Special Techniques in Cytology
the task. However, th is project allow ed investigators to appreci-
ate and develop approaches to the problem s encountered in the
early investigative process. Parts o f the cytoanalyzer were used
b y M endelsohn's group to b u ild CYD AC (C yto p h o to m e tric
D ata C onversion), a system designed tow ard com puter-assisted
description and id e n tific a tio n o f in tra c e llu la r cell patterns.7,11
In the 1960s, a n um b er o f cytochem ical scanning m ethod s and
instrum ents o f the integ rating type were described.1,12-14 The
m ost sophisticated optical and m echanical design entered in to
the U niversal Recording M icrosp ectrop hotom eter (U M S P -1) o f
Carl Zeiss,15 w h ic h incorporated an ingenious hig h-precision
m echanical stage and achrom at lenses corrected fo r the u ltra -
v io le t and visib le range. L im ita tio n s precluded clinical applica-
b ility , however, inc lu d in g the slo w scanning m echanism .
In 1968, W ie d and collaborators introd uced the ta xon om ic
in tra c e llu la r analytic system (TI-C A S),16 a device fo r cell id e n tifi-
cation based o n com p uter evaluation o f the intra ce llula r patterns
o f image scan data. T his e ffo rt stim u la te d the interest o f others
in Europe and Japan, leading to later systems such as the VC SA ,17
Cerviscan,18 CYBEST ,19 T U D A B ,20 SAM BA,21 Diascanner,22 FAZY-
T A N ,23 LEYTAS,24 M agiscan,25 M c G ill System ,26 and TU LIP S .27,28
Studies o n the com p ositio n o f cellular samples showed
th a t the m a jo rity o f ab norm al cells in conventional Papanico-
lao u smears were in clusters.29-33 This understanding led to the
conclusion th a t better cell presentation was needed to o p tim ize
the perform ance o f scanning devices. T h ro u g h o u t the 1970s, the
emphasis m oved to developm ent o f techniques o f cell sam pling,
preservation, disaggregation, and m onolayering o f cell prepara-
tio n s .34-37 Such specimens allow ed fo r better visu a liza tio n o f cells
fo r in d ivid u a l cell analysis, a process better suited to the com pu-
ter hardware and software available. These processes have c u lm i-
nated in the present w ith successful intro d uc tion s o f autom ated
cytology products in to the m arketplace. The 1980s saw the fo r-
m a tio n o f com panies such as N eopath, Cytyc, N eurom edical Sys-
tems, and Roche Image Analysis Systems, w h ic h b rou g h t forw ard
and tested the products th a t we n o w ro u tin e ly u tiliz e in d aily
practice. (Several organizations have sponsored and sustained
those w o rkin g in the fie ld o f a u to m a tio n fo r years. These include
the Engineering Fou nd a tion o f Am erica, the Concerted A ction
o n A utom ated and A nalytical C yto m etry o f the European Eco-
n o m ic C o m m u n ity, and Professor George L. W ied o f the U n ive r-
sity o f Chicago, w h o organized academic m eetings and specific
conferences o n a rtific ia l intelligence and cytology a u to m a tio n .)
The Rationale for Automation
Cervical cytology has always been a h ig h -vo lu m e test, one o f the
last such tests to be p erform ed in a com p letely m anual fashion.
Because m anual in te rp re ta tio n o f cells fo r the id e n tific a tio n o f
ab no rm a lities has always fo llo w e d a m orp h olog ic criteria-based
approach, it was o n ly reasonable to hypothesize th a t such a
process w o u ld be am enable to an "im age analysis"-based com -
p u te riza tio n approach in w h ic h the com p uter softw are w o u ld be
able to re lia b ly "q u a n tita te " w id e ly recognized m orp h o m e tric
param eters such as nuclear size, nucleus to cytoplasm ic ratios,
c h ro m a tin texture, and d istrib u tio n , and th en sum (o r otherw ise
analyze) such id en tifie d cellular changes in to an inte rp reta tion ,
free o f any h u m a n inte rven tio n . A t first, such a hypothesis m ig h t
seem to be very reasonable; however, th is task was fo u n d to be
fa r m ore com plex due to a variety o f factors. The conventional
cervical cytology slide is very com plicated. It m ay contain up to
500 000 cells o f various types th a t m ay be displayed in an in n u -
m erable array o f patterns and variations, w ith com plex over-
lapping, obscuration b y b lo o d and in fla m m a tio n , and a w id e
variety o f presentations in w h ic h benign o r reactive cells m ay
m im ic m alig nancy and m a lig n a n t cells can m im ic the benign.
As such, early attem pts at a uto m a tion , geared tow ard m achine
fin a l inte rp reta tion , were generally unsuccessful. P h iloso p hic
changes in approach were necessary i f a u to m a tio n was to be
successfully introd uced as a m ainstream process. Therefore the
concept o f th e device as an aid to h u m a n in te rp re ta tio n came to
be regarded as a reasonable approach in d istin c tio n to m ethods
seeking fu ll device-only inte rp reta tion . T his p h iloso p hic overlay
is in h e re n t in the tw o current approaches being u tiliz e d in a uto-
m ated screening: (1) the triage o f slides b y the device to those
req u iring m an ua l review and those n o t req u iring m anual review
(the Becton D ic kin so n /T riP a th FocalP oint P rim a ry Screening
System, FPPS), and (2) the location-g uid ed screening approach
in w h ic h the device id entifies cells o r areas o n the slide w ith
h ig h p ro b a b ility o f a b n o rm a lity fo r presentation to hum ans fo r
fin a l review and in te rp re ta tio n (Cytyc ThinP re p Im aging Sys-
tem , N eurom ed ical Systems PapNet, Becton D ic kin so n /T riP a th
FocalP oint GS).
O ver th is tim e, parallel efforts tow ard im p ro vin g the cervi-
cal specim en were taking place. C lin ica l trials o f liquid-based
technolog y u tiliz e d new sam p ling devices such as endocervical
brushes and "b ro o m s " th a t provided a m ore cellular sample,
p a rtic ularly o f the im p o rta n t tra n s fo rm a tio n zone. L iq uid -
based specimens are generally better d istrib uted on the slide,
w ith im p roved cell to cell separation o r "se g m e ntation ." Such
specimens are also less prone to cellular obscuration b y b lo o d
and in fla m m a tio n . Hence liquid-based specimens create a m ore
o p tim ize d "p la tfo rm " fo r autom ated screening, and the m ost
m od ern clinical devices take advantage o f th a t fact.
A n o th e r im p o rta n t piece o f the a u to m a tio n puzzle is the role
o f the reg ulatory agencies, m ost n o ta b ly the US Food and D rug
A d m in is tra tio n (U SFD A ), w h ic h has to pass on the safety and
effectiveness o f such devices p rio r to com m ercial release. It was
clear th a t the in itia l u tility o f autom ated screening devices des-
tine d fo r com m ercial release w o u ld be tested and first released
in a q u a lity c o n tro l (Q C ) use mode. A utom ate d screeners w o u ld
be u tiliz e d to "check" p reviously m a n u a lly screened slides lo o k -
ing fo r false-negative cases. D u rin g such activities, the overall
perform ance o f th e devices could be tested. In a d d itio n, the fo r-
m u la tio n o f the types o f trials necessary to adequately investigate
and d ocum ent these perform ance o r operating characteristics
w ith o u t c om p rom ising p atient care could also be developed.
Professional societies p ublished docum ents proposing per-
form ance measures,38,39 and the U SFD A and its advisory panels
judged the data presented b y m anufacturers.
F o llo w in g acceptable Q C perform ance, the in tro d u c tio n and
vetting o f p rim a ry screening devices could be initia te d .
Cytology Automation: Accuracy
and Productivity
There are tw o u ltim a te goals o f a uto m a tion : (1) to im p rove the
accuracy o f the in te rp re ta tio n in term s o f b o th false-negative
and false-positive perform ance, and (2) to im p rove p rod u c tivity
o f the w orkforce. A false-negative specim en is defined as one in
w h ic h an in te rp re ta tio n is rendered o f n o disease being present
w h en indeed such is n o t the case. False-negative specimens are
previous page 1006 ComprehensiveCytopathology 1104p 2008 read online next page 1008 ComprehensiveCytopathology 1104p 2008 read online Home Toggle text on/off