Bioimaging Bank: Linking Medical Imaging and Biobanking Data for - - PowerPoint PPT Presentation

Bioimaging Bank: Linking Medical Imaging and Biobanking Data for Cancer Research

Daniel Moses

BSc(Hons) MBBS Hons MEngSc PhD FRANZCR DABR

Medical Director Research Imaging NSW Level 1 Building 3, Prince of Wales Hospital. https://research.unsw.edu.au/research-imaging-nsw Professor, Graduate School of Biomedical Engineering, UNSW

The Project in a Nutshell

EXISTING DATABASES: POWH and UNSW HSA Biobank project:

• standardised universal consent process, collecting and linking both clinical and biological data to specimens
• dedicated biobanking database contains biological information related to the tumours from approximately 5000 patients treated at POWH.
• specimens are stored, managed and distributed from the UNSW Biorepository

Medical image data:

• stored in a large RIS/PACS in the Medical Imaging Department of Prince of Wales
• includes all radiology and nuclear medicine studies since 2008

AIM: Linking clinical imaging data to the established biobanking database facilitating much easier examination of both biological and imaging data in specific cohorts of interest and allowing:

• improved imaging feature extraction from medical images for predicting biological properties of tumours
• combining medical image and biological features of tumours into a prognostic model
• improved personalised treatment

OUTPUTS:

• Extend the pathway for extracting medical imaging data to other partners in the South Eastern Sydney Local Health District
• Promote and support research by linking medical imaging data and biological tissue data for cancer patients through a single comprehensive

database

• Bioimaging bank would be available to any researcher in Australia with a suitable research project and with ethics approval on application

Key Issues

Ethics Extraction and Matching of Existing Patients from Biobank Database

• csv file

Radiology Information System (RIS)

• installed in 2008 (Centricity RIS-i 4.2 Plus) and runs on Windows Server 2003 and an Oracle database
• contains all of the patient report and examination data performed in the SESLHD (including Prince of Wales Hospital)
• since 2008 and includes 3.4 million recorded procedures

Importation of Medical Imaging Reports into the Biobank Database

• OpenSpecimen 6.1 with an Oracle database
• more fields need to be created within the database for each patient
• attempt to import the large number of medical imaging reports into the current kept crashing the system

Participant_PPID Participant_Registration Date Participant_Gender Participant_Date Of Birth Participant_Consent Date Participant_MRN Site ID Participant_MRN Site Participant_MRN Participant_MRN Site ID - 1 Participant_MRN Site - 1 Participant_MRN - 1 Participant_MRN Site ID - 2 Participant_MRN Site - 2 Participant_MRN - 2 Participant_MRN Site ID - 3 Participant_MRN Site - 3 Participant_MRN - 3 Participant_Consent Comments

Lessons Learnt

Ethics Coordinating Between Large Institutions Dealing with a Diverse Operational Team Communication and Collaborations with Key IT Staff Outside of the Project Patient Privacy and Data Security

• Transfer of data
• RIS managers in the Medical Imaging Department at POWH allowed password access to a database

Dealing with Outdated and Fragile Software/Computer Systems

• RIS: used a “view” of database - functioned as a protective mechanism

→ only allowed us to query a confined portion of the database without using too many database resources

• Biobank: support team advised an upgrade to a mySQL database which works better with the OpenSpecimen & is more stable

Custom Writing Software

• Python programming language

○ cx_oracle python library to connect to and query the database (https://oracle.github.io/python-cx_Oracle/) ○ python csv library to read & write csv files (https://docs.python.org/3/library/csv.html)

• patients to query are listed in the csv file supplied by Biobank
• each time query is run it will generate all reports for one patient

○ script loops through each patient, querying the database and recording the result ○ recorded in a local database that the query has been performed on that patient on such-and-such a date ○ built-in time delay to avoid spamming the server

Acknowledgements

Dr Laughlin Dawes (Radiologist and Computer Scientist at POWH) Dr Jitendra Jonnagaddala (Data Scientist, UNSW Medicine) Dr Anusha Hettiaratchi, (Manager, UNSW Biorepository, Mark Wainwright Analytical Centre[MWAC]) Dr Carl Power (Head, Biological Resources Imaging Laboratory in the MWAC) Dr Stephanie Macmillan (Operations Manager,TCRN) Professor David Goldstein (Director, TCRN, UNSW)