Digital imaging in Pathology: Clinical Perspective

1
Digital imaging in PathologyClinical Perspective

• Michael Feldman, MD, PhD
• University of Pennsylvania
• feldmanm_at_mail.med.upenn.edu

2
Outline

• Historical overview
• Clinical/Research
• Primary diagnostics
• CAD
• CBIR
• Telepathology (Distributed diagnostics)
• Microscope replacement
• All cases
• Subset cases
• Immunohistochemistry
• Quantitative
• Flow on a slide, cell based analysis of one or
  more markers
• Automated QA/QC
• Radiology Pathology integration

3
How much progress since the 18th C.?
van Leeuwenhoek (1632-1723) microscope with an
aspherical lens--magnifies 295x.
http//www.brianjford.com/wav-spc.htm and R
Levenson Futurescape CAP presentation CRI inc
4
Developments in microscope-based pathology
1700s Microscopy 1850s Hematoxylin Cell
theory of disease and cancer (Virchow) 1870s Eo
sin 1890s Formaldehyde 1940s Immunohistochem
istry (direct fluorescence) 1970s In-situ
hybridization (radioactive DNA) 1990s The
computer
R Levenson Futurescape CAP presentation CRI inc
5
Outline

• Historical overview
• Clinical/Research
• Primary diagnostics
• CAD
• CBIR
• Telepathology (Distributed diagnostics)
• Microscope replacement
• All cases
• Subset cases
• Immunohistochemistry
• Quantitative
• Flow on a slide, cell based analysis of one or
  more markers
• Automated QA/QC
• Radiology Pathology integration

6
Computer Assisted Diagnosis

• Not an MD replacement !!!
• Assist MD in rendering best
• Diagnosis
• Prognosis
• Theragnosis
• Perform function human cannot
• Attend to high dimensional data
• Attend to fused data
• Attend to scaled data

7
Computer Assisted Diagnosis

• Morphometrics - Use case Bladder Cancer

Primary Metrics - Size - Shape - NC
ratio Higher order metrics - Chromatin - Nuclear
contours - Z-axis
Dysplastic/TCC
Reactive
8
Morphometrics Already Recognized in Some
Classification Systems but Manual Collection of
Stereological Data Prohibitive

• ANCONA INTERNATIONAL CONSULTATION SYSTEM
• for Flat Bladder Cancers
• BENIGN EPITHELIAL ABNORMALITIES
• EPITHELIAL ABNORMALITY
• REACTIVE EPITHELIAL ATYPIA
• FLAT UROTHELIAL HYPERPLASIA
• PRESUMMED PRENEOPLASTIC LESIONS AND CONDITIONS
• KERATINIZING SQUAMOUS METAPLASIA
• GLANDULAR METAPLASIA
• MALIGNANCY ASSOCIATED CELLULAR CHANGES
• (CYTOLOGICAL ABNORMALITY NOT RECOGNIZED BY
  LIGHT MICROSCOPY)
• PRENEOPLASTIC LESIONS
• DYSPLASIA
• CARCINOMA IN SITU

9
Computer Assisted Diagnosis

• Morphometrics - Use case Bladder Cancer

Feulgen stain for ploidy Proteins P21,p27, p53,
pRB NMP22 C Factor H Bladder Mucins CEA Urovisio
n FISH for centromeric Ch 9p21, 3, 7, 17
Reactive
Dysplastic/TCC
10
Prostate cancer CAD

• Each pass of the image rejects pixels, and then
  only the positive pixels are analyzed at higher
  scales
• This allows us to efficiently analyze the image
  at higher scales by only looking at interesting
  pixels
• Increasing accuracy does NOT increase execution
  time

11
Higher scale analyses
12
Non-Linear dimension reduction followed by SVM
G3 vs. BE 85.43 G4 vs. BE 92.60 G3 vs. G4
95.80
Blue CAP3 Green CAP4 Red - Benign
13
Content Based Image Retrieval

• Pathology is visually based
• How do you get case examples/images for
  Differential diagnosis
• Query by textual metadata. Only as good as your
  annotation
• Query by image as a web service would be very
  valuable

14
Telepathology

• Support remote diagnostics
• Streaming video
• Virtual slide /- lookback
• Business Model
• Health System Networks project expertise
• Expert consultation
• Clarient Inc
• CAD and CBIR additive in this use case

15
Microscope replacement

• Is the microscope dead
• Can we do a full clinical case load by Vslide?
• Current systems not fast enough for large
  workload
• Integration into AP digital workflow still
  primitive
• Cost is very high if try and replace scope
• Added cost vs conventional microscope
• Image quality ???
• Selected slide scanning
• More reasonable in todays environment
• Tumor boards
• Reduced historical look back
• Health networks with distributed experts
• Second opinion business where slides have to be
  returned

16
Outline

• Historical overview
• Clinical/Research
• Primary diagnostics
• CAD
• CBIR
• Telepathology (Distributed diagnostics)
• Microscope replacement
• All cases
• Subset cases
• Immunohistochemistry
• Quantitative
• Flow on a slide, cell based analysis of one or
  more markers
• Automated QA/QC
• Radiology Pathology integration

17
Immunohistochemistry Support

• Historical technique 1970s
• Used for
• Tumor of origin (Qualitative, maybe quantitative)
• Prognostics (quantitative 0-3 manual scoring)
• Manual interobserverreliability issues

18
Color and Spectra
Light has no color. Color is an observer
interpretation. These two yellows appear
identical to the human eye, yet they have very
different spectral components.
19
3. Image processing to resolve individual stains
based on spectra
teach computer the spectral profile
of chromogen/fluorophore
extract spectra of individual stains
A, B, C
A
B
C
20
6. Computational assignment of immunostains to
each nucleus
breast tumor stained for p-ERK (DAB) hematoxylin
segmentation of nuclei (based on hematoxylin)
p-ERK- stromal cells (grey)
p-ERK tumor cells (yellow)
21
Data display AnalysisFrequency histogram of
intensity of p-ERK staining of stromal and tumor
cell nuclei in a breast tumor
tumor cell nuclei
stromal cell nuclei
22
pERK
Ki67
23
(No Transcript)
24
Immunohistochemistry QA/QC

• Daily QA function in Immunohistochemistry
  laboratory
• Monitor stain quality
• Manual process make sure stains worked as
  expected
• Highly repetitive on similar tissue type with
  similar patterns and intensity
• Amenable to automation (Pattern, location and
  intensity)
• MD FTE 2-3 hr/day for a busy lab

Cam 5.2
CD20
25
Rare Event - SLN Melanoma
26
Outline

• Historical overview
• Clinical/Research
• Primary diagnostics
• CAD
• CBIR
• Telepathology (Distributed diagnostics)
• Microscope replacement
• All cases
• Subset cases
• Immunohistochemistry
• Quantitative
• Flow on a slide, cell based analysis of one or
  more markers
• Automated QA/QC
• Radiology Pathology integration

27
Radiology- Pathology ConvergenceThe Elements
28
Multi-modal Registration Paradigm
Histology section with ground truth
Similarity metric
Cancer ground truth on registered MRI
Registered MRI
Corresponding MRI
29
MRI of Prostatectomy with 4T Magnet at 0.8
mmHistopathology Correlation
17









Magnetic resonance imaging provides signal
contrast (MR stain) that allows for the
identification of carcinoma and benign
hyperplasia similar to a 2-4X optical lens.
Adenocarcinoma interrupts normal curvilinear
architecture


BPH without carcinoma

30
Core Philosophy Multi-modal data Integration
31
Radiology-Pathology Convergence- During Your
Career ? Answer Just Around the Corner
32
Diagnostic Imaging today
MRI
In Vivo Optical
Pathology
Molecular
Proteomic
33
Thanks

• Rutgers
• Anant Madhabushi
• Scott Doyle
• John Chappelow
• James Monaco
• Cri
• Rich Levenson
• Cliff Hoyt
• Radiology
• Mitch Schnall
• Mark Rosen
• Penn
• Bill Lee
• Wiem Lassoud
• RPI
• Badri Roysam