High Volume Slide Scanning Architecture and Applications Dr. André Huisman Department of pathology...

22
High Volume Slide Scanning Architecture and Applications

Transcript of High Volume Slide Scanning Architecture and Applications Dr. André Huisman Department of pathology...

High Volume Slide Scanning

Architecture and Applications

Department of pathology UMCU

• UMC Utrecht

• >1,000 beds

• >10,000 employees

• Department of pathology:

• 20.000 surgical pathology cases

• 156.000 glass slides (histology, cytology, IHC)

• 15 pathologists, 10 residents

Digital pathology - advantages

• Digital Archiving

• Instant access from multiple locations by multiple people

• No searching for slides

• Constant quality

• Telepathology

• Consultations, revisions and panels

• Education

• Research

• Automated image processing

Project background

Clinico pathological conferences:

• 900 meetings every year

• No (multi headed) microscope needed

• Quicker preparation of meetings

• No retrieval of glass slides from archive

Aim (2007)

Digitize all diagnostic slides we have

(prospectively)

Challenges

• Scanners

• Image size: up to 1 GB x 500 slides per day

• No existing infrastructure present for storage of this size at our facility

• Image presentation and software integration

• Logistics

Scanners

• Different manufacturers:

• Speed

• Focusing method

• Acquisition technique

• z-stack acquisition

• File format policy

• Application integration

• 2007: 3D Histech (Zeiss), Aperio, Hamamatsu, Olympus (US: Dmetrix,

BioImagene)

• 2010: Leica, Menarini, Philips, Omnyx, BioImagene

Scanning logistics

• 3 Aperio XT scanners (120 slides per scanner)

• Morgue assistants

• Mark slides after scanning

• 6 Hours per run at 20x magnification (~3 minutes/scan)

Storage – HSM

• HSM = Hierarchical Storage Management• Sun Microsystems (Oracle)• 6 TB available on very fast fibre channel disks• 120 TB available on tape (750 GB each)• 2 Tape drives

• Completely transparent archiving and retrieval (robot)

• Access time from tape: 1 - 3 minutes

Linking systems

• 1D Barcodes• U-DPS: reporting system• LMS: Laboratory Management System• Spectrum: Aperio’s image management solution• Storage system

• Own development: integration layer

Scanner workflow

Image / Data server

Scanning

Database

Storage

Other images (e.g. macro)

pathology users

U-DPS

LIS

HSM

120 TB on tape6 TB fast disks

Order formscanner

Speech recognition

Validation

• Aim: validate diagnostic use of digital slides

• Method: reevaluate diagnosis with same pathologist on scanned

slide after washout period (1 year) for several organs

• Gold standard: original diagnosis using ‘traditional’ microscopy

Preliminary results validation

GI tract (%) Skin (%)

Same 92 94

Similar 7 5

Discrepant 1 1

• GI tract discrepancy

• Different interpretation of abnormality

• Glass slide and WSI contained same information

• Skin discrepancy

• Clinical information not used for interpreting digital slide

Pitfalls of digital archive

• Costs

• Huge storage needs – 40 TeraByte per year (over 57,000 CDs)

• Largest storage in the UMCU

• Logistics of scanning up to 500 slides per day

• Currently scanning almost 24 / 7

• Place of scanning in process

• Speed of image retrieval

• Image compression (JPEG 2000?)

• Backup

Education

• All students view the same “best slide”

• Slide images can be integrated with

• Annotations

• Questions

• Macroscopic images

• Other multimedia

• Most UMCU microscopy practical sessions are digital

• Student satisfaction is high

• Place (small) slide scanners at different labs

• Upload digital slides to UMC Utrecht

• Aurora mScope Clinical

• www.pathoconsult.com

Teleconsultation

Upload new media

Digital slide panel discussions

Image processing research

Image processing applications on virtual slides:• Detecting mitotic figures in breast cancer slides• Use texture features

• Establishing histological grade in breast cancer• Segmentation of individual nuclei (on H&E stained

slides)• Detect points of interest• Use marker-controlled watershed segmentation• Post processing

Conclusions

• Routine scanning is possible and makes sense

• Future of pathology is digital

• Digital pathology is expensive

• Digital pathology is just starting..

Together we are shaping Pathology 2.0

Discussion

• Limitations current system

• Cytology

• Speed

• Magnification (20x / 40x incidental)

• Backup

• Quality Control

Discussion

• Archive heavily used

• Educational use still growing

• Teleconsultation network growing (www.slideconsult.com)

• Need for standards

• DICOM / JPEG 2000

• Images, annotations and reports

• Mixing scanners and integration with other software platforms

(middleware?)

• Image management central in workflow for pathologist?

Questions?!

Huisman et al., Creation of a fully digital pathology slide archive by

high-volume tissue slide scanning, Human Pathology, 2010

May; 41(5): 751-7

[email protected]