Will WholeSlide Imaging Change the Practice of Pathology Point CounterPoint

1
Will Whole-Slide Imaging Change the Practice of
Pathology?Point Counter-Point
John H. Sinard, MD, PhD Director, Pathology
Informatics Yale University School of
Medicine Department of Pathology and
Ophthalmology New Haven, Connecticut john.sinard_at_y
ale.edu
Ulysses J. Balis, MD Director, Pathology
Informatics Harvard Medical School
Massachusetts General Hospital Department of
Pathology Boston, Massachusetts balis_at_helix.mgh.ha
rvard.edu
APIII, 2005 August 24, Lake Tahoe, CA
2
Educational Objectives

• Salient Review of Underlying Wide-Field
  Technology
• Review of Wide-Field Capture Strategies
• Current and Anticipated State-of-the-art
• Review of Likely Application Settings
• Pros
• Cons
• Open Debate
• Audience questions and Open Discussion

3
The CCD the fundamental enabling tool of
digital image capture
4

• Pixel - A picture element a single location of
  a digital image made of x by y locations, or
  pixels.

Overall Digital Image
5
Image
Analog to digital converter

Each image location contributes a series of
analog voltages specific for its location only.
10011011011 00100110111 00010111010 11011101001 00
101110111......
Array of photosensitive elements
A binary number is finally obtained for each
unique location within the rectangular array.
6
Wide-Field Image Capture (type 1)formerly known
as the store and forward model

• Acquire the whole slide into the digital realm
• Image is scanned and reconstructed in some
  predefined time interval (minutes to hours)
• The data set is then available for display or
  dissemination.
• Better performance is achieved with increasing
  computational power and system memory

7
Wide-Field Image Capture (type 2)(RANA
Robotically Actuated Network Appliance model)

• Utilize remotely operated stage/focus/lighting/obj
  ective robotics to control microscope.
• Scan field of interest in a just-in-time digital
  delivery model, using the internet as the
  delivery conduit.
• Better performance is achieved with better
  communication bandwidth.

8
Wide-Field Image Capture Strategies

• Multiple technical approaches
• Area scan CCD camera utilizing conventional X-Y
  stage robotics for tiled reconstruction.
• Push Broom X-Y capture where the camera moves
  with the slide in one axis
• Line scan CCD with image strip reconstruction
• MEMs-based (microfabricated electromechanical
  systems) lens array technology

9
Representative tiling artifacts
10
Current State of Wide-Field Slide Scanning

• Single slide (and small set) scanning reduced to
  practice.
• Generally confined to a single acquisition plane
• Storage technology currently based on multiplanar
  TIFF / JPEG 2000 storage/compression technology.
• Optical path engineering is approaching the
  quality of modern brightfield microscopes

11
Wide-Field Microscopy Competing factors.

• Compression Ratio
• Too low digital storage is prohibitively
  expensive
• Too high Image is useless, diagnostically
• Resolution (image quality)
• Too low Image is useless, diagnostically
• Too high Image acquisition too timely to allow
  for conversion to an all digital signout paradigm.

12
Conventional Loss-basedImage Compression
Raw Data
Restored Data
Compression Algorithm
Restoration Algorithm
Compressed data (may or may not preserve spatial
organization of original data)
Depending on the selected compression ratio,
restored loss-compression imagery may or may not
be of diagnostic quality.
13
Vector Quantization
Original Image
Division of image into local domains
Extraction of Local Domain Composite Vectors
VKSLx0y0Order , LxnymOrder
Vectorization of each local kernal
Individual assessment of each composite vector
14
Vector Quantization
VKSLx0y0Order , LxnymOrder
Established Vocabulary
Query Against library (Vocabulary) of established
vectors
Novel Vector
Previously Identified Vector
Assignment of a unique serial number and
inclusion into global vocabulary
Assembly of compressed dataset
38857448643
15
VQ - BasedImage Compression
Raw Data
Restored Data
Compressed data (preserved spatial organization
of original data)
Depending on the selected compression ratio,
restored loss-compression imagery may or may not
be of diagnostic quality.
16
Resolution

• CCD Facts
• Number of pixels (i.e. MegaPixel Count) is
  directly proportional to the maximum number of
  transistors that current microphotolithographic
  techniques can allow on a single substrate
• Transistor count for both CCDs and CMOS imagers
  closely follows Moores Law, which states that
  total number of possible transistors on a chip
  doubles every 18 months. This has been generally
  accurate since the mid-1960s
• Current State of the art (mid-2005) in
  single-device imagers
• Consumer grade imaging 16.2 Megapixel (Canon)
• Scientific-grade imaging 22.6 Megapixel (Dalsa
  Corporation)
• This capability will likely double by the close
  of 2006

17
Some Observations concerning our revered subject
matter (the slide)

• Charactistics
• 2.5 by 7.5 cm
• 1/3 used for label
• 2.5 x 5.0 cm for tissue display
• Typical light microscopy is diffraction-limited
  to 0.25 microns
• Yields an effective required pixel count of 100K
  by 200k pixels (2.3 Gb) or a 20k MPixel Image
• This is the same things as saying that one would
  need to capture 20,000 images with a 1 MPixel
  camera to obtain a single slide
• Herein lies the essence of why telepathology has
  been so long in approaching an operational
  reality.

2.5 cm
7.5 cm
5 cm
(1000 x 25) / 0.25 microns 100,000 linear pixels
(1000 x 50) / 0.25 microns 200,000 linear pixels
This is a 20 GPixel image
18
What happens as Moores Law is applied to the
pathology problem?
19
Observations

• Current scanning throughput capabilities (based
  on 2000-2001 technologies when scanners were
  designed) take about 42 minutes to scan a single
  slide at diffraction-limited resolution
• At some point in the future, imager MPixel
  density will reach a critical point where
  throughput will no longer be a critical factor
  rather storage will then prevail as the single
  largest operational concern
• Various Scanners circumvent the 42 minute rule by
  playing tricks with the data and with the optics

20
Current Exploits in High-Speed Wide-field Capture

• Capture at less than diffraction-limited
  resolution (i.e. 20x or even 10x)
• Pre-scan slide at very low magnification to
  identify where tissue is present for subsequent
  high-resolution scanning
• (given that the average histological slide
  uses 38 of the available surface area, this
  makes the worst case scan time of approx. 16
  minutes for some scanner topologies)
• Throw multiple imagers into the fray, thus
  obtaining parallel acquisition channels
• (the latter two have no impact on image quality)

21
So, While Were waiting for 2010 (scan of 1.89
minutes / slide or better)

• What is there for us to debate?
• Venues in which current technology can be
  leveraged
• Educational potential
• Archival potential longitudinal view of patient
• Academic/research potential

22
Uses Today

• Educational compendia
• Rare Slides Database
• Cytology
• Small Biopsies where the block is easily
  exhausted
• Surrogate for FedExTM Pathology for real-time
  consultative pathology
• Cost effective replacement for recuts

23
Uses Today (continued)

• Electronic Proficiency Testing
• Use as a daily documentation tool in routine
  clinical practice
• Rapid access to archival image repositories
  without manifest dependence on conventional
  filing systems (slide file rooms, etc.)
• Digital representation of pathology domain images
  is the first step to true domain-specific
  interoperability (using Radiology as the
  prototypic specialty, via the use of DICOM
  Digital Image Communications of Medicine)

24
Education Arguments in Favor

• Image Acquisition need only be performed once and
  scan duration is not a gating factor for
  assemblage of salient case sets
• Image quality has the potential to approach that
  of a slide, if focal plane issues are addressed
  during scanning
• With identical teaching images, students are
  exposed to a controlled set of archetypical cases
  for enhanced communication of salient image-based
  learning

25
Current Capabilities of Wide-Field Scanning
Technology Education

• Make an exact, diagnostic grade, digital copy of
  one or more slides, for assembly into a
  convenient and inexpensive repository.
• DVD single layer technology enables one to store
  about 15-45 typical slides in JPEG 2000 format
• DVD dual layer technology hold the promise of
  30-90 slides per disc.
• Simplified content dissemination and media use
  for derivative academic works.

26
Education Arguments Against

• The current generation of scanners are relatively
  expensive.
• Scanning benefits from some degree of technical
  imaging expertise, thus the platform may not be
  suitable for all use-cases.
• High cost associated with storage of wide-field
  digital imagery
• High compression ratios, to alleviate storage
  requirements, may render the images to be
  sub-par, even for teaching.
• Some departments may lack infrastructure to
  switch from microscopy-based instruction to a
  computer-based model (need for computer
  labs/workstations etc.)

27
Digital Consultative Pathology Arguments in Favor

• Often (but not always), a subset of the slide
  will suffice for a consultative question, making
  the scan time less than 42 minutes.
• Image quality (with diffraction-based scanning)
  is on a similar caliber with direct viewing of
  microscopy.
• Images may be sent in near real-time over the
  internet, with the availability of suitable
  bandwidth.

28
Recuts Arguments in Favor

• No limit to the number of copies
• Copies are an exact reproduction of the original
• Z representation from digital capture for
  cytology is now possible (Synthetic Microscopy)

29
Synthetic Aperture Microscopy

• Use of mathematical technology from other
  synthetic reconstruction fields (synthetic
  aperture radar, etc) to efficiently model the Z
  continuum for fluid motion focus simulation
• Added benefit of data compression along with fast
  playback
• Reduced to practice

30
Synthetic Microscopy
Based on Agard and Sedat, 1992
QA0A1ZA2Z2A3Z3..AnZn
31
Maintenance of Competency / Proficiency
TestingArguments in Favor

• Simplified generation of cohort sets of challenge
  slides
• All slides identical
• Creates the possibility to share en masse small
  and even minute biopsies
• Increased statistical power of evaluation from
  comparing viewer performance based upon identical
  imagery
• No attrition of disseminated precious material
• Ability to track area of the slide reviewed by
  individual program participants.

32
Centralized Digital Image Repositories Arguments
in Favor Immediately helpful for Integrated
Patient Care Activities

• Centralized repositories for multidisciplinary
  tumor board meetings and longitudinal patient
  treatment planning sessions
• Access point for direct patient education
• Simplified tumor registry image access

33
Centralized Digital Image Repositories Arguments
in Favor Immediately helpful for Integrated
Patient Care Activities

• Immediate personal access to salient imagery
  data, with prior diagnostic annotations (voice,
  overlay, cartographic fiduciary marking for
  relocation of points of interest)
• Reduced dependence on HE recuts
• Simplified sharing of interesting cases with
  colleagues

34
Image Repositories and Interoperability

• Goal create central repositories of electronic
  medical record data
• improve statistical power of retrospective
  studies
• improve outcomes analysis
• simplify reporting of required diagnostic
  entities
• improve economy of healthcare delivery
• enhanced medical education

35
Interoperability Argument in Favor

• Interoperability
• simplify data exchange across disparate
  information systems
• typically carried out with a messaging standard
• based upon pre-coordinated usage of medical
  terminology data structure
• metadata

36
Metadata of Images

• Specimen / patient demographics / prior history
• Accession / slide / block number
• Anatomic location
• Stain / antibody
• Magnification
• Capture equipment
• Packaged with associated XML-based synoptic report

37
Current Promising Pathology Metadata Projects

• DICOM (Digital Image and Communication in
  Medicine) Visible Light Supplement for microscopy
  and endoscopy
• LDIP (Association of Pathology Informatics)
• www.pathologyinformatics.org
• National Cancer Institute (U.S.) Spin Project
  (Shared Pathology Informatics Network)

38
Wide-Field Slide Scanning(a summary)

• Strengths
• Multiple commercially available platforms
• Largely automated image capture process
• Image quality can be of acceptable diagnostic
  quality
• Already useful as an educational / repository
  tool
• Ability to make unlimited copies
• Potentially faster than postal consultative
  pathology
• A necessary gating factor for pathology becoming
  an all-digital specialty

• Weaknesses/limitations
• Can be slow
• Need for extensive data storage infrastructure
• Z-data not routinely captured
• Focus algorithms imperfect
• Interoperability not currently available for
  metadata
• Potential dataset size can be excessively large,
  especially with the use of Z-encoding and storage