11/1/96

discussion of the types of software that investigators are using for the

preparation of figures for publication (i.e. taking multiple confocal

images and organizing them into a single figure with appropriate labeling,

scale bars, etc.). As you might guess I am currently teaching myself how

to do this and in the process have discovered a number of inadequacies in

the software I'm using. For example, Photoshop is fine for making

adjustments in image size, resolution, contrast, etc., but when it comes

to applying labels and lines pointing to objects in the images, it leaves

much to be desired since it's not an object oriented program (i.e. even

using layers, modifications to labeling take longer than they would if the

label could be reselected and modified). At any rate, some of the

questions that come to mind are,

1. Are programs such as Adobe Illustrator a good choice for preparing

figures? If not, what do you suggest and why?

2. What sort of resolution (dpi) is it best to format images at for

eventual printing?

Thanks for any input any of you might have.

Steve Kempf

kempfsc@mail.auburn.edu

for dealing with bitmaps. We use it for overlaying channels, boosting the

blue for printing (that's another story all together), for adjusting

contrast and brightness, and other occasionally for noise reduction

(filtering). We then use Canvas from Deneba. We continually evaluate the

software world and consistently come back to Canvas for scientific

graphics. It has the easiest tools to access (i.e. not hidden under many

menu layers), it understands bitmaps and objects. It has the best

align/distribute functions of any software that we have used (quite useful

for labelling multiple panel figures) and it generates clean export files.

Canvas 3.5 is a standard for all of computers, and yes, it is available

for WINTEL machines.

We are just starting to explore the latest version (Canvas 5.0) which

includes many of the functions that we use Photoshop for, including

accepting Photoshop plug-ins. Unfortunately, some of the robustness of the

printing functions has been lost and will hopefully be fixed in the near

future. If you do get v5.0, be sure to get the patches from Deneba's web

site (www.deneba.com).

Paul Goodwin

Image Analysis Lab

FHCRC, Seattle, WA

pgood@FHCRC.ORG

We use CorelDraw. It seems to do everything we want to do wrt

labelling and combining different sorts of figs.

Regards

Mark Cannell

mcannell@SGHMS.AC.UK

and Canvas (excelent!!!. Besides the functionality Paul talked about, you

can almost use it instead of Powerpoint for slide preparations as well).

Between the three we can do everything we ever wanted to do (and more...)

Stamatis Pagakis snp@physiol.ox.ac.uk

University Laboratory of Physiology stamatis.pagakis@physiol.ox.ac.uk

University of Oxford, Parks Rd. FAX: +44 (0)1865 272469

Oxford OX1 3PT, United Kingdom Tel: +44 (0)1865 272551

exporting them to Harvard Grapphics or Power Point for

lettering/labeling, and re-exporting as . tif for printing on a

Codonics.

Exporting and re-exporting does not degrade the quality of the tif images.

__________________________________________________________

Renato A. Mortara

Disciplina de Parasitologia

Departamento de Micro, Imuno e Parasitologia

ESCOLA PAULISTA DE MEDICINA

Universidade Federal de Sao Paulo

Rua Butucatu, 862 6o andar

04023-062 Sao Paulo SP

Brasil

Email: epmpara@eu.ansp.br

FAX 55 11 571-1095

PHONE: 55 11 570-8306

We have had success using a combination of Photoshop and Micrografx' Picture

Publisher (now sold as part of a suite of products). Picture Publisher uses

objects, and one can work with dozens of them (one each for each image in

the montage and one each for each label and for scale bars etc.) before

combining them. We would use Picture Publisher exclusively, were it not for

the fact that it does not resize images well - probably because the

interpolation function is poor. Photoshop interpolates beautifully.

Picture Publisher also has thumbnails, which are VERY useful, and it allows

you to open multiple images at once. In general, PP has much better file

management features than Photoshop. Maybe the new version of Photoshop

(4.0) will be better in this respect.

Hope this helps.

Peter Sargent

sargent@ITSA.UCSF.EDU

advantage of the new version of Canvas (Version 5.0) is that permits use of

Postscript text as well as 24 bit bitmapped images. Photoshop has three

major flaws when used for this purpose:

1) If you put the text on the same layer as the image, it permanently

obliterates the underlying pixels of the picture.

2) Although Photoshop 3.0 attempted to deal with this by permitting

additional layers, thus allowing you to enter text without destroying the

image, and then move it around, this is still not vector based, or even

PostScript, text.

3) Almost as bad, when you then add additional layers to in PhotoShop to

prevent destruction of the original image, the file doubles in size as soon

as you add text, since the new layer is also another RGB layer of the same

dimension. This can be a disaster when dealing with 26 MB images as in Leaf

Lumina files. Adding the simplest text increases the file to 52 MB. If I

add another layer to store information about the image, but don't want it

printed in the final form, that increases the size of the file to 72 MB.

Canvas 5.0 solves this very nicely. The bitmapped 24 bit image is retained

as a bitmap. You can then lay on text in vector format which can be moved

about on top of the original image, without degrading it. Alternately, you

can (and probably would be wise to do so) can make a separate layer for the

text. All this only increases the size of a 26 MB file to 26.01 MB.

An important reason for adding the text in a separate layer is that if you

make the mistake of rotating the text, it can convert it to a bitmapped

item. This might overwrite your original valuable microscope image.

Canvas 5.0 is definitely the way to go for image preparation. But tread

carefully. It is a new version, and is still buggy. Back up everything!

Harvey J. Karten, M.D.

Dept. of Neurosciences

Univ.California San Diego

La Jolla, CA 92093-0608

EMail: Kartenh@sdsc.edu

also: HJKarten@ucsd.edu

Phone: (619)-534-4938

FAX: (619)-534-6602

Retinal WebSite: http://www-cajal.ucsd.edu

comments (I appologise if I'm stating what you might consider obvious, but

it is based on experience):

1. First, check with the publisher/printer. Some basic questions at an

early stage can avoid a lot of problems later on.

2. Scale bars - PLEASE. I continually receive micrographs without scale

bars marked on the micrograph. Whilst it is sometimes possible to add scale

bars during production, often it is not and this is usually at a point when

deadlines are approaching.

References to magnifications aren't quite useless but nobody knows the

final scale of reproduction in the publication until quite a late stage - I

frequently have to recalculate magnifications during final proof reading by

comparing the proof with the original.

On occassions, authors quote what seems to be the instrument magnification

at image acquisition and then supply an 'original' which has clearly been

enlarged.

With electronic image formats, images can be visualised at any scale and

talk of magnifications really is useless.

3. Don't try to be too clever. For example, don't supply text files with

embedded illustrations. Even if the editor and the printer are using the

same platform and same applications as the author, it still has to all be

taken apart and layed out again according to 'house style'. Don't provide

DTP-type files - it will cause problems. Supply a basic word processed file

for the text and separate files for images and illustrations.

4. Use as standard a file format as possible - you may have access to a

sophisticated application which has its own very efficient file format, but

the publisher/editor/printer probably don't!

Generally, I would recommend a TIFF format for images but be aware that

there are several varieties of TIFF. In addition, there are RGB and CYMK

varients for colour TIFF - use RGB TIFF. The printer is expecting this.

For illustrations, EPS is probably the most widely acceptable.

Note that both TIFF and EPS are inefficient in regard to the amount of

space they occupy. You may want to consider compression but either check

with the publisher first, or make your compressed archive self-extracting.

5. Always supply hard copy, both for text and for figures. Figures don't

have to be 'photographic' quality but the essentials should be obvious. If

you don't, be prepared to see, for example, some odd colours.

With regard to the text, always provide an ASCII version of the file. It

probably won't be used but I find it a lot easier to open an ASCII file

than to extract text from a file format that I can't read - and most

publishers/editors/printers won't even try. And some text file formats

store the data in a way that isn't accessible via file editor applications.

6. Be aware that many printers are Apple Mac based - the Mac is much

commoner in the printing industry than among general users. If you can

supply a Mac version on a Mac formated disk, everything will be a lot

easier.

7. With regard to image 'resolution', again the first step is to ask the

publisher. MICROSCOPY & ANALYSIS, for example, prints full colour images to

a high standard - other publications work to lower standards, usually

because of cost considerations.

However, almost any 'professional' publication will be printing at a much

higher resolution than most microscopists would get anywhere near 2000 -

3000 dpi for typesetting.

In addition the whole issue is complicated by scaling of images at various

stages and dpi is relevant to output but total pixels are what is important

for the microscopist.

8. Colour - try to avoid non-RGB colours, e.g. Pantone colours. They may

look attractive but you're more likely to run into problems with the

printer. OK, the printer probaly will handle Pantone OK, but thats a

special colour for printing and the publisher almost certainly won't want

it - because it'll probably cost extra!

Dr L. P. Stoter Technical Editor, MICROSCOPY & ANALYSIS

Technesis

17, Rocks Park Road email: LPS@teknesis.demon.co.uk

Uckfield, E. Sussex Phone: +44 (0)1825 766911

TN22 2AT Fax: +44 (0)1825 766911

United Kingdom

our confocal images and our 12-bit CCD images. The image

analysis is performed on an SGI using VoxelMath (that's your

program isn't it Steve?) and Photoshop is one of the only PC-based

imaging programs that can handle the 24-bit TIFFs we get from

the SGI. We get the best results, however, when we do everything

on the SGI using the "Showcase" application. The only problem

is that we don't have a high resolution color printer or slide

maker connected to the SGI so we have to save the files as

EPS and send them out or (try to) port them into another

program that has a driver for the slide maker. As you can

see, this can get quite messy. You never know what you have

with an EPS file until you print it out!

Tracy Richmond McKnight email: richmon@ece.ucdavis.edu

Department of Human Physiology TEL: (916) 752-5584

School of Medicine, MS-1A FAX: (916) 752-5423

University of California

Davis, CA 95616

professional designers and publishers usually use. Some journals, e.g.

"Development", explicitly asks the contributers for composing digital

figure using either of these two programs.

Advantages:

1. These programs handle figures directly in the postscript format, which

is the de-facto standard in the publishing and printing industry. Since no

format conversion is needed, they are accurate and cause less problems when

you submit your electronic data to the DTP folks.

2: When you place images (confocal and other photos) in the figure, these

programs keep the original image data separately. The figure has only

"links" to the originals. This is good, first: even if you place megabytes

of photos in your figure, the program only handles the low-resolution

preview images, thus the program does not slow down. And second: the

publisher often want to modify the original bitmap data for good

reproduction; this is impossible if the photos are integrated in the

figures.

3: The postscript image format is among the most precise. These programs

offer better and more precise control for text handling, gradation, colour,

etc.

Disadvantages:

1: You need a postscript printer to print the data. Many color printers,

such as inkjet printers and slide makers, may not be able to handle this

format.

2: Though using Freehand is as easy as Canvas or Corel Draw, Illustrator

may have a bit steeper learning curve.

3: They are mostly Mac-oriented. As Dr. Stoter puts it, the apple system is

disproportionaly common in the publishing industry.

Several people have recommended Canvas and Corel Draw. They are good

programs and should be useful enough 1: if you don't intend to submit the

digital data to the publisher, and 2: if you don't have megabytes of images

placing in the figure. Although Canvas and Corel Draw are able to save data

in the postscript format and to place large images, they don't perform

these tasks as perfectly, precisely and quickly as Freehand and/or

Illustrator do. Whether you should go for the later two programs depends on

what you intend to do. But I would like to say that using Freehand and/or

Illustrator is not as difficult as some people might imagine. They are

actually fun to use.

2. What sort of resolution (dpi) is it best to format images at for

eventual printing?

An image setter that publishers use usually has the resolution of 1270 DPI

or 2540 DPI. Characters are printed at this resolution. To print colour and

B/W photos they "half tone" the image at the screening resolution of 150

LPI (line per inch). Some journals use 133 LPI for worse-looking but

cheaper reproduction, while some may use 200 LPI for better reproduction.

The rule of thumb is: the resolution of the original data should be two

times higher than the screening resolution. If the publisher used 150-LPI

reproduction, you need 300 PPI (pixels per inch) resolution. If the width

of the final figure has 3 inch in the published journal, you need 900

pixels in the original data.

If the resolution (PPI) is higher than "2 x screening resolution (LPI)",

you have too large data without increasing the quality of the final print.

Photo CDs and slide scanners can generate 2000 x 3000 pixels of image data,

but they are too large for most purpose.

Terms:

DPI: dot per inch : the resolution of printers and image setters

(how many dots per inch it can place on the paper)

PPI: pixel per inch: the resolution of the image data

(how many pixels you intend to print)

LPI: line per inch: the resolution of the screen (mesh) for reproducing

half-tone images

In the CD ROM of the Photoshop program, there is a very nice introductory

course about resolution. I would recommend to go through this tutorial.

Hope this helps,

-----TU NE CEDE MALIS, SED CONTRA AUDENTIOR ITO.------

Dr. ITO, Kei

ERATO, Mitsubishi-Kasei Institute of Life Sciences,

194 Machida-shi, Minami Ooya 11, Tokyo, Japan

Tel: 0427-21-2334 (from Abroad: 81-427-21-2334)

Fax: 0427-21-2850 (from Abroad: 81-427-21-2850)

email : itokei@fly.erato.jst-c.go.jp

WWW:

http://papageno.jst-c.go.jp/ (Jfly home page)

http://papageno.jst-c.go.jp/brain_k_ito.html (Fly sections)

http://flybrain.jst-c.go.jp/ (Flybrain visual database)

publication and have converged on a mixture of programs that work well. We

aim to have publication qaulity hard-copy prints, since that is what most

journals still want. Things would almost certainly be different if we had

to submit digital files for publication in the Desktop Publishing mode. We

use digital imaging for both our widefield fluorescence and confocal

systems, and are now working out ways of doing it for electron microscopy...

Much of our initial image processing is done with NIH Image as 8 bit TIFF

files. We also use a Silicon Graphics workstation for 3D reconstructions

etc. which can produce 24 bit or 32 bit TIFF files. As already pointed

out, Photoshop is the only program we have found that reads the SGI files

properly. Our final image manipulations are done in Photoshop, and then we

use Powerpoint, CorelDraw or Canvas (depending on personal preference) to

make the final plates up, which are printed on a Kodak dye-sublimation

printer.

Although I shudder to say that I like a Microsoft product, Powerpoint

really makes it very easy to make up plates and label them for publication.

All the text and labels etc are vector based and can be moved, edited,

scaled anyway you like without affecting the underlying bit-mapped

illustrations. Also Powerpoint has a surprisingly good (and easy to use)

image scaling routine that does not degrade the original image. It will rad

most standard types of image files, except some TIFF files from the SGI. We

solve this problem by opening them as large images in Photoshop and use the

clipboard to cut and paste them into Powerpoint, and then rescale them in

Powerpoint ( it sounds dumb, but it works really well!). The people in our

Illustration and Media Department like a presentation program like

Photoshop because all the pictures are in one file and they can easily

batch process them.

For publication quality prints, your final images need to be at about 150

DPI or better. This is easy to achieve, if you initial files are reasonably

large (eg 512 x 512 of more), since a rescaling down to usual publication

size will bring you to better than 150 DPI anyway. Of course this assumes

you are using the full image. If you start croppping too heavily, your

original image may not have much marging for rescaling to a smaller size

(and higher resolution).

There can be some problems in rescaling images from video cameras, in that

you can get a series of vertical lines due to noise in the CCD or frame

grabber or both (I forget exactly which: see Russ's Image Analysis Handbook.

.). At full screen size, these bands are barely visible but when an image

is rescaled to 150 or 200 DPI they become very noticeable, especially in

images with lower signal to noise ratios. There is a very easy way to get

rid of this with Fast Fourier Transforms, which we use in the FFT version

of NIH Image. The only restriction is that the images must be square. It

works really well!

We have now published about 5 or 6 papers in the last 12 months where all

the fluorescence microscopy has been prepared in the ways we described, and

we have received very favourable reviews of the illustrations (the text is

something else!!!). The really great thing about all this is when you

rewrite the paper, you can easily edit the figures without going to the

darkroom for hours on end. (We have just about closed down our three

darkrooms...).

A final word in using these techniques for EM. We do have a Gatan CCD

camera in our TEM, so we can get 8-bit digital images straight off the

microscope. These are terrific for working with (we do a lot of serial

section work, so they make quick and easy records of serially sectioned

things..), but the quality is not good enough for publication, mainly

because it seems that the camera has a pretty limited dynamic range

compared with film. So in the end we still use photographic negatives

and prints to get our best EM images. But even so, the easiest way to make

plates is to get 35mm slides of the best images scanned as high resolution

TIFF files and then ship them off into Photoshop / Powerpoint...

PS. I have seen that there is now a CD-ROM Image Processing Tool Kit to

complement the Image Processing Handbook, which turns Photoshop into an

Image Processor... It sounds great. Any one used or seen it yet?

Hope that helps...

IAN

Professor Ian Gibbins Flinders Microscopy &

Department of Anatomy and Histology Image Analysis Facility

Flinders University of South Australia

GPO Box 2100 Adelaide 5001 Centre for Neuroscience

AUSTRALIA

Phone: +61-8-2045271

FAX: +61-8-2770085

e-mail: Ian.Gibbins@flinders.edu.au

the page in Clarisdraw. I find this an easy program to use, inexpensive

and does everything I need to set up the graphics. I have considered

Canvas since you can apparently do much of what Photoshop does, but in our

multiuser facility, Clarisdraw seems less intimidating for new users. Dave

Dr. David Knecht

Department of Molecular and Cell Biology

University of Connecticut

U-125

Storrs, CT 06269

Knecht@uconnvm.uconn.edu

[Return to Tips & Tricks Menu]