A New Era in Automated Cell Image Analysis

18 September 2014 – Pfäffikon
A New Era in Automated Cell Image Analysis
David Zahniser, PhD- Chief Science Officer
Roche Diagnostics Hematology, Inc.
Boston, Massachusetts USA
Page 1 - September 18, 2014
Disclosures and Notice
Dr. Zahniser has financial interest in
Roche Diagnostics Hematology, Inc.,
manufacturer of the cobas m 511
The cobas m 511 is not cleared by
The US Food and Drug Administration,
and is not commercially available.
Page 2 - September 18, 2014
Microscopy – The Foundation of Clinical Diagnosis
George N. Papanicolaou
1883-1962
Paul Ehrlich
1854-1915
Page 3 - September 18, 2014
Microscopy – The Foundation of Clinical Diagnosis
Question… How do we provide the BEST and the most DEFINITIVE
Morphologic Diagnoses?
Answer… Optimization of:
Sample Collection
Sample Preparation
Staining
Microscopic Interpretation
Quantitation
Page 4 - September 18, 2014
Examples of Automated Preparation, Staining, and Imaging
Cytology
Histology
Page 5 - September 18, 2014
Modern Automated Hematology Analyzers
Counts: RBC, WBC, PLT
Measures: Hgb, Hct, RBC metrics
Classifies: WBC Differential
Page 6 - September 18, 2014
“Flags” reflect Uncertainty
“Flags” are messages that the measurements cannot determine
the cell type or abnormality (10%-40% of cases)
About 50% of flags trigger a manual review of the microscope slide
Page 7 - September 18, 2014
Manual Wedge-smears of Whole Blood
• The “Gold Standard” for cell morphology and WBC differential
But…
• Quality of results dependent on preparation and stain
• No consistency in selection of review areas, so results vary
Page 8 - September 18, 2014
23543566
Doe, John
What is the CBC on this slide?
Easy answer, you use your eyes to count:
7 RBC
6 WBC
2 PLT
Page 9 - September 18, 2014
23543566
Doe, John
What if … an instrument could:
•
•
•
•
•
•
•
•
Paint a precise and reproducible amount of blood onto a slide
Have all cells countable in a thin layer
Have the cells be morphologically the same as under the scope
Have fresh, reproducible staining
Obtain all CBC parameters from the slide using imaging
Perorm an automated 5 part differential using imaging
Produce images for evaluation when needed
Do this at a rate of 60 samples per hour…
Page 10 - September 18, 2014
The cobas m 511 System: Four-in-One
Slide Maker
Stainer
Analyzer
Imaging System
Page 11 - September 18, 2014
Whole blood is printed onto a glass slide…
Printing process overview
 The system prepares each slide by printing whole blood from a stainless needle in
adjacent rows
Page 12 - September 18, 2014
Uniform Printing of a Precise Blood Volume
Page 13 - September 18, 2014
…and then the slide is stained…
Staining process overview
 Each slide stained individually
 Uses a specially designed Romanowsky-type stain
 Stain comprised of 4 separate solutions including a fixative, 2 dyes, and a rinse
Page 14 - September 18, 2014
…before the system captures cell images…
Imaging process overview
 Uses multispectral imaging at low and high powered magnification
 Computer-driven classifiers locate, group and count different types of cells
 Computer quantifies and classifies using morphologic descriptors
 Size, shape, color, and optical density are assessed and used with statistical classifiers
Four wavelengths of light from LEDs illuminate the cell fields for the B&W camera
Page 15 - September 18, 2014
The Bloodhound System: Four-in-One
Page 16 - September 18, 2014
Interior View
Page 17 - September 18, 2014
Imaging enables key measurements
Bloodhound measurements
 Measures Hemoglobin using specific wavelengths of light from an LED
 Wavelengths based on where hemoglobin absorbs optimally on air dried slide
Hemoglobin
Hemoglobin
Component
 Measures Mean Cell Volume (MCV) by finding the height (by optical density at specific
wavelengths) at dozens of points within each red blood cell
Mean cell
volume
 Volume of the cell calculated from the total of these heights
 Tens of thousands of RBCs used to calculate the mean
.. ...
.
. . . ..
.
...
..
.
..
.
..
.
...
...
..
..
..
...
.
..
..
.
...
.
..
..
.
..
.
..
..
...
.
..
..
...
.
..
..
...
.
.. .
. .
... . . .
.
 Volume occupied by RBCs can be calculated after determining the MCV and the total
number of RBCs in one microliter
Hematocrit
 The percentage then gives the Hematocrit
Hematocrit =
MCV x Total RBCs
Total volume
Page 20 - September 18, 2014
Unique viewing station…
Viewing station highlights
 Comprehensive review screen
displays:
‒ CBC and 5-Part WBC Differential
data
‒ Individual images of red blood cells,
white blood cells, platelets,
reticulocytes and unclassified cells
 For flagged cases, medical
technologist reviews data and images
to determine final result
 iMac and analyzer use custom CMI
software not accessible by user
Page 21 - September 18, 2014
…and with optimized interaction
Page 22 - September 18, 2014
…that makes review of flagged cases efficient
Review of sample flagged cases
Leukemia
Neutrophilia
Leukopenia
Anemia
Page 23 - September 18, 2014
Calibration, QC and proficiency testing
Overview of calibration, quality control and proficiency testing
 Based on stabilized RBC, PLT, WBC, nRBC, and
Reticulocytes
 Development underway with commercial supplier
 Calibrators and controls used in compliance with
applicable laboratory regulations including CLIA
and CAP
 Parallel 510k clearance process
Page 25 - September 18, 2014
A 26 Parameter CBC and Diff by Microscopy
Digital Imaging:
• Counts RBC, WBC, PLT, nRBC,
Retics
• HgB, MCV, MPV
• WBC Differential
Human Review:
• Classifies unclassified cells
• Assesses morphology
• Finalizes differential
Potential Efficiencies:
• Improved TAT
• Better staff utilization
Page 26 - September 18, 2014
Testing Data
Comparability
Purpose and timing
Clinical data
 Determine correlations between common
parameters when identical samples were
processed on a Sysmex XE-5000 and the
Bloodhound Instrument
 Data based on verification studies performed in
February 2013
Parameter
r
Slope
Bias
RBC
0.99
0.98
0.04
WBC
0.99
1.02
0.25
PLT
0.97
0.96
-7.08
MCV
0.91
1.14
-0.24
MCH
0.98
1.00
-0.08
HGB
0.99
1.05
-0.12
Method, standards and statistics
Conclusions
 Routine samples (normal and abnormals)
 Strong correlation observed for large sample
processed on a Sysmex XE-5000
set between the Bloodhound Instrument and
the Sysmex XE-5000 for all measured
parameters
 1,183 samples processed on an both
instruments for CBC and WBC diff parameters
 Standard: H26-A2
 Statistics: Deming regression and Bland-Altman
 Minimal bias between instruments
plots for bias assessment
Page 27 - September 18, 2014
Comparability – RBC
Page 28 - September 18, 2014
Comparability – WBC
Page 29 - September 18, 2014
Comparability – Platelets
Page 30 - September 18, 2014
Imprecision: Repeatability
Purpose and timing
Clinical data (%CV)
 Ability to obtain consistent results when
Parameter
Repeatability
(95% CI)
WBC
2.99
(2.87-3.12)
RBC
1.18
(1.14-1.23)
PLT
3.11
(2.99-3.24)
HGB
1.59
(1.53-1.66)
HCT
1.66
(1.60-1.74)
MCV
1.19
(1.14-1.24)
MCH
1.06
(1.02-1.10)
Method, standards and statistics
#Neut
4.11
3.95-4.28
 Normal samples spanning the reference ranges
#Lymph
10.44
(10.03-10.89)
#Mono
15.67
(15.06-16.35)
#EO
27.18
(26.11-28.34)
running a sample multiple times on the same
instrument (repeatability)
 Data based on verification studies performed in
February 2013
were run the 31 times on 3 Bloodhound
instruments
 Abnormal samples assessed medical decision
levels for RBC, HCT, HGB PLT and WBC
 Standard: H26-A2
 Statistics: Percent CV calculation
Conclusions
 Excellent repeatability (low CVs)
 Equal to or better than the published %CV of
automated hematology analyzers
Page 31 - September 18, 2014
What else can you learn from the morphologic analysis of
each cell on the slide, and…
….What clinical value?
Page 32 - September 18, 2014
How can remote review affect quality and staffing?
Potential for an integrated delivery network
 Allows flexibility in lab layout and technician time
 Ability for pathologist or supervisor to view remotely
Pathologist
Supervisor
Technologists
Page 33 - September 18, 2014
Thank you for your attention.
Roche Diagnostics Ltd
[email protected]
6343 Rotkreuz
Switzerland
COBAS, COBAS M, BLOODHOUND and LIFE NEEDS ANSWERS are
trademarks of Roche.
All other trademarks used or mentioned are the property of their respective
owners.
This presentation is our intellectual property. Without our
written consent, it shall neither be copied in any manner, nor
used for manufacturing, nor communicated to third parties.
Page 34 - September 18, 2014