This presentation in mainly focused of understanding of automation and its utility in cytopathology. It will be very usefull for postgraduate in pathology, cytopathologist and cytotechnicians.

1. AUTOMATION IN CYTOLOGY
PRESENTER: DR Manan Shah

2. Automation
• Defined as the technology by which a process or procedure
is performed without human assistance.
• Automation in a clinical laboratory is defined as a process by
which analytical instruments perform many tests with the
least involvement of an analyst.

3. Introduction and History
• Despite the success of manual screening, some faults do exist,
prompting the development of viable automated systems.
• Automation in cytology started and was focused for a long
time on PAP smears.
• Later the same principles were applied to other areas of
cytopathology like organ cytology , body fluids and so forth.

5. Historical Background in Pap smears
• The first attempts to automate the screening of cervical smears
dates back to the early 1950s.
• Mellors and coworkers, among them Papanicolaou, developed
a scanning device, the “Cytoanalyzer, which could gather
data on nuclear size and nuclear optical density of a large
number of cells.
• Lacking computerization, the process was slow.

6. The “Better” Pap Smear
• In May 1996, the ThinPrep ® Pap test was approved by the
FDA.
• Clinical trials confirmed increased sensitivity compared to the
conventional smears.
• Although adding cost, a number of studies suggested that a
reduction of ASCUS and unsatisfactory Pap tests have saved
lots of money spent on unnecessary recall visits and negative
colposcopic examinations and biopsies.

7. Need for automation
• The most challenging cases in cytology are those representing
failure to detect abnormalities existing at the time of screening
• Mainly difficulties in reporting occur due to overlapping of
cells and nuclei and obscuring factors.
• Automated instrumentation may improve sensitivity, reduce
unsatisfactory specimens and provide for reasonable bottom
lines.

8. Expectations from the Auto analyser
• The system should be able to comment on smear adequacy
• Scanning should be rapid and reliable and reproducible.
• The system should select suspicious cells (or slides) and
present them to the cytologist for final classification
• Sensitivity of the automated device (plus the cytologist) should
equal or exceed the sensitivity of the conventional method.

9. Different types of Automation tools
• Semi automated
• Fully automated
Principle
Artificial
Neural
Network
Robotics

10. Artificial Neural Network
• An artificial neural network (ANN) is a statistical classifier that
can be trained to recognize and distinguish patterns.
ANN
Input
units
Raw
information
that is fed
into the
network.
Hidden
units
It is determined by the
activities of the input
units and the weights on
the connections between
the input and the hidden
units.
Output
units
The behaviour of the output
units depends on the activity
of the hidden units and the
weights between the hidden
and output units.

11. Feed forward networks
1. Unidirectional flow of
information.
2. Good at extracting
patterns Generalisation
and prediction.
3. Parallel processing of
data.
4. Not exact models, but
good at demonstrating
principles
Recurrent networks
1. Multidirectional flow of
information.
2. Memory / sense of time
3. Complex temporal
dynamics.
4. Various training methods
5. Often better biological
models.

14. Goals of automation
1. Improving the accuracy of test results.
2. Shortening the length of time needed to perform the tests.
3. Obtaining a slide that is representative of the original sample
collected from the patient.

15. Automated devices are placed in 3
different categories
Specimen
collection
and
preparatio
n devices
Thin
Prep
Surepath
Manual
screening
adjunctive
devices
Pathfinder
The
papnet
system
AUTO
PAP 300
System
Autocryte
interactive
system
Automated
screening
devices
Focal
point
slide
profile
Thin
Prep
imaging
system
Focal
point
GS
system

16. Automation in Cervical Cytology

17. Specimen collection and preparation
device
The FDA has approved 2 automated systems:
1.Thin-Prep Processor
2.AutoCyte Prep / Surepath - now part of TriPath Imaging,
– Both systems use fluid-based collection devices for the
collection of the specimens.

18. Thin prep
• Utilizes the controlled membrane transfer technology
• Vial is spun gently to breakup the mucus, blood, debris and large
cell cluster, mixes the sample
• Series of negative pressure draws the fluid through the thin prep
membrane
• Epithelial cells and organism are trapped and blood, mucus and
debris pass thro it.

19. Thin prep
• Draws thin evenly layered diagnostic material
• Cellular material is transferred to glass slide using computer
controlled mechanical positioning and positive air pressure
• Slide with thin evenly layered circle of epithelial cell (20mm) is
made
• Slide is ejected into a cell fixative bath ready to staining and
evaluation.

23. Surepath technique
• Layering of cell sample on to a liquid density gradient-
vortexing and centrifugation
• Vortex –breaks up large cell aggregates, mucus and blood
• Density gradient centrifugation separation of cellular
elements from obscuring inflammation and debris
• Filtrate is placed in a chamber and applied glass slide by
gravity sedimentation
• Even layered circle of cells on slide (13mm)
• Automatically stained by surepath processer.

25. Advantages
• Decrease in number of inadequate smears and interpretation
time
• Randomised representative sample of cells-more accurate
diagnosis
• Back ground environment absent (!!!!!)
• Improves sensitivity and specificity
• Infective organisms, benign cellular changes, endocervical
atypia and carcinoma have similar features
• Increase relative sensitivity of ASC-US, ASC-H, and LSIL

28. Pitfalls of LBC
• Smear patterns altered because of randomization of cells.
• Abnormal cells are dispersed.
• Scanty LBC preparations can be difficult to screen and
interpret.
• Blood mucous inflammation and malignant diathesis are very
difficult to interpret

29. Pitfalls of LBC
• Epithelial cells appear mostly as single cells and are slightly
smaller than they appear in conventional smears especially
endocervical cells and immature metaplastic cells.
• LBC is more expensive than conventional test.

34. Ancillary Testing
• One of the most compelling reasons for using LBC over
conventional cytology is the ability to perform ancillary tests
on the remaining cells in the LBC medium.
• The first ancillary test taken from LBC to be evaluated and
proven to be useful in multiple studies was testing for HPV.
• The sensitivity for CIN II or III of HPV testing of residual LBC
from specimens interpreted as ASC-US reported in these
studies varied from 89%96 to 92%.
• Testing for chlamydia and gonorrhea also possible - from the
sample taken from thin layer preparations.

35. Automated Device for
Screening

36. 1. Manual screening adjunctive device
• It speeds up the manual screening process.
• Maps out specific fields on slides that the cytotechnologist
needs to review as opposed to the technologist screening the
entire slide.
• These are computerized microscopes which can electronically
and physically dot abnormal cells or even mechanically drive
the stages to the coordinates of previously identified
abnormal cells.

38. Pathfinder
• The Pathfinder is considered an adjunctive screening device
because slides are manually screened by cytotechnologists.
• It consists of monitor, a keyboard, and a storage device attached
to the microscope.
• During the screening process, the Pathfinder maps area of each
smear that has been screened by cytotechnologist, calculates the
average percentage of fields overlapped, records the time that
the cytotechnologist spent evaluating the smear.
• It is no longer manufactured or marketed.

39. Review scope

40. 2. The Papnet System
• This automated screening device is designed to detect rare
abnormal cells when present in a conventionally prepared
slide.
• It uses the principle of neural network processing,

42. 3. The AUTOPAP 300 System
• The AutoPap automates the
screening of conventionally
prepared cervical smears.
• The system uses the principle
of image analysis algorithms
and field of view (FOV)
computers to classify cell
images.
• In the primary-screening
mode, the instrument
screens the slides and ranks
them into 2 categories:
– Archived or no further review
required
– Review required

43. 4. Autocryte interactive system
• The AutoCyte is undergoing FDA clearance to be approved for
screening of monolayer cervical smears.
• It uses the same principle of algorithmic classifiers as does the
AutoPap, presents a computer evaluation derived from the
population histogram analysis, and allows the technologist to view
specific fields on the slide.

44. Automated screening system in gynec
cytology-Outline
• Focal point slide profiler
• Thin Prep imaging system
• Focal point GS system

45. 1. Focal point slide profiler
• Smears or SurePath
slides
• 8 slides/tray, 36 trays
• Capacity: 288 slides per
24 hours
• High speed video
microscope
• 3 cameras operate on
different focal planes :
dynamic focussing.
• Strobe light used to
acquire 25 images/sec
• 4x magnification: map of
entire slide and 1000
fields captured at 20x
magnification
• Image analysis performed
using preset algorithms
• Score assigned to each
slide (range: 0 to 1)

46. Focal point slide profiler-Sensitivity for
conventional smear
25,125 cases ASCUS LSIL HSIL
Current practice 79% 86% 93%
Focal point 86% 92% 97%
Significant Significant Not significant

47. Summary for focal point slide profiler
• At least as accurate as manual screening.
• False-negatives do occur.
• Modest productivity enhancement (15-20% saving
in screening time).

48. 2. Location guided imaging with the thin
prep Imaging system
• For thin prep slides only
• Image processor is computer based system run on window NT
• 25 slides/cartrige, 10 cartriges and Capacity 300 slides/day
• Measures integrated optical density of nuclei
• Identifies 22 fields on each slides that are most likely to harbour
abnormal cells
• If all 22 field are judged normal –Negative without further review
• If any field are judged abnormal-Full slide screening

49. Sensitivity
9550 CASES ASCUS LSIL HSIL
Manual screening 76% 80% 74%
Image assisted 82% 79% 80%
Significant Non significant Non significant

50. Specificity
9550 CASES ASCUS LSIL HSIL
Manual Screening 97.6% 99.0% 99.4%
Image assisted 97.8% 99.1% 99.2%
Non significant Non significant Non significant

51. Thin prep Imaging system-Summary
• At least as accurate as manual screening
• As with focal point ,false negative do occur
• More significant productive enhancement (25-50%) than the
focal point
• Many favourable post approval studies
• 70% of thin prep slides in US are evaluated using TIS

52. 3. Focal point GS imaging system
• Similar in design concept to thin prep imaging system
• FDA approval granted in 2008
• Slides imaged by FP slide profiler
• Field of vision examined for all adequately scanned
slides
• 10 FOV presented in order of decreasing score. All to
be examined

53. Summary of focal point GS imaging system
• Improved sensitivity
• Less false negatives
• More significant productivity enhancement

54. COMPUTER VISION TECHNIQUES

55. Computer vision Techniques
• Automated systems for cytology are static image analysis
systems which comprise a cell scanner (Digital camera) which
“see” images by measuring the light intensity and colour
properties being received by their electronic sensor elements.
• If stained cytology samples is placed in an apparatus which
has lenses and a digital light sensor (camera) one can “train”
the computer to react to chromatin clumping as well as some
of the other criteria we use, such as nuclear size, form etc.

56. Computer vision Techniques
• The optical images caught by the camera are converted into digital
images inside the camera and stored on a magnetic disc.
• The computer is programmed to analyse and classify the images.
• The computer selects images/ smears which are most likely to
contain abnormal cells and presents them to the cytotechnologist
for further triage under the microscope.

57. Computer version techniques
1. Pattern recognition
a) Segmentation
b) Image pre-processing
c) Feature extraction
d) Feature pre-processing
e) Feature selection and discrimination measures
f) Classification
g) Evaluation of classifier performance
2. Texture analysis

59. Segmentation
 Extraction of:
1. The background
2. The heaps- Separation of the isolated cells and
the heaps.
3. The position of the nuclei
4. The boundary of the nuclei

62. RECENT ADVANCES

63. AUTOMATION IN LUNG LESIONS

64. The lung cell evaluation device (LuCED)
• Early Detection of Lung Cancer in Sputum Based on 3D
Morphology.
• It produces 3D volumetric cell representations in isometric,
sub-micron resolution based on computed tomography.
• VisionGate, Inc. in collaboration with the University of
Washington, is developing LuCED test to score sputum
samples processed by the Cell-CT for evidence of cell
dysplasia or cancer.

65. • The LuCED test comprises a series of steps starting with cell
preparation including fixation and staining with hematoxylin.
• Based on cellular prevalence counts, its estimated that LuCED
sensitivity exceeds 90% as specificity approaches 100% for
patients with cancer cells in sputum.
• Cell analysis in 3D provides an unobstructed and
unambiguous representation of normal and cancer cell
morphology.

66. AUTOMATION IN URINE CYTOLOGY

67. Automated Urine Microscopy Analyzer
• Automated instruments have reduced the need for labour
intensive manual microscopy.
• There are 3 systems currently available to automate manual
microscopy.
1. An image-based analysis system that uses a video camera and
strobe lamp (stops fluid motion) to detect and sort particles based
on predetermined particle dimensions.
2. The other type is based on principle of flow cytometry, it classifies
particles based on fluorescent intensity, electrical impedance, and
forward angle light scatter
3. A next-generation automated image-based urinalysis system, the Iris
iQ200 Elite recently received US FDA clearance.

68. • Images are stored and can be viewed on the workstation
screen, thereby eliminating the need for manual microscopy
in most cases.
• Only urine samples containing crystals and/or yeast that
would require review images for confirmation.

72. Conclusion OF ARTICLE
The results from the automated analyzers for erythrocytes, leukocytes
and epithelial cells were similar to the result of microscopic
examination. However, in order to avoid any error or uncertainty, some
images (particularly: dysmorphic cells, bacteria, yeasts, casts and
crystals) have to be analyzed by manual microscopic examination by
trained staff. Therefore, the software programs which are used in
automatic urine sediment analysers need further development to
recognize urinary shaped elements more accurately. Automated
systems are important in terms of time saving and standardization.

73. Automation in Molecular Cytopathology
• Growing field
• Time-consuming
• Need for standardization
• Increase Efficiency
 Automation
 FISH
 Laser Micro dissection (mutation analysis)

74. • Diagnostic
Urine (multi-target)
Mesothelioma (9p21)
Biliary tract (9p21)
Translocations
Few others…
BRAF - Melanoma, pap. thyroid carcinoma, NSCLC
C-kit – GIST
• Predictive
Breast (HER-2)
Lung (EML4-ALK)
Lung (MET)
Lung (EGFR)

76. Laser Capture Microdissection (LCM)
• Laser Capture Microdissection (LCM) -technique for isolating
pure cell populations from a heterogeneous tissue section or
cytological preparation through direct visualization of the cells.
• Molecular profiling of diseased and disease-free tissue,
permitting correlation of cellular molecular signatures with
specific cell populations.
• DNA, RNA, or protein analysis may be performed with the
microdissected tissue by any method with adequate sensitivity.
• Automated LCM platforms combine graphical user interfaces
and annotation software for visualization of the tissue of
interest in addition to robotically controlled microdissection.

77. Laser Microdissection
Laser Pressure Catapulting(PALM®)

78. • The principal components of LCM technology are
– Visualization of the cells of interest through microscopy,
– Transfer of laser energy to a thermolabile polymer with formation of a
polymer—cell composite, and
– Removal of the cells of interest from the heterogeneous tissue section.
• Automated LCM is compatible with a variety of tissue types,
cellular staining methods, and tissue preservation protocols
allowing micro-dissection of fresh or archival specimens in a
high-throughput manner.

79. Short term implication
• Decision on which system of automation.
• Calculation of minimum workload required for cost effective
implementation of the machine.
• Reduction in workforce.
• Calculation of the minimum and maximum workload for the
screener.
• QC/QA procedure for the machine.
• Revision of the patient information leaflet and report format.

80. Long term implication
• Room for new technologies.
• Risk for litigation- human versus machine errors.
• Who takes medico-legal responsibility for machine error?
• Effect on workload of machine on the volume and pattern.

81. Conclusions
• Automation in cytology has taken a long time to be realized,
but it is now a reality.
• The technology is exciting and, if given time to develop in
tandem with standard good laboratory practice, i.e., parallel
studies in routine settings, then the most effective
components of these systems will prevail.
• Cooperation among pathologists, clinicians, and
manufacturers will ensure that the technology performs as
expected and contributes to affordable and reliable patient
care.

83. History in other organs

85. REFERENCES
• Koss’s Diagnostic cytology-Volume II
• Henry’s clinical diagnosis and management by laboratory methods.
• The Automation Trend in Cytology Laboratory Medicine. Volume
31(4) April 2000
• Cervical Cytology Automation: the U.S. Experience .17th
International Congress of Cytology, Edinburgh, Scotland -2010
• Textbook on adaptive multi-scale and texture analysis with
applications to automated cytology.
• Image segmentation applied to cytology.

86.  Thank you 