Clinical Validation of High Throughput AQUA and Multispectral Imaging for Breast Cancer
Cristen K. Hays
1
, Kerry Fitzgerald
1
, Bana Mouwakeh
1
, Peter Miller
2
, Clifford Hoyt
2
, Jason Christiansen
1
, Danielle Murphy
1
1
Genoptix Medical Laboratory, Carlsbad, CA;
2
Perkin Elmer, Hopkinton, MA
2679
INTRODUCTION
The high rate of discordance currently identified in the
scoring of breast cancer biomarkers ER, PR, Her2, and
Ki67, which directly impact patient management, drives a
growingclinicalneedforthedevelopmentofdiagnostictests
which offer enhanced reproducibility while maintaining
accuracy in a high throughput environment. We currently
offer AQUA
®
technology in our clinical lab for breast
cancer. A significant limitation to using this technology
within a diagnostic laboratory is low throughput and
manual annotation of patient invasive tumor. To address
this, we evaluated the Vectra2
™
automated multispectral
slide analysis system in a clinical environment with a
focus on quality and pathologist involvement.
MaTeRIals aND MeTHODs
All results were generated using de-identified clinical
samples. The Vectra system was used to acquire
breast biomarker expression in invasive tumor regions
scored by AQUA technology. The multi-step process
of image acquisition was assessed in conjunction with
the reproducibility of region selection for scoring as
determined by invasive cancer tissue classifier algorithms
developed through the use of Inform and Nuance
applications. We also developed a review application
to allow remote access to images prior to reporting of
patient results.
ResUlTs
Normalization of system intensity compensation produced
intensity values approaching unity. Algorithms defining
invasive tissue score were found to be highly correlative
(ER, PR, Her2 AUC 0.966-0.967; Ki67 R2) to the current
clinical platform with pathologist annotation. Together,
these features resulted in acquisition of conserved regions
of invasive tumor in a highly reproducible fashion, such
that tissue scores were the same for the same sample on
multiple days and with different instruments, p=0.7335.
Scores were consistent amongst varying parameters,
including repeated scanning (~9 %CV, n=13 runs) and
differing field sampling parameters while decreasing time
to 15 min per slide for processing with parallel analysis.
Our development of the Vectra Review software enables
visualization of regions selected for biomarker scoring
that can be reviewed by clinical staff remotely, while not
impacting overall workflow.
CONClUsIONs
We show that the Vectra 2 platform is as accurate as our currently clinically validated platform for
AQUA scoring of ER, PR, Her2, and Ki67. The introduction of advanced tissue-finding algorithms
concisely recognize relevant tissue regions unique to each biomarker without manual annotation
and greatly improved the clinical workflow. This deployment of digitalized pathology resulted
in a reduction of time to generate scores, which marks a significant advance for our clinical lab.
Overall, this clinical validation demonstrated that the AQUA-paired Vectra 2 system is a highly
robust, reproducible, and high throughput advance to our currently available diagnostic breast
cancer platform that still allows for clinical lab and pathologist review.
•AQUAtechnologyandthestabilityoftheVectra2platformenableshighreproducibilityof
AQUA scores between Vectra machines and during repeated scanning.
•User-deinedt issuealgorithmsoftheVectra2systemenablerobustHPFcollectionand
AQUA scoring with clinical relevance.
•Vectrascoresarehighlycorrelatedwiththoseofourcurrentclinicalplatform.
•Vectra2withAQUAtechnologyprovidesareproducibleandrobustassaywhichimproves
workflow in a clinical setting.
Genoptix
®
is a registered trademark of Genoptix, Inc. © 2013 Genoptix, Inc.
Background of Vectra 2 and Vectra Review
Vectra 2 Platform Improves Efic iency for Automated Quantitative Analysis
Figure 1. The Vectra 2 platform is a multispectral imaging
system for acquisition and analysis of brightfield/
fluorescent microscopy with tissue microarray (TMA)
and whole tissue sections (WTS). It relies on associated
software, Inform and Nuance, to generate algorithms
used to identify regions of interest (ROI) at 4X for further
sampling at 20X while reducing background fluorescence
for pure imagery. Images are easily reviewed with Vectra
Review software. Vectra Review displays a stitched image
of the entire tissue section and allows the reviewer to
toggle between 4X and 20X images in order to accept or
reject images based on quality prior to scoring.
Figure 2. Introduction of the Vectra
2 platform into the clinical lab
offers a significant improvement to
workflow for performing AQUA. A.
Automated QUantitative Analysis
(AQUA) is used to quantitatively
measure biomarker expression
by assessing marker presence
(target) within a defined tumor
mask. B. The current workflow
involves review of a hematoxylin
and eosin (H&E) section to
determine invasive tissue ROI for
fluorescent immunohistochemistry
(FIHC)section.AFIHCsection
is scanned and annotated before
AQUA scoring and includes
several manual steps. Use of
Vectra eliminates the need for
H&E staining before scoring, while
utilizing tissue-finding algorithms
to identify ROI and reduce the
number of high powered images
captured. The time to acquire and
analyze a specimen is ~15 to 20
minutes. A pathologist reviews
images with the associated H&E
section prior to final scoring by
AQUA through Vectra Review.
Figure 3. AQUA technology combined with the Vectra
platform provides a stable and reproducible system for
AQUA analysis. A. Scoring is dependent upon changes
inlightoutput,useofaFluorRefslideallowsfor
consistency in score reporting independent of changes
in light output. B. Repeated scanning does not result
in significant changes in AQUA scoring in (left) TMA,
p=0.4972 or (right) WTS ~9%CV, n=7. C. Imaging
of WTS results in highly reproducible AQUA scores
between four different Vectra 2 machines p=0.7335.
Development of Robust Software Parameters
Figure 4. A. User-defined ROI finding
algorithms reliably identify the same regions
during repeated scanning on the same and
different machines. B. Score generation with
varying numbers of high powered field images
of whole tissue section samples generates
comparable scores by AQUA.
Concordance of AQUA Scores Generated by New Platform
ER
Her2
PR
Ki67
Figure 5. A. We
used ROC analysis
to investigate the
concordance of Vectra
2 generated AQUA
scores for ER, PR,
and Her2 based off of
validatedcutpoints.For
Ki67 we used linear
regression analysis
and observed a high
correlation of the
Vectra platform to our
current system.
A.
B.
A.
B.
C.
Stability and Reproducibility of Instrumentation
A.
B.
Hays_AACR_poster.indd 1 4/3/13 4:52 PM