PROCEEDINGS
`
`FIFTY-SECOND ANNUAL MEETING
`
`MICROSCOPY SOCIETY OF AMERICA
`
`TWENTY-NINTH ANNUAL MEETING
`
`MICROBEAM ANALYSIS SOCIETY
`
`NEW ORLEANS, LOUISIANA
`
`31 July-5 August 1994
`
`Editors
`G. W. BAILEY
`and
`A. J. GARRATT-REED
`
`
`III In nllllllfl “III
`II
`‘M’
`
`
`
`Box 426800. San Francisco, CA 94142-6800. USA
`
`1 9 9 4
`
`ZEISS, et al., Ex. 1014
`
`

`
`G. W. Bailey and A. J. Garratt-Reed. Eds., Proc. 52nd Annual Meeting of the Microscopy Society of America
`Copyright © 1994 by MSA. Published by San Francisco Press. lnc., Box 426800, San Francisco, CA 94142-6800, USA
`
`PUBLISHER’S NOTICE
`
`The Microscopy Society of America (MSA) and San Francisco Press, Inc., are not responsible for
`the information and views presented in this volume by the several contributors.
`Permission to reprint portions of this volume in professional journals published by scientific or
`technical societies is automatically granted; for reprint rights given to book publishers, commer-
`cial journals, company publications, and other media, apply to the Society's secretary.
`
` Back Volumes
`
`a San Francisco Press is offering all MSA, ABM, and MAS proceedings published before 1990 at
`a special low price of $10 each, while they last. The following volumes are available:
`MSA: 1983, 1987, 1988, 1989 at $10 each
`AEM: 1981, 1984, 198'? at $10 each [Note: XII ICEM vol. 2 (see below) serves as the 1990
`
`ABM]
`
`MAS: 1976, 1977, l9't'9, 1980. 1981, 1982, 1986, 1987, 1988,1989 at $10 each
`
`0 Proceedings published in 1990 through 1993 are offered at half price, as follows:
`MSA 1990 (XII ICEM), four volumes, also available individually: (1) imaging sciences, (2)
`analytical sciences, (3) biological sciences, (4) materials sciences, $32.50 each, $130 the set
`
`MSA 1991, $32.50
`
`MSA 1992, two parts (also contains MAS 1992 papers), $65 the set
`
`MSA 1993, $40
`
`MAS 1990, 1991 at $32.50 each
`Add postage and handling of $3 for first book, $2 for each additional book (foreign orders, $6 and
`$4); Californians add sales tax. (No credit cards.)
`
`Copies of this 1994 Proceedings are available to members of MSA and MAS who send prepay-
`ment by persona! check at $56 (a 20% discount from the list price); orders from organizations are
`filled at the list price, $70.
`‘
`
`Aiso available from San Francisco Press:
`Electron Microscopy Safety Handbook (Barber & Mascorro, Eds.), 2nd ed. (1994), $15
`Electron Microscopy and Alzheimer's Disease (J. Metuzals, Ed.), $10
`Early History ofrhe Electron Microscope (L. Marton), 2nd ed. (1994), $15
`Eiectricity and Medicine: History of Their Interaction (Rowbottorn & Susskind), $30
`Microstrnctu re ofMaterials (K. M. Krishnan, Ed.), Gareth Thomas Festschrift, $30
`
`Sena’ orders to:
`
`San Francisco Press, Inc., Box 426800, San Francisco, CA 94142-6800, USA
`
`1 !
`
`Printed in the U.S.A.
`
`ISSN 0424-8201
`
`

`
`FAST AUTOMATED MOSAIC SYNTHESIS METHOD FOR 2-D.-"3-D IMAGE ANALYSIS 0]?
`SPECIMENS MUCH WIDER THAN THE FIELD OF VIEW
`
`D.E. Becker*, H. Ancin*, B. Roysam* and J.N. Turner**
`
`*Rensselaer Polytechnic Institute, Troy, New York 12180-3590
`“Wadsworth Center for Laboratories and Research, Albany, NY 12201-0509
`
`We present an efficient, robust, and widely-applicable technique for computational synthesis of wide-area
`images from a series of overlapping partial views. The synthesized image is the set union of the areas covered
`by the partial views, and is called the "mosaic". One application is the laser-scanning confocal microscopy
`of specimens that are much wider than the field of view of the microscope. Another is imaging of the retina]
`periphery using a standard fundus imager. This technique can also be used to combine the results of various
`forms of image analysis, such as cell counting and neuron tracing, to generate large representations that are
`equivalent to processing the total mosaic, rather than the individual partial views.‘-2
`
`The synthesis begins by computing a concise set of landmark points for each partial view. The type of
`landmarks used can vary greatly depending on the application. For instance,
`in the retinal
`imaging
`application, the vascular branching and crossover points are a natural choice. Likewise, the locations of cells
`in Figs.
`1 and 2 provide a natural set of landmarks for joining these images. In the cell counting application,
`these locations would be computed anyway.‘ Even when no such ‘obvious’ landmarks are available, image
`processing can often provide useful landmarks.’ The next step in the processing aligns pairs of partial views
`by searching for possible correspondences between landmarks.‘ Associated with each set of correspondences
`is a spatial transformation (translation, scaling andlor rotation) relating the two partial views. Knowing which
`landmarks correspond allows us to transform one partial view to align it with the other. Feasible
`transformations are evaluated by counting the number of points that coincide in the aligned images, and the
`best possible transformation is computed. The two partial views are then merged, and duplicate landmarks
`are removed. Further partial views are combined with the resulting larger partial view similarly, until a
`mosaic consisting of all the partial views merged together results.
`
`The data collection requirements of this method are minimal. It only requires each partial view to have an
`adequate overlap with the adjoining partial view. Adequate overlap is needed to ensure a sufficient number
`of common landmark points. The partial views can be irregularly transformed relative to each other.
`Figs.
`1 and 2 are 2-D projections of representative confocally-acquired 3-D images of dopaminergic neurons
`from a rat's substantia nigra.5 These images were processed with an automated cell counting system.‘ The
`numbers on these figures indicate locations of detected cells. Fig. 3 shows the mosaic synthesized from Figs.
`1 and 2. This gives an accurate estimate of the cell counts in the combined area. The mosaic generation took
`less than one second on a Silicon Graphics workstation. Fig. 4 shows one of six retinal partial views that were
`acquired at differing magnifications by directing a patient
`to follow a movable lighted target. The
`superimposed dots indicate detected landmarks corresponding to vasculature branches.-‘crossovers. The
`resulting mosaic is shown in Fig. 5.‘
`
`References
`
`1. B. Roysam et al., J. Microscopy, l'.~’3:2(l994)l03.
`2. A. Cohen et al., J. Microscopy, l73:2(l994)1lS.
`3. W. E. L. Grimson, Object Recognition by Computer, Cambridge:MIT Press (1990).
`4. L.G. Brown, ACM Compuririg Surveys, 24:4(l992)326-376.
`5. M. A. Chishti et al., Proc. Ann. M02. 0fSoc. Toxicol. (1993).
`6. Acknowledgements: Dr. H. L. Tanenbaum, D. H. Szarowski, NSF, DEC, AT&T Foundation.
`
`224
`
`

`
`Figs 1-2: Projections of 3-D confocal microscope images of rat neurons. White numbers indicate cell locations.
`Fig 3: Mosaic resulting from synthesis of‘ Figs.
`1 & 2. The combination process is totally automatic.
`Fig 4: Representative partial view of retina taken from conventional fundus camera. Black dots are landmarks for
`matching.
`Fig 5: Mosaic resulting from 6 partial views of the type shown in Fig. 4. Partial views have differing scales and
`locations.
`
`225