Automatic 3D alignment of projection images of randomly oriented objects

• Published in: Ultramicroscopy Vol. 52; no. 2; pp. 141 - 156
• Main Authors: Farrow, Neil A., Ottensmeyer, F.Peter
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 01.11.1993; Elsevier Science
• Subjects: Analytical biochemistry: general aspects, technics, instrumentation; Analytical, structural and metabolic biochemistry; Biological and medical sciences; Fundamental and applied biological sciences. Psychology; Simulation; Projection; Biological macromolecule; Density distribution; Experimental study; Image; Electron microscopy; Micrography
• ISSN: 0304-3991; 1879-2723
• DOI: 10.1016/0304-3991(93)90185-Z

Reconstruction of the 3D density distribution of macromolecular structures from their 2D projection images or electron micrographs requires knowledge of the relative orientations of the 2D projections. Using a technique that combines the principles of the common axis theorem and quaternion mathematics, the projection directions may be determined a posteriori from the images. This technique is applied to stimulated electron micrographs of the Klenow fragment of DNA polymerase I. These simulated images, with graded signal-to-noise ratios, are subjected to image processing procedures, such as 2D translational and rotational registration and multivariate statistical classification, that would normally be applied to electron micrographs to improve their quality. The alignment procedure is able to determine the relative 3D orientations of the projections correctly even when the images have an initial signal-to-noise ratio of 3.1 and when images within a single class of views are deliverately perturbed in orientation. A relationship is obtained between the parameters used to describe the success of the 3D alignment algorithm and the quality of the 3D reconstruction that might be expected.