|
|
|
|
EFAN Cell
2&3 Workshop - Universitat de les Illes Balears
on the
“Use of digital
tools for exchanging
annotated images of
calcified structures”
4
– 5 October 2000
under the auspices
of the EC
Concerted
Action
(FAIR
PL.96.1304)
|
|
EFAN
Report 8-2000 |
" Use of digital tools
for exchanging
annotated images of
calcified structures "
Proceedings of a EFAN Cell
2&3 workshop held at Universitat de les Illes Balears,Palma de Mallorca,
Spain,
4 – 5 October 2000
Editor: Hervé Troadec
IFREMER DRV/RH
BP 70
29280 Plouzane, France
Tlf. - 33-298224388
Fax- 33-298224547
e-mail: htroadec@ifremer.fr k
|
European Fish Ageing Network
(EFAN) |
|
Phone: (47) 37 05
90 00; Fax: (47) 37 05 90 01; Email: bente.lundin@imr.no Office address:
Flødevigvn. 49, Hisøy (Arendal), Norway. Coordinator:
Erlend Moksness, Phone (direct): (47) 37 05 90 41; E-mail: moksness@imr.no |
CONTENTS
I. Introduction.......................................................................................................................... 34
A. Background...................................................................................................................... 34
B. Goals for digital image users.............................................................................................. 34
II. Creation of a digitised reference
collection............................................................................. 45
A. Software choice................................................................................................................ 45
B. Image digitisation.............................................................................................................. 45
1. Image resolution and colour versus black
&white........................................................... 45
2. Viewing limitations........................................................................................................ 56
C. Image annotation.............................................................................................................. 56
1. Image format................................................................................................................ 67
2. Use of digital images for control
collections:................................................................... 89
3. Use of agreed digital collections for
training................................................................... 910
4. Digital validated collections........................................................................................... 910
5. Use of digital images for exchanges............................................................................... 910
6. Use of digital images for workshops.............................................................................. 910
III. Prospects for the future....................................................................................................... 1011
A. Need for custom software............................................................................................... 1011
B. Intensification of internet use............................................................................................ 1011
IV. References......................................................................................................................... 11
V. List of participants.................................................................................................................. 11
List of
participants
Table 1: Participants of the Cell 2&3 2000
Workshop held in Palma (Spain)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
I.
Introduction
The use of digital images for age estimations
of calcified structures can provide many benefits. Their ease of circulation and exchange should facilitate dialogue
at national and international levels on methodology and interpretation
protocols (see glossary of EFAN report 3/98).
The methods and technologies used in the
capture and processing of digital images is a dynamic area that requires
researchers to exchange information regularly. Hence, guidelines provided below
are frequently supposed being updated.
As a first step in the interpretation of
calcified micro and macro structures, used in the determination of age, a
detailed analysis of each structure is required. Analyses should identify and
describe basic structures and their associated problems and incorporate these
in the development of general, standardised interpretation procedures.
Resulting protocols should be objective and
comparable, recognising intra and inter specific structural variability in
their construction and incorporating recognised terminology in the formulation
of instructions.
Such objectivity will assist in the resolution
of subsequent, subjective interpretation of the image (e.g. subjectivity due to
species life histories, geographical areas of occurrence) and the associated
age estimation of the calcified structure.
A. Background
EFAN Cell
2 (Information Processing), a collaboration of software developers and users of
digital image processing software, developed simple protocols for the exchange
of interpreted digital images as a result of a workshop held in the University
of Porto (Portugal, 1998; EFAN Report 3-98). These protocols were used and adapted in
a subsequent workshop on whiting otoliths (Hirtshals, Denmark) where a
reference collection of annotated images was created (Newton, 1998). These
protocols are in a continuous state of update and development and depend on
continued communication between users and software developers for their
successful application (EFAN Report 1/2000,
Iraklion, 1999).
B.
Goals for digital image users
·
The
establishment of a digital image reference database of calcified structures
(e.g. otoliths, statoliths, scales, spines etc.) of species aged with computer
assisted age estimation tools (semi-automatic or automatic). Such a database
should preferably be web based for maximum accessibility and include control,
agreed and/or validated reference collections (e.g. EFAN Cell 2 “Digital Image Reference
DataBase of Interpreted Calcified Structures”),
·
The
establishment of digital image exchange programmes using standardised
protocols,
·
The
organisation of ad hoc workshops to
resolve interpretation problems,
·
The
initiation, where possible, of dialogue between the users and developers of
image analysis software.
II.
Creation of
a digitised reference collection
A. Software choice
The choice of a specific software package must be guided by the kind of use that is required (e.g. creation of the image set, interpretation of images, automatic post-processing of the results) and by the tools that are required to produce a correct interpretation (e.g. measurements, back-calculation, focusing). Software available on the market can be divided into three types :
1. graphics packages or photo retouching packages providing tools enabling basic image and graphic overlay manipulation (e.g. Paintshop-Pro, PhotoShop). They allow the marking of basic structure information required for calcified structure interpretation. Digitisation is generally limited to TWAIN drivers that are not sufficiently flexible in most cases require the use of more specialist imaging packages.
1.scientific
imaging packages provide the features described in (1) along with sophisticated
digitising capabilities and measurement tools. Some packages provide functions
more or less adapted to age estimation on calcified structures (e.g. Image-Pro,
NIH, Optimas, Visilog).
2. specialised imaging software can provide in addition to all the features described in (1) and (2) more specific functions like real-time back-calculation, ring location, curved profile acquisition, mosaic building (e.g. OTO, Bony-Parts, TNPC/Visilog). They are also able to manage basic structure co-ordinates, which can be of a great help for post-processing of large amounts of readings.
B. Image digitisation
The digital image of a calcified structure is only a representation of the ‘operators’ preferred view for interpretation. The ‘operator’ may be a group of experts or a single technician. Associated with each reference collection, files should be provided detailing the source of the images and the method used to capture them. As a collection evolves, it is important that the images within that collection remain comparable regarding acquisition conditions.
1. Image resolution and colour/Black &
White
Sensor resolution is constantly increasing but digital images are still not of a quality comparable to that observed by the human eye through a microscope. The specification of the otolith collection should attempt to provide images with the best resolution and the best dynamic range. Cameras providing images in excess of 1Mb are becoming more widely available and must be preferred to those of lower resolution. The choice of colour or black and white images should be made after discussion with expert readers who can give guidance on their use of colour gradients when making interpretations.
2. Viewing limitations
When an age reader is working with an original sample it is possible to alter all the viewing parameters e.g. lighting, orientation, magnification or focus, using digital images restricts the parameters that can be altered by each reader. This effectively restricts the reader to choices made by whoever digitised the sample.
1.1
Focusing
problem
One of these limitations is that digitising only provides images in one focal plane. The choice of focal plane can vary according to the operator and can be considered as subjective. Today there is no solution available to this problem but ideally a sequence of images could be acquired while varying the focus i.e. in different focal planes. If measurements are required, a microscope with a motorised Z-axis will be necessary in order to precisely interpret this third dimension. This will produce a huge image file and its ease of manipulation will rely heavily on the memory (RAM) available on the computer. Additionally, it will require some programming and tests prior to validation and approval.
1.2
Limited
field of view
An image may be a single captured image showing the whole structure or a mosaic of captured images (elements) joined to form the whole structure. When creating a mosaic image it is important that as many of the viewing parameters remain the same for each element of the mosaic. Mosaics can be built by manually overlaying each element (e.g. with a graphics or photo retouching package) or by using automatic mosaic builders, which attempt to find common areas. When using automatic mosaic building systems the stability of the viewing parameters is more important than that when building mosaics manually. Altering the viewing parameters can reduce the likelihood of an automatic system finding a correlation between elements.
1.3
Image format
Digitisation should be made in a format that does not result in degradation of image quality. A lossless image format is preferred e.g. TIFF LZW and users should be aware of the manner in which their software manages file compression.
1.4
Image
ownership
With increasing use of digital images the ownership of an image may become an issue. For each collection of images, ownership should be detailed and their allowed use specified. Digital watermarking of images is a way of indelibly stamping ownership into an image without materially affecting it.
C. Image annotation
When collecting images for a reference collection the previous comments regarding operator preference should be given consideration (part II-A and II-B).
The image collection
will be composed of calibrated digital images with a resolution that still
enables the reader to estimate the age. A graphic interpretation will be
associated with each image in order to describe the location of each basic
structure identified by the reader as being required for its age estimation.
Rejected information produced by the interpretation process (e.g. false annuli)
will be also included in the graphic information.
Large numbers of images can become a problem to manage. To aid the manager or co-ordinator an agreed file-naming convention should be adopted as one of the first steps in building an image collection. The convention should take into account the platforms across which the images may be used as some systems may not support extended filenames.
1.
Image format
If no measurements or back-calculations are required from the image collection then Paintshop-Pro (version 5 and later) can be used as a tool for producing annotated images. Otherwise the selection of software will be governed by the tools required (see II-A), and similar procedures as described for Paintshop-Pro can be applied.
The PaintShop-Pro file format (version 5) ™ (JASC Corporation, http://www.jasc.com) will be used as the file format
with additional layers containing symbols and/or text annotation regarding the
interpretation of the image.
Graphic interpretation : Layer Descriptions
The layer identified as ‘Background’ contains the image
of the calcified structure. When there is a need to merge layers from different
files, a square reference point will be marked in an area away from the
structure (Figure 1). This reference point must be duplicated in each new layer
created by readers. The reference point allows accurate overlaying of marks
from different readers into 1 image file for comparison.
Figure 1 Layer identified as "Nucleus", identifies the position of the
nucleus with either a dot or a ring. Layer identified as "Rings" contains the agreed checks or rings
that the age reader should be attempting to identify. The rings are be
identified with dots that contrast with the sample colours. Layer identified as "Info" highlights other features within the
image that the reader feels would aid other readers when interpreting the
visible structures e.g. false annuli.
These identified features should be backed up with remarks in the text
layer. Layer identified as "Text" contains remarks relating to
features highlighted in the info layer and comments relating to the image and
its interpretation. Co-ordinators should consider the
facility to use different colour marks for the parameters they wish to study
(Figure 2) e.g. For a between reader analysis a different colour would be used
for each reader. Using different colours will also facilitate the
post-processing of exchange exercises.
Figure 2 In addition to the
image files, there will be a data file for each image set. This file contains
information about the sample, how it was prepared and information on the
equipment used to capture the image. The data file will be CSV format with 1
line per image file. Data Required : 1.
Submitting
institute 2.
Species: Latin name 3.
NODC code 4.
Institute assigned
identifier -- Allows the original sample to be re processed if necessary. 5.
Type of hard part
(Otolith, Scale, Statolith, Spine, Vertebra, Tooth) 6.
Length of
individual (mm) 7.
Sex of
individual (M,F,U) 8.
Maturity stage 9.
Estimated age 10.
Known age 11.
Age units (Years,
Days, Seasons) 12.
Date of capture
(DD/MM/YY) 13.
Position of capture
14.
Image presented by
(Contact person in institute) 15.
Preparation method
of the sample 16.
Lighting method
(Reflected, Transmitted, Polarised, Filtered) 17.
Camera type 18.
Frame grabber type 19.
Image resolution at
capture 20.
Magnification
(pixel size or calibration for measurements) 21.
Pre-processing
completed before submission 22.
Filename -- note
possible problems with long filename support on some platforms 23. File size -- Kb Digital control collections allow
a greater flexibility for quality assurance (QA) procedures as they enable the
control of a number of parameters that can affect an age estimate (e.g.
lighting, sample degradation). The
problem of sample degradation over time, which reduces precision, is solved
with the use of a digital source. A
digital file will remain static for as long as its storage media is viable. The
image file would remain in a read-only state on a master device with copies
being made each time the image is needed. By providing annotations in addition
to age when interpreting a control collection the reader can give the QA
manager a way to investigate the source of any detected imprecision. The image format used for control collections will need,
as a minimum, the ’Background’ layer (i.e. the basic image of the structure)
and the ‘Rings’ layers. There must be
also in place measures to keep track of information on the reader and the date
of the estimation (e.g. this may be in the form of agreed filenames). A larger control collection would reduce the problem of
‘reader memory’ memory’ (i.e. recognition of individual otoliths and the recollection
of prior age estimates) but requires
larger manpower input at the beginning of the process to capture images
and provide a first estimate for each reader.
Although the viewing parameters were fixed when the image
was captured it is possible for an age reader to adjust the image by the use of
filters (e.g. brightness, contrast, etc). Due to the effect that digital filters can have on bias in the estimations,
the use of such filters should be considered as part of the reading protocol. When a reader is allocated a sample of digital images to
read they will be given copies of the master image containing only the
‘Background’ layer. In the process of making their estimation, they will create
a new ‘Rings’ layer duplicating the reference square. They will then mark their interpretation of the structure onto
the new layer and save the file. Once
they have completed all the samples required the marked files will be returned
to the co-ordinator with their age estimations. If a problem in precision arises the co-ordinator can discuss the
problem with the reader using their interpretation and historically marked
images. Because the images within an agreed collection contain the
results of multiple interpretations of the structures they become useful to new
readers to see a wider range of interpretations without the need for all the
experienced readers to be present. The image format for this type of collection should use
all the layer types described above (i.e. IIC1 et seq.). The ‘Rings’ layer should contain the agreed
interpretation with additional ‘Rings’ layers containing the interpretation of
each of the agreeing readers. Validated collections are an extremely valuable resource.
Digital versions of validated collections would allow much wider use of these
collections without the inherent risks involved in transporting original
samples. The image format for validated collections should include
all the described layer types with text layer including information on or
references to the method of validation. One of the major problems with standard sample exchanges
is the time-lag as the samples are passed to each reader. In order to increase
the speed of the exchange, database and software access might be proposed in
different ways: ·
By sending the database on CD-Rom (650Mb) or DVD-Rom
(4.7 Gb), ·
By downloading the database from an ftp site. Co-ordinators should survey the participating institutes
to identify a suitable level of hardware
compatibility (e.g. the media type to be used for distribution, the
operating platform to be used). Digitisation is the first stage in all uses of images and
should form part of the introductory session of the workshop. The protocols used by individual institutes
should be available for discussion and the co-ordinator should consider whether
it is appropriate to digitise a sample of the structures during the first part
of the workshop. The images that will be used in the workshop should, where
possible, be digitised to a previously agreed standard. The co-ordinator should consider the use of networked
facilities for holding the source images, in a read-only form, and the result
images of the workshop sessions. These
files must be included in routine secure backups. At some point during the workshop it will become necessary
to discuss images and results in a plenary session and facilities such as a
large screen projector connected to a PC should be available. If information such as length, region of capture, season
of capture are considered important factors in the estimation of age, it should
be made available in an accessible form.
It may be appropriate to include the information as part of each image. Using collection management software it would be possible
to allocate samples to be read by a selected reader, pass those images to the
reader, collect the resulting age estimations and compute agreement levels for
constructing an agreed collection. The management software would assist the
post-processing of the interpretation results (e.g. compare ring locations
automatically, track and report changes in precision and provide additional
analyses). A more intensive use of internet would provide a solution
for many current problems : ·
A virtual workshop would ensure a quick and cheap
collection of the data, ·
Standard web browsers may be the solution to the
hardware compatibility problems, ·
Remote access to existing software via Web conferencing
software (e.g. NetMeeting, Microsoft ä) or to a specific
Web-designed software would ensure
interpretative tool harmonisation of tools and protocols and eventually
more in a more efficient post-processing of the data. EFAN, 1998. Report of the EFAN
Cell 2 Workshop on the " Establishment
of a digital image reference database” held at the University of Porto,
28-29 September 1998, EFAN Report 3/98, Editor: H. Troadec, 42 p. EFAN, 1999. Report of the EFAN
Cell 2 Workshop on the Contrast enhancement and Newton, A. W., 1998. Report of the ICES/FAIR Otolith
Ageing of North Sea Whiting Working Group at Hirtshals, Denmark October 1998. Table 1: Participants of the Cell 2&3 2000
Workshop held in Palma (Spain) First name Last name Country Richard Ayers UK Manuela Cunha Portugal Peer Doering-Arjes Germany Gianfranco Giannetti Italy Toni Lombarte Spain Beatriz Morales-Nin Spain Ana Moreno Portugal Andrew Newton UK Conor Nolan Ireland Sandra Ramos Portugal Filipa Ribeiro Portugal Hervé Troadec France 
1:
Reference point in the 'Background' layer.
2:
Schematic representation of a
calcified structure image annotated with graphic layers.
External data file
2. Use of digital images for control
collections:
3. Use of agreed digital image collections
for training
4. validated digital image collections
5. Use of digital images for exchanges
6.
Use
of digital images for workshops
III.
Prospects
for the future
A. Adapted software requirement
B. Intensification of internet use
IV.
References
pre-processing of
calcified structure images” held at the Candia Maris Hotel Heraklion,
Greece, 4-5 October 1999, EFAN Report 1/2000, Editor: H. Troadec, available at
:http://www.efan.no/rep2000/C2Rep99.htm.V.
List of participants