Automated microscopy and image analyses are revolutionizing biological research by: 1) adding objective quantification to classically subjective imaging, 2) turning months-long sequential trial-and-error slide-based experiments into days-long parallel multi-well plate experiments, and 3) enabling fast, large-scale image-based screens for active chemical compounds and genes (e.g., RNAi and cDNA) (1
). Advances in automatic acquisition, measurement, comparison and pattern classification tools for cellular images continue (17
). Autofocus is a critical component of microscope automation and is most challenging to implement when using high numerical aperture objectives (NA ≥ 0.5), which exhibit high resolutions and small depths of field. Precise autofocus overcomes problems that include mechanical instability, movement of live specimens, variable thickness, drift (e.g., due to thermal expansion) and irregularities of biological substrates (e.g., slides and microtiter plates). Reflective positioning using laser-based methods (30
) can be faster but finding best focus by measuring the resolution of the images is more direct and can produce sharper images, especially with higher NA objectives.
Automated microscopes most commonly use incremental scanning, in which the stage is motionless while acquiring the image and then moved to the next field. At each field, a motorized stage moves the specimen (or objective) along the optical axis (z-direction) to collect a stack of images for measuring the position of sharpest focus (33
). The first field of each specimen can be focused manually (as is typical for manual loading of slides or microtiter plates) or automatically using a longer axial search range (as is typical in robotic slide/plate loading). Entire slides (with thousands of images) and hundreds of 384-well plates in a row (with millions of images) are routinely scanned automatically using autofocus. The axial positioning is performed by an objective positioner, or as for the experiments reported here, a motor on the fine focus knob of the microscope, or more recently, piezoelectric stage inserts (e.g., Mad City Labs, Madison, WI, www.madcitylabs.com/
and Physik Instrumente, Karlsruhe, Germany, www.physikinstrumente.com/
). Focus is moved through a specific number of steps (Δz) in the z-direction. Autofocus is performed by collecting and analyzing a sequence of images by applying various measures of resolution, acquired at different test object planes from different z-positions (33
). The highest relative focus measurement value corresponds to the best focused image. After finding best focus, the image is acquired and the stage is moved to the next field. In our experiments, maintaining focus in large scale scanning (thousands of fields) requires measuring focus on at least seven test focal planes (34
). For incremental scanning, we demonstrated real-time focus measurements on a dedicated circuit board to perform autofocus in as fast as 0.25 s on adjacent fields of view for cells on coverslips; including stage motion, the fastest scanning rate was about 3 fields/s (34
With vibrations blurring image acquisition for faster stage movements, we implemented parallel acquisition of multiple image planes with multiple real-time resolution-measurement circuits to achieve on-the-fly autofocus during continuous stage motion (37
). Time-delay-and-integration (TDI) scanning using large format area CCD cameras synchronized to the stage motion enables high sensitivity and high speed collection of long 2-D image strips. The combination of continuous stage motion with multiplanar image acquisition can increase the scanning speed four- to ten-fold, depending on the application. A “volume camera” is used to acquire multiple focal planes simultaneously, as shown in the cartoon in (realized via imaging fiberoptic bundles or beamsplitters) (40
Figure 1 A cartoon of simultaneous multiplanar image acquisition is shown.(40)
Multifocal imaging is also useful for 3D imaging of live specimens. Prabhat et al
. utilized a 4-camera multi-focal-plane system to image fast events in living cells in 3D(41
), including a study of the sorting of endosomes to exocytosis at the plasma membrane by direct imaging (41
). Mulitplanar imaging systems are somewhat complex, making them cumbersome, expensive and prone to additional aberrations.
Constructing a stack of images from the differences in foci produced by chromatic aberration on an RGB camera could create a simpler optical configuration in continuous-motion scanning and would speed conventional incremental scanning. Although microscope objectives are corrected for chromatic aberration, the correction is not perfect. In recent study where chromatic aberration correction with deconvolution was used to measure colocalization of sterols in lipid droplets in cells, 0.8–1.0 µm and 2.0 µm axial chromatic aberrations were reported for fluorescence emission wavelengths of 405 nm vs. 610 nm on multicolor 0.1 µm fluorescent beads using a 63X 1.4 NA oil objective on two different Leica microscopes (44
). Each microscope objective should be assumed to have different aberrations (this may be true even comparing two objectives with the same part number), and objectives of increasing quality and cost have better aberration corrections than those of lower cost/quality. In addition to aberrations in the microscope optics themselves, the optics in a 3-chip camera may also contribute to the differences in foci at different wavelengths. Controlling differences in foci may only be practical with a mechanism for automatically adjusting chromatic aberration in a precalibrated manner for each objective/optical configuration.
The addition of a dispersive planar glass element in the diverging (or converging) beam path of the optical system produces chromatic aberration. It should thus be possible to control the amount of chromatic aberration (i.e. the separation of the focal planes acquired simultaneously on an RGB camera) by changing the thickness of the glass element and choosing glass with the appropriate refractive index and Abbe number (dispersion). To determine if the use of chromatic aberration might be practical for simultaneous acquisition of multiple focal planes we: 1) measured differences in focal planes with a 3-chip CCD camera and 2) explored insertion of different types of glass into the optical path to determine the potential for controlling the degree of chromatic aberration.