Implementation of Subspecialty Whole Slide Image Teaching Sets

Implementation of Subspecialty Whole Slide Image Teaching Sets for Residency Education Bryan Dangott, M. D. University of Utah Hematopathology Fellow

The UPMC Health System • Diversified environment ( academic hospitals, community hospitals) • 19 hospitals with another 18 affiliations • 2 million + outpatient visits per year

University of Pittsburgh Medical Center www. upmc. com 29 Western PA Counties 19 Hospitals 40 Cancer Center Locations

Pathology Centers of Excellence (COEs) at UPMC Presbyterian Dermatopathology ENT/Thyroid Pathology Gastrointestinal Pathology Hematopathology Neuropathology Thoracic/Mediastinal Pathology Transplant Pathology UPMC Magee-Womens Hospital Breast Pathology Gynecologic Cytopathology Obstetric/Gynecologic Pathology UPMC Shadyside Pathology Informatics Genitourinary Pathology Bone/Soft Tissue/Melanoma Pathology UPMC Childrens Hospital Pediatric Pathology

Background • UPMC is a multi-site institution – Residents rotate at different facilities • Problems with glass sets – Only available to one person at any given time – Missing slides – Broken – Disorganized – Degradation (cover slip adhesion, stain quality)

Digital Teaching Material • AP Teaching Site (http: //path. upmc. edu/cases/index. html) • 700 cases – Prepared by the residents – Gross images, microscopic snap shots – Comprehensive discussions

TELE-EDUCATION: Online Conferences

Creating Whole Slide Image Sets • Scanning began in 2005 with GU • Over 2900 whole slide images – General Unknowns collection – Genitourinary – Head and Neck – Thoracic – Hematopathology • Under Development – GI, Bone and Soft tissue, Cytology

Infrastructure • Storage space – Storage Area Network for image files • 20 terabytes. – Supplemental file storage space. – Web hosting space.

Servers • Image server – Aperio Image. Server • Webscope, Flash viewer • Image. Scope, Desktop Application – Zeiss, Java viewer • Web servers – Apache – IIS – Cold. Fusion

Database Backend • Oracle – Unknowns – GU – Thoracic – Head and Neck • Microsoft Access – Hemepath – Testing Interface

Front End Language • Cold. Fusion – Database searching – Page presentation and formatting • Javascript – Used to selectively display portions of the page – Selects the viewer to launch based on link

Basic Procedures • Gather data – Clinical Information – Diagnostic Material • • Slides Radiology Gross Images Flow, Cytogenetics, Molecular data • De-identify data

Scan Hardware • • Aperio Scanscope XT (120 slide loader) Aperio Scanscope CS (5 slide loader) Zeiss Mirax MIDI (12 slide loader) Hamamatsu (210 slide loader)

Scanning Challenges • • • Tape style cover slips Scratched / damaged slides Uneven tissue sections Very thin sections of fatty tissue. Peripheral smears / Aspirate smears Cytology specimens

Scanning Challenges • Cytology and hematopathology – Focus points – Depth of field • Scan time – 40 x hemepath • 30 bone marrow aspirate slides (~24 hrs) • Batch scanning

Personnel • • • Web Developer – Thomas Harper Slide Scanning – Jon Duboy Server Administration – Ishtiaque Ahmed Project Coordinator – Doris Dayoub Project Oversight – Anil Parwani, MD, Ph. D

Implementation by Subspecialty Unknowns Head and Neck Genitourinary Thoracic Hemepath Database Oracle Access Front End Cold. Fusion Image. Server Aperio / Zeiss Aperio Data Entry Web Interface Direct Entry Updates Weekly One Time Batch Continuous Clinical History Limited Moderate Extensive

Stats by Subspecialty Unknowns Head and Neck Genitourinary Thoracic Hemepath Total Cases 340 130 300 806 173 1, 749 Total Slides 400 158 340 983 1083 2, 964 Slides per Case 1. 18 1. 22 1. 13 1. 22 6. 26 Uncompressed Total Data (GB) 2, 122 897 3, 141 6, 266 20, 450 32, 876 Compressed Data /Storage Space (GB) 98 66. 2 79. 9 282 1520 2, 046 Overall Compression Ratio 21. 67 13. 5 39. 3 23. 4 14. 1

Capabilities Unknowns Head and Neck Whole Slide Links 8 Ancillary Image/File Links 3 Free Text Fields 1 8 7 Genitourinary Thoracic Hemepath 8 5 Unlimited 10 Unlimited 7 14 7 Other Data Fields 12 Diagnosis X X Discussion X X ICD-O Codes X X X

Challenges • Managing images, web page code, clinical data sources. • Customization of individual sites – Each of the subspecialties has different needs

From Template to Flexible • Hemepath vision – More slides – Higher magnification (40 x minimum) – More slide area (smears) – More data types

Two Projects at a Time • Hemepath teaching set was built simultaneously with a testing interface. – Required architecture flexibility – Required more involvement with the developers

Hemepath Data Types • • • CBC Differential counts Molecular data Immunophenotyping Flow Cytometry Cytogenetics

Hemepath Database Design • Access – Up front design required a lot of interaction – Hemepath wanted to manage the database • This eliminated building a web interface for data entry – Allowed additional development to enhance data entry

Work Flow Improvements • Batch data download from APLIS – Tab delimited text file • Batch data processing – Vbscript – Notepad++ vs Excel • Data Entry – Copy and paste data fields – Link integration

Hemepath File Naming • File naming conventions –CASE# - TYPEID - FREETEXT – WSI • 001 -CBX-HE – Supplemental Files • 001 -FLO-CD 2, CD 3, CD 4, CD 8

Hemepath File Naming Slide_Type Slide_Abbv aspirate ASP clot CLT core touch prep cytogenetics CYT CBx flow FLO CTP molecular MOL other OTH reference REF static slide image SSI whole slide WSI lymph node LN lymph node touch prep LNTP other OTH peripheral blood PB soft tissue Link. Type_Name Link_Abbv ST

Special Features Unknowns Head and Neck Testing Interface X Genitourinary X Thoracic Hemepath X Auto Import WSI Links X Auto Import Ancillary Files X Link Sorting X Naming Convention X Multiplatform Scans X

Testing Interface

Future Goals • Cross site searching • Expansion of the testing interface to other sites • Further development of other subspecialties

Credits • Contributors – – – Jon Duboy Li Li Joseph Sirintrapun Gaurav Sharma Doris Dayoub Ishtiaque Ahmed Thomas Harper Christine Garcia Fiona Craig E. Leon Barnes Samuel Yousem – Project Oversight – Anil Parwani, Liron Pantanowitz, Trevor Macpherson