34
Automation in Cervical Cytology
T a b le 3 4 .3
FocalPoint/SurePath Study: 9665 Cases Studied
N o fu rth e r
F irst
S e co nd
T h ird
F o u rth
F ifth
re v ie w
HSIL+
85
22
2
0
0
0
LSIL
47
21
11
10
5
6
ASC7AGC
54
23
12
5
4
2
AGC, atypical glandular cells of undetermined significance; ASC, atypical squamous cells of undetermined significance; HSIL, high-grade squamous intraepithelial lesion;
LSIL, low-grade squamous intraepithelial lesion.
“Total ASC-positive over third rank: 84%.
(After Vassilakos et al. 2002,90 with permission.)
T a b le 3 4 .4
FocalPoint Ranking of SurePath Slides
Q u in tile
H S IL+ (% )
HSIL (% )
A IS (% )
C a rc in o m a (% )
1
60 (58)
19 (66)
0 (0)
41 (59)
2
25 (24)
7 (24)
1
(20)
17 (25)
3
11 (11)
3 (10)
2 (40)
6 (9)
4
6 (6)
0 (0)
2 (40)
4 (6)
5
1
(1 )—210
0 (0)-30
0 (0)
1
(1)-180
No further review
0 (0)
0 (0)
0 (0)
0 (0)
Total
103-21
29-32
5
69-87
Quintiles 1
and 2
85 (83)
26 (90)
1
(20)
58 (84)
Quintiles 1-3
96 (93)
29 (100)
3 (60)
64 (93)
AIS, adenocarcinoma in situ; HSIL, high-grade squamous intraepithelial lesion.
“Low cellularity cases automatically placed into Quintile 5.
(After Parker et al. 2004,89 with permission.)
T he FPPS has several o th e r op erational features w o rth n o t-
ing. Slides th a t cannot be successfully scanned are designated as
"Process review ." T his p o p u la tio n o f cases is generally less than
3% w ith SurePath slides (D. W ilb u r, unp u b lishe d data), and users
are required to send such cases on fo r a fu ll m an ua l review. O ne
o th e r im p o rta n t feature is th a t the FPPS makes an assessment o f
squam ous c e llu la rity o n all cases and is program m ed to never
place p o o rly cellular samples in to the N FR p o p ulatio n. This is
im p o rta n t because invasive carcinom a samples, w h ic h m ay be
p o o rly cellular w ith obscuring b lo o d and diathesis m aterial,
m ay receive a lo w device score. As a safety measure, the FPPS
always places such cases in to the fifth q u in tile (the low est cat-
egory req u iring a m anual review ). Therefore these cases w ill
always receive a m anual screening despite a score th a t m ay have
otherw ise caused them to be placed b e lo w the p rim a ry thresh-
old. D ata fro m a seeded stud y w ith large num bers o f invasive
cancers have show n the value o f th is system, as lo w -c e llu la rity
cancers have been re lia b ly placed in to th is review category.89
It therefore has been noted th a t the highest "ris k " FocalP oint
q u in tile s are the first (in w h ic h m ost HSILs are present) and
the fifth (in w h ic h p o o rly cellular invasive carcinom as can be
fo u n d ) (Table 3 4.4) .
Table 34.4 shows FocalP oint ran kin g data fo r SurePath slides
and the im p ortance o f q u in tile 5 "sc a n tly" cellular specimens.
T he FocalP oint system has a b u ilt in "safety n e t" a lg o rith m
designed to ensure th a t p o o rly cellular samples w ill always be
review ed b y h u m a n screeners. This was designed as such because
it is w e ll k n o w n th a t p o o rly cellular samples m ay p refere ntially
h a rb o r m alig nancy because o f d ilu tio n a l effects caused b y excess
b lo o d and necrotic m aterial. As such, lo w -c e llu la rity cases are
always placed in to q u in tile 5 to ensure a m anual review. N o te
th a t in th is study, w h ic h included 87 cancers, 18 were "lo w cel-
lu la rity " specimens placed in to th is group b y the m eth od noted
above. T his p rincip le m akes q u in tile s 1, 2, and 5 the "hig hest
p ro b a b ility "
q u in tile s — 1
and
2 fo r high-grade squam ous
in tra e p ith e lia l lesion-p ositive and 5 fo r cancer.
Location-guided Screening Application
As indicated above, the FocalP oint system has the a b ility to n o t
o n ly ra n k order slides b u t also to id e n tify the locations o n the
slide w here ab norm al cells are m ost lik e ly to be fo un d . The slide
score is a cum ulative process in w h ic h the scores o f in d ivid u a l
FOVs are sum m ed to achieve the to ta l slide score. T he highest
scoring FOVs w o u ld therefore be m ost lik e ly to contain ab nor-
m al cells. The a b ility to present these FOVs to m an ua l screen-
ers w o u ld be a fu rth e r im p ro ve m e n t in the overall process by:
(1) increasing p ro d u c tivity and (2) increasing accuracy b y apply-
ing the p reviously noted p rincip le o f "increased prevalence
leads to increased detection sen sitivity." W ith the device h o n -
ing d ow n the p o p u la tio n o f p o ten tial "needles in the haystack,"
the m an ua l screener w ill have a hig h er lik e lih o o d o f successfully
id e n tifyin g them . The process allow s fo r FO V review to triage to
a fu ll m anual review i f a b n o rm a lity o r p o ten tial a b n o rm a lity is
id entified , o r to triage direct to sign o u t as N IL M in the instance
in w h ic h no p o te n tia l a b n o rm a lity is id e n tifie d on F O V -o n ly
review. A t present, the device (in its SurePath-m ated configura-
tio n ) id entifies 10 FOVs per case, w h ic h are ra n k ordered fro m
highest to low est p ro b a b ility o f c ontaining ab norm ality.
1033
previous page 1017 ComprehensiveCytopathology 1104p 2008 read online next page 1019 ComprehensiveCytopathology 1104p 2008 read online Home Toggle text on/off