (hMSC) organization, gene expression, and fibrillar collagen deposition. Both micro- and meso-scale architectures directed cell alignment and resulting tissue organization, and when combined, meso cues could enhance or compete against micro-scale cues. As meso boundary aspect ratios were increased, meso-scale cues overrode micro-scale cues and controlled tissue alignment, with a characteristic critical width (∼500 μm) similar to boundary dimensions that exist in vivo in highly aligned tissues. Meso-scale cues acted via both lateral confinement (in a cell-density-dependent manner) and by permitting end-to-end cell arrangements that yielded greater fibrillar collagen deposition. Despite large differences in fibrillar collagen content and organization between μPP architectural conditions, these changes did not correspond with changes in gene expression of key matrix or tendon-related genes. These findings highlight the complex interplay between geometric cues at multiple length scales and may have implications for tissue engineering strategies, where scaffold designs that incorporate cues at multiple length scales could improve neo-tissue organization and resulting functional outcomes.
    Biomaterials 09/2014; 35(38). DOI:10.1016/j.biomaterials.2014.08.047 · 8.56 Impact Factor
Show more