Digitisation of electron microscope films: Six useful tests applied to three film scanners
Introduction
New types of digital camera, such as direct-detection back-thinned CMOS, hold the promise of performance superior to that of film, especially for detection of high-voltage electrons, but these are not yet commercially available. In the meantime, film still offers the best medium for recording high-resolution electron microscope images despite its well-known disadvantages of having to be extensively desiccated, physically loaded, developed, fixed, dried and digitised.
Film has a high intrinsic noise due to fog, dust and defects. To overcome this, it is desirable to use it at a magnification that gives the highest possible exposure whilst avoiding saturation. In practice, this means it should be exposed to an optical density (OD) of around 1.0. When used in this way, it is a superb medium for recording images. It can record and store images of 150 Mpixels with a detective quantum efficiency (DQE) and modulation transfer function (MTF) higher than that of the best available phosphor/fibre optic CCD cameras, although multiple binning (e.g. 2×2 or 4×4) of the data from phosphor/fibre optic CCD cameras can improve the CCD data at the cost of a reduced field of view [1], [2], [3]). In particular, at high voltages (300 keV), the effects of backscattering of electrons from the plastic support of the emulsion and the film holder are less than those found from the fibre optics of a CCD camera. In due course, new types of electronic camera, such as Medipix [4] and direct detection CMOS [5], will be developed further and should eventually replace film.
In structural biology, radiation damage limits the electron dose that can be used to study unstained biological macromolecules embedded in ice to around 10 el/Å2 and this limits the signal-to-noise ratio of the resulting images. As a result, images of many thousands of individual molecules must be averaged before the structure of the average molecule can be determined [6], [7], [8], [9]. It is therefore critically important that all steps between image formation and the availability of a digitised image in the computer should be carried out with the greatest fidelity.
The digitisation step should ideally capture all the high-resolution information and not introduce further noise or geometric distortions. In electron microscopy, we wish to measure the incident electron flux, which is proportional to OD, and this is obtained by conversion of the light transmission (T) measured by the scanner into OD=−log10 T. The requirements on geometrical accuracy and reproducibility during digitisation depend on the nature of the specimen. For images of molecules which form two-dimensional crystals, the procedure of indexing the diffraction pattern and unbending any lattice defects [10] automatically compensates for small non-linearities, differential (XY) magnification and non-orthogonality of the scanner axes. Similarly, for images of helical arrays, the layer line indexing and other scale factors serve the same purpose [11]. For single-particle images, however, there is no underlying lattice to act as a graticule to remove scanner defects [9]. It is therefore important that the digitisation should be done precisely on an orthogonal grid with no differential XY magnification and no significant distortions (e.g. pincushion or barrel distortion in a lens coupled system). Since film typically has a useful MTF for Fourier spacings beyond 1/10 μm−1, corresponding to pixel sizes of 5 μm or less, it is therefore also important that positioning should be precise and reproducible to within a small fraction of this (e.g. 1 μm).
We have been involved with the development and evaluation of film scanners for many years, and in this paper, we describe the tests and results that we have been using to evaluate and to monitor performance. Some of the tests are very quick and serve to provide a routine health check, whereas others measure distortions or geometrical parameters that are permanent features of a particular scanner. The results of the tests show that relatively inexpensive desk-top film scanners are surprisingly good. However, more precise instruments, which satisfy the requirements of high-resolution electron microscopy (EM), particularly single-particle EM, are needed if the results of image processing are not to be degraded by the digitisation step.
Section snippets
Scanners tested
We have tested three film scanners, the Nikon Coolscan 8000, The Zeiss Intergraphics SCAI and a laboratory prototype, which we refer to, after its principal components, as the KZA scanner. All three scanners use a linear CCD to record the image of a line on the film, which is moved in a direction at right angles to the CCD to create the image. Depending on the magnification used, the entire area of the film can be digitised with either a single scan or several successive scans in which the
Six tests
Each of the tests is carried out using a special test film, and detailed descriptions of how to make the films are included in the details that follow:
Discussion and conclusions
We have found the above tests provide all the information needed to determine whether a film scanner is contributing noise, systematic geometrical errors or random positional errors to the digitisation of an electron micrograph. All six tests would normally be used in evaluation of a new scanner. Two tests, A and F, are particularly quick to carry out and can be used to check routinely that there has been no degradation in the performance of the instrument. The computer programs described in
Acknowledgements
We are very grateful to Frank Mallett who made many contributions to film scanner construction and evaluation at the MRC Laboratory of Molecular Biology until he retired. He put forward the original concept for the KZA scanner described here. We also thank Joanita Jakana and Wah Chiu for digitising our test films on their Nikon Coolscan 8000 scanner, and Klaus Neumann for his help over many years in improving the performance of our Z/I film scanner. This work was supported in part by research
References (15)
- et al.
J. Struct. Biol.
(2003) - et al.
J. Struct. Biol.
(2004) - et al.
J. Struct. Biol.
(2005) - et al.
Nucl. Instrum. Methods
(2005) - et al.
Nucl. Instrum. Methods
(2005) - et al.
J. Mol. Biol.
(2003) - et al.
Ultramicroscopy
(1986)
Cited by (16)
-
Direct imaging detectors for electron microscopy
2018, Nuclear Instruments and Methods in Physics Research, Section A: Accelerators, Spectrometers, Detectors and Associated Equipment -
An algorithm for estimation and correction of anisotropic magnification distortion of cryo-EM images without need of pre-calibration
2016, Journal of Structural BiologyCitation Excerpt :To better understand the variations of elliptic distortion, we have investigated the particle location dependence of (Δjspr, θjspr); Tulane virus particles located at the bottom and top right corners on films were found having larger Δjspr, while no significant location dependence for θjspr was observed (Supplementary Fig. 3C). Such location dependence of Δjspr was likely due to scanner defects during digitization (Henderson et al., 2007) and it was not observed for datasets recorded with the K2 detector. All these tests suggested that the spread of the elliptic distortion parameters (Δjspr, θjspr) among particles revealed by the anisoscale aligner indeed reflected true variable magnification distortion within a dataset to a large extent.
-
Tilt-pair analysis of images from a range of different specimens in single-particle electron cryomicroscopy
2011, Journal of Molecular BiologyCitation Excerpt :Briefly, solutions of β-galactosidase at a concentration of 1 mg/ml were applied to glow-discharged Quantifoil grids (Agar Scientific), blotted, and plunge frozen using a homemade apparatus similar to that described by Dubochet et al.5 Grids were transferred to an FEI Polara G2 microscope and images were recorded on film at 39,000× magnification and 80 keV with defocus between 1.5 μm and 2.0 μm using an electron dose of 8 electrons/Å2 and developed for 12 min in full-strength D19 developer. The micrographs were digitized on the KZA film scanner62 in 6-μm steps, defocus was estimated using CTFFIND3,46 and particles were picked manually using Ximdisp63 and then processed using EMAN239 to give a structure at ∼ 13 Å resolution, which will be described in more detail in a later publication. We also calculated a 3D map starting with the atomic coordinates (Protein Data Bank: 3I3E) of E. coli β-galactosidase.43
-
Structure of rhomboid protease in a lipid environment
2011, Journal of Molecular BiologyCitation Excerpt :The quality of the negatives was evaluated by optical diffraction for the absence of astigmatism and tilt, and those exhibiting strong reflections to 6–8 Å were selected for further processing. The whole negative was digitised with a pixel size of 6 μm on a KZA scanner50 and manually inspected using the interactive FFT option in the Ximdisp program.51 The best-ordered area of 4000 × 4000 pixels was selected using the program LABEL and further processed using the MRC image processing program to correct for lattice distortions and contrast transfer function.34,52
-
Images of paraffin monolayer crystals with perfect contrast: Minimization of beam-induced specimen motion
2011, UltramicroscopyCitation Excerpt :Although this produced much larger files, it allowed diffraction spots out to a 1.5 Å resolution to be seen in the best images. Note that the SCAI film scanner, manufactured originally by Zeiss and in wide use by the EM community, has a similar MTF to the KZA film scanner used here [24], but has a pixel size limited to 7 μm. The Medipix detector has 55 μm pixels, so this was equivalent to 1.8 Å pixels, which again allowed only the lower resolution paraffin lattice spots to be observed.