,.
.,,
:
t_
:;
PACS Mini Refresher Archival
Image Meryll
M. Frost,
A typical day
Course
Jr, BSEE
radiology
and
much
data
decreases
.Janice
C. Honeyman,
department
as much
which
Technologies1
as
can
1 terabyte
are substantial
create
of data
storage
#{149} Edward
many
per
problems.
data
PhD
gigabytes
year.
One
and
is data
and
MD
of image
Archiving
solution
requirements
V. Staab,
increases
data
per
accessing
this
compression, the
rate
of data
transfer; however, standards are not yet available. Other solutions involve improvements in archival media. Jukebox subsystems allow automated access
to multiple
units.
Digital
magnetic
large amounts of information more practical technologies videotape allows storage of transfer. Optical disks, now a large storage capacity and tape is also being investigated technology to choose depends tution and the cost, stability, tern. U INTRODUCTION One of the ultimate
outcomes
tape,
the
of picture
archiving
and
(PACS) will be the elimination of film as the preferred ogy. Although with present technologic and economic unlikely
in the
near
future,
standard
medium,
can
store
and enables easy updates or replacements; have been introduced in recent years. Digital digital video data and features a high rate of data the preferred permanent archival medium, have provide excellent long-term stability. Optical as a solution to the archiving dilemma. Which on many factors, including needs of the institransfer time, and storage capacity of the sys-
many
pertinent
clinical
communication
systems
medium in diagnostic constraints this goal problems
in image
nadiolis highly
storage
can
be solved available
with existing technologies
PACS technology. In this article, we describe the currently for storage or archival of digital images in a typical PACS and
evaluate
their
capacities
U BASIC Different
storage
CONCEPTS IN diagnostic imaging
quirements. sional
A digital
array
of picture
Abbreviations: read
MO
term:
Picture
RadloGraphics From
1992;
the
Department
Gainesville, Address C
data
transfer
STORAGE modalities that
magneto-optical,
CAPACITY produce widely
would
elements
be stoned
called
PACS
=
characteristics.
pixels.
picture
The
archiving
varying
on a PACS
and
value
digital
consists
stored
storage
at each
communication
ne-
of a two-dimen-
system,
pixel
WORM
=
is a nu-
write
once
many
Index
I
image
and
RSNA,
FL 32610. reprint
requests
archiving
and
communication
system
(PACS)
12:339-343 of Radiology, From
the
1991
University RSNA
of Florida scientific
College
assembly.
of Medicine, Received
November
BoxJ-374 18,
JHMHC. 1991;
1600
accepted
SW Archer
Rd.
November
19.
to M.M.F.
1992
339
approach
merical representation of the magnitude of image data (ie, level of intensity) at the particular x and y coordinates of the original image.
How
accurately
the digital
image
latter
bits,
is referred
10 bits,
quoted
to in numbers
12 bits,
and
ofbits:
referred
plained
in terms
a yardstick.
Ifwe
inch,
are
there
8
age such
reside for some magnetic disks,
storage.
as magnetic
tape,
optical
ex-
something to the
possible
with
nearest
values.
digital
systems
normally
tape.
‘/
Ifwe
and
of the individual
eral also
so
digi-
possible forms been proposed
U DATA In lossless nal digital
of data (3,4).
COMPRESSION data compression, the image can be reproduced
ated that, when decompressed, a close approximation of the
storage, 2 bytes
is i0
Ifwe
consists
bytes,
assume
of 50
5 12 pixels
with
2 bytes
is used
would be A megabyte
512
and
that
images,
lossy
tomographic X
since
the CT image on 0.5 Mbytes.
a gigabyte
bytes.
5 12
but
the
is
average
total
1012
CT study
storage
re-
quired for that study would be 25 Mbytes. If 50 studies are performed per day, 1.25 Gbytes of data will need to be stored.
Recent
research
indicates
tnix of at least 2,048 x 2,048 for digital chest radiography we assume use of 12-16-bit of storage (or the equivalent diskettes)
will
be
If a department and lateral chest Gbytes ofstorage These numbers
required
that
a pixel
ma-
may be required (1,2). Again, if pixels, 8 Mbytes of six 3.5-inch for
that
one
image.
obtains 75 postenoantenion radiographs per day, 1.2 will be required. indicate that the typical
radiology
department
gigabytes
of image
can
data
create
many
pen day,
and
upward
of 1 Thyte of data per year. The problem is, How do we maintain this much digital data in an archive and still have reasonable access to many years worth of digitized images? The digital
image
archive
typically
has
a staged
the
a reduced
is currently
amount
tolerated diagnostic
facIn
set
is cre-
to determine
compression Images
as 10: 1 may
onigifrom its
produces only original image. that
still maintaining
image.
much
exact
data
underway
oflossy
while
have
Lossless compression are readily achievable.
compression,
Research x
is 106 bytes,
a terabyte
the
12
for
x 512
(eg,
compression
example,
computed
media
oped, such as the optical or tape jukebox. To maximize the available storage capacity, sev-
mea-
store
each
compressed form. tors of two to four
contains
on
optical disk, magnetic tape) does not have the total storage volume necessary, methods of handling multiple volumes have been devel-
Most
pixel,
disk,
be
values,
scan
Eventually
can
possible
per
of
in bits
1,024
(CT)
period usually
from the working permanent stor-
tal data in multiples of 8 bits (1 byte). Therefore, 8-bit data are stored in 1 byte, but 10-, 12-, and 16-bit data are stoned in 2 bytes. For
bits
digital
acquisition
optical
measure
a typical
The the
typically
10 bits
forth.
from
are
sure to the nearest ‘/8 inch, there are 288 possible values. The number of bits needed to represent a pixel value in a digital image also depends on the precision of the measurement. Eight bits represents 256 possible values,
retrieval.
16 bits
of measuring 144
and archive
images are migrated to a form of long-term
Because
as measured
the
to as working
these store
by manufacturers.
Precision
enter
system and usually time on high-speed
represents
the original image depends on both the sampling frequency and the precision of the numerical value used to represent each pixel.
The
to storage
images
can
compressed
still
be
an acceptable by as
be clinically
acceptable
(5).
Data
compression
fits. First is the age requirements.
duced
data
can have
direct
several
bene-
reduction in data storSecond, since only a re-
set is sent
to the archive,
an
apparent increase in the rate of effective data transfer is obtained. The negative aspects of compression are the lack of standards and intolerance of data errors. Eventually, data compression will be added to the standards of the committee formed by the American College of Radiology and the National Electrical Manufacturers
until
that
vendor’s general addition,
Association
time
the
compression use of image because
(ACR-NEMA),
proprietary
algorithm compression
of the
ture of a compressed can be allowed from
nature
highly
but
of each
will make difficult. encoded
the In na-
image data set, no errors the archive system. With
an uncompressed image, a single-bit error in an image of several million bits will have no impact on the diagnostic quality of the image. In the case of highly compressed image data,
even
a single-bit
error
can have
devastating
results.
340
U
RadioGraphics
U
Frost
et al
Volume
12
Number
2
U
TYPES
.
Magnetic
OF
developed have been the primary method of software distribution, system software backup, and data interchange. Only in recent years have practical replacements such as CDROM (compact disk, read-only memory) and floppy digital disks been introduced. Al-
MEDIA
ARCHIVAL
Disk
The front end of any will contain a number
imaging
digital
archive
of conventional
mag-
netic disks. Normally, fault-tolerant designs must be incorporated to ensure image data integrity until migration to the permanent archive media has been confirmed. Fault tolerance is usually achieved by one of two
though
methods. The first is disk mirroring, in which the data are duplicated in two identical places. The second method, usually used with multiple-disk-array systems, involves the addition of one or more parity drives. The parity drive allows efficient restoration of data after the replacement of any failed drive.
fixed-head
by
Although
the
storage
capacities
of individual
archival media (eg, digital tape cassette, optical disk) are rapidly expanding, ready access to multiple units will be a necessary part of any imaging archive. To solve this problem,
jukebox
subsystems
cess have been system consists
back
units,
that
allow
designed. of one
storage
automated
The or
typical
more
shelves
Jukebox
systems
magnetic
digital
designed videotape,
and
then
time
usually
.
to the record-playback
access
required
8-20
Magnetic
the data to fetch
data and expanding
programs) industry.
produced In addition,
dardized
method
dates and The digital
March
1992
the
of providing
the
oxide
on the
tape
as the
to the
a signal
originally
this
are
sensed
ideally
recorded
the
The
is
informa-
is grossly
describe
mechanism.
Durhead,
that
explanation
it does
of
by the
patterns
creating
created
material.
passes
recorded
head,
identical tion.
tape
field alignment
over-
basic
storage
recordcapacity
of
that
and
is usually
unit,
The is
of both
by this rapidly an easy, stan-
software
causes
previously the
a magnetic
head
bits
medium
form
the
designs.
and
has been an archive stabeginning of the combeginning, computer dethe need to provide a ofstoring the enormous (in
on
for
Tape
of information
reading,
playback
seconds.
Digital magnetic tape pie almost since the puter age. From the velopers recognized safe, secure method amount
desired
of
a magnetic tape is determined by multiplying the usable length of the tape by the track density and the number of tracks on the tape. The track density refers to the number of digital
on that volume.
the
ing
Although
optical disks. Storage capacities in excess of 28 Thytes are possible with such configurations. In the case ofjukebox systems, the process to access any digital image requires that the robotics locate the correct medium, trans-
fer that medium
head
the head,
simplified,
for the individual
have been tape,
magnetic
jukebox
record-play-
to be a multitude
technologies available all implementations two basic categories,
moving
record
the
by
ac-
and
passes the
the
media, and the necessary robotics to allow fast movement of the media to and from the record-playback units.
digital
may seem
In fixed-head technology, the record-playback magnetic heads remain stationary while the tape is moved mechanically past the heads. The number of heads and tracks on the tape vary from type to type, but the basic pninciple is the same. During recording, as the
tape
Jukebox
.
there
different digital tape the market, in general can be classified into
can
example, as the
be
stored
the
digital
computer
mation
pen
expressed
unit
tape
used
industry
exchange
length
in bits
was
pen for
inch
For
many
standard ‘/2
of tape
inch. for
wide,
years infor-
stored
1,600 bits per inch, and contained nine tracks. The rate of data transfer from digital magnetic tape is a function of the track density
multiplied
the
rate
sec.
Although
can
by
the
be as high this
speed
tape
speed. was
Digital
Kbytes/
adequate
die digital audio needs, another necessary to expand digital tape tal video arena.
.
Typically,
as 200-300
to
approach into the
han-
was digi-
Videotape
The moving-head originally designed this system, both
magnetic tape machine was to record analog video. In the tape and the heads
up-
new software was also necessary. magnetic tape standards that were
Frost et al
U
RadioGraphics
U
341
move, larger
allowing bandwidth.
accommodation Typically,
of a much or more
one
heads are mounted on a rotating drum, the tape is positioned such that contact the drum is maintained for most of the diameter.
in addition,
lower
on one
the
side
tape
an optical sensor reads the reflected light, thereby recreating the original data. Sustained data transfer rates of up to 1 Mbyte/sec for both reading and writing have been demonstrated.
and with drum
One
is positioned
of the drum
and
higher
on
the other. With the head mounted on a drum, the path traced by the head on the tape is a helix, thus the term helical scan tape. Normally, the rotation of the drum is synchronized with the video information such that one video field is stored per rotation of the drum. The D 1 and D2 digital video standards were created to allow the utilization of helical scan techniques for storing digital video information.
The
D2
standard
specifies
data
Optical
Disk
The large storage capacity of optical disks along with their excellent long-term stability have made them the preferred permanent archive medium. Laser optical disks come in several varieties, although only the writeonce-read-many (WORM) and magneto-optical (MO) or erasable optical varieties are typically
used
The
in digital
capacity
imaging
archives.
ofWORM
disks
has
reached
more than 10 Gbytes, with 2 1 Gbytes expected in the near future. The WORM medium has the advantage of being very stable after being written. It is not influenced by external magnetic fields and will retain the data for at least 10 years. WORM technology uses a high-energy ing
laser
disk.
If the
burned
into
beam
laser
By modulating
digital
data,
digital
data
beam
beam
the surface
dium.
laser
focused
an
is on,
a pit
laser
accurate
actually
the
rotatis
of the optical the
is stored
on
me-
beam
with
representation on
the
is focused
the of the
optical
disk.
on the
The
disk
sun-
face by a mirror-lens system mounted on a movable arm. By moving the arm, random access
to any
individual
track
on
the
disk
can
a digital
to the disk, mirror-lens
data
pattern
a lower-power system irradiates
has been laser with the the pattern
written same and
drawbacks
MO
disks
and erased. The MO earth material coated This
material
ofWORM
tech-
can
can
be
written
on,
read,
disk medium is a rareonto the disk substrate.
maintain
localized
magnetic
The polarity can be changed by apa magnetic field while the material is
polarity.
heated.
A focused
laser
beam
is used
to pro-
duce the local heat necessary to create the polarity change. With this disk material, the light polarization is different for each of the two material states. Therefore, once a digital pattern
has
heating
laser
been
written
with
the
by modulating
incoming
the
digital
data,
tion pattern. The MO tive
than
reading the
process
WORM
is much
technology,
less
sensi-
resulting
in
significantly slower rates of data transfer. Data transfer rates are on the order of 100 Kbytes/ sec, although recent advances indicate that this limit will significantly improve in the futune. . Optical Tape Recent developments in flexible optical make it a viable technology for long-term age of digital images. Thirty-five-millimeter
flexible
optical
as high
as
media
1 Thyte
per
with
storage
reel
have
media
stor-
capacities been
demon-
strated. Data transfer rates of up to 250 Kbytes/sec have been routinely achieved. To date, no jukebox has been designed to use this technology, but the evolution of this product bears close scrutiny. U
SUMMARY
are undergoing and performance.
rates
for digital radical For
of MO drives
have
image
changes example,
archives in both price the transfer
increased
so dramati-
that they may become the preferred cal medium for image archiving. The real lemma is that there is no clear-cut answer cally
342
U
RadioGraphics
U
Frost
Ct
al
a
lower-power laser can be used to read back the data by filtering the respective polariza-
The technologies
be achieved.
Once
technology.
plying
transfer
rates as high as 14.5 Mbytes/sec, greater than all but the fastest magnetic disks. In addition, four full digital audio channels are simultaneously recorded with the digital video. I
of the major
nology is the lack of standardization. There is no guarantee that the disks produced by one manufacturer’s equipment can be used by another. In fact, even the size of the media has not yet been agreed on. The MO disk is another rapidly evolving
Volume
12
optidito
Number
2
the question ofwhich technology should be chosen for digital image archiving. The choice must be based on careful evaluation of the particular needs of a specific institution. The long-term stability of both WORM optical disk and optical tape makes them attractive as permanent archive media. Unfortunately, the fact that the information cannot be
changed tient
will lead
records,
to fragmentation
leading
this
increase
in operating
If the
problem
to increased
major
with
only
access
effective
function by other
of the factors
archival such
is the
rates
time
sets, digital rates of 14 benefit. Howare
media as network
not but
just are
speed,
1992
for
fault-tolerant
REFERENCES Dwyer
SJ III,
Cox
Templeton
AW.
tion
gray
digital
SPIE 1990;
GG,
Cook
LT,
Experience scale
McMillanJH,
with
display
high
resolu-
systems.
Proc
Seshadri
SB,
1234:132-139.
Arenson
3.
del HL. Digital imaging workstation. Radiology 1990; 176:303-315. Lo SC, Huang HK. Compression of radiologi-
4. 5.
images
Jam
AK. IEEE
Chen by
6.
Chakraborty
with
Radiology
server
a
RL,
ces.
Proc
affected net-
redundancy
2.
ne-
work activity, working storage speed, and display speed. The most feasible approach is to incorporate any digital imaging archive into a cohesive network resource system (6). This approach will allow one to add newer technologies to an existing archive scheme rather
March
1.
cal
consideration
transfer
U
time
a moderate
to provide
it
operation.
costs.
quired to transfer large image video technology with transfer Mbytes/sec can be ofsignificant ever,
tant,
of the pa-
because all of the images from a patient may be on multiple disks. However, periodic regrouping of patient studies on new media can alleviate
than totally replace it. Under this approach, becomes possible to optimize the technology for the specific application and, more impor-
Image 1981;
J, Flynn detection
irreversible
512, 1986;
DP,
1,024,
and
2,048
matni-
161:519-525.
data compression: 69:366-406.
a review.
MJ, Gross B, Spizarny of image degradation data
compression
D.
Obcaused
processes.
Proc SPIE 1991; 1444:256-264. Ackerman L. Digital storage of images. puters
in Radiology,
Frostetal
Kun-
March
1991,
U
pp
Corn43-48.
RadioGraphics
U
343