![Orthanc
• Small
• Flexible
• Quick deployment
• Stable
• Scalable
• Upload via DICOM & Web
• Multi language
Has web interface including drag/drop
Plain text config files
DICOM nodes in separate config
Configuration synchronised by Git
# orthanc_nodes.json
{
"DicomModalities" : {
"siteone": ["SITEONE", "1.2.3.4", 104 ],
"sitetwo": ["SITETWO", ”2.3.4.5", 104 ]
}
}
orthanc and nginx daemons are stable
Auto restart with supervisordOrthanc easily handles our scale
Images stored in external DB
Compiles and installs easily on AWS Linux
Simple directory/symlink config files
Install in /opt/orthanc
Language-independent with REST
We develop & test with curl & Postman
Clients in Python and Matlab
Runs easily in 2 GB, 1 core instance](https://image.slidesharecdn.com/crabbcasestudy-191215213539/85/Image-Management-Using-Orthanc-in-a-Multi-Centre-Reserach-PET-Study-5-320.jpg)
Image Management Using Orthanc in a Multi-Centre Reserach PET Study
- 1. Image Management Using Orthanc in a Multi-Centre Research PET Study OrthancCon I, 2019 Andrew Crabb The Johns Hopkins University I Do Imaging
- 2. Overview • We are a Radiology lab running multiple studies, all different • Five-year research neuro PET pharmaceutical imaging study • Four US academic medical centres • 600 subjects, 2 scans each • Images uploaded by scanner techs or Radiology admins • Research subjects: No Protected Health Information entered • High variability: Subject codes, sites, scanners, staff • Some scanners could not have a DICOM node configured • Some sites insist on encrypted transport
- 3. Neuro PET
- 4. Requirements • A small, flexible solution with quick deployment • Stability – no memory leaks, restart on crash/reboot • Scalable to ~ TB scale • Client upload via DICOM transport and Web interface • Not every DICOM modality can be configured for our node • Security – Encrypted transport • Potential to achieve HIPAA compliance • Client multi language/environment support • Hierarchical development environment
- 5. Orthanc • Small • Flexible • Quick deployment • Stable • Scalable • Upload via DICOM & Web • Multi language Has web interface including drag/drop Plain text config files DICOM nodes in separate config Configuration synchronised by Git # orthanc_nodes.json { "DicomModalities" : { "siteone": ["SITEONE", "1.2.3.4", 104 ], "sitetwo": ["SITETWO", ”2.3.4.5", 104 ] } } orthanc and nginx daemons are stable Auto restart with supervisordOrthanc easily handles our scale Images stored in external DB Compiles and installs easily on AWS Linux Simple directory/symlink config files Install in /opt/orthanc Language-independent with REST We develop & test with curl & Postman Clients in Python and Matlab Runs easily in 2 GB, 1 core instance
- 6. Why a Cloud Solution? • Problems with DICOM transport: • Not (easily) encrypted over public internet • Ports (104 or custom) usually blocked • C-MOVE not possible behind NAT • Cloud location is available from anywhere • Can allow only certain IP ranges • Many features are easily deployed • Encryption of storage and transport • Backups of database instances • Logging of users and actions
- 7. Implementation EC2 RDS 11112 443 Site 1 Site 2 Study Manager
- 9. HIPAA Compliance Requires (amongst others): • Encryption of data in transport • Encryption of data at rest • Access control • Logging of all actions • Configuration monitoring • Verified backup/archive procedure
- 10. Possible HIPAA-Compliant Implementation EC2 RDS 443 Site 2 Study Manager Encryption of data in transport Encryption of data at rest Access control Logging of all actions Configuration monitoring Verified backup/archive procedure
- 11. Clinical Medicine vs. Medical Research Clinical Read individually, now Read by humans Immediate reading Research Processed in batch, later Processed automatically Delayed processing = William / Will ≠ Nicknames = William / Wiliam ≠ Misspelling = Sarah / Sara ≠ Alternate spellings = A010 / AO10 ≠ Spoken digits
- 12. Accommodating Error Every human process includes random error Medicine has a strong culture of reducing and correcting error Automation achieves zero ambiguity by accommodating error
- 13. Why not just fix the errors? • Difficult • Dispersed backups immediately created • PACS systems have limited access • Slow, heavyweight process • No history (provenance) of edits • Only conforms to one view of correctness idoimaging.com/turku2017 ‘Fixing’ is an attempt to say that something did not happen, that did happen The ‘fixing’ process itself will contain error
- 14. Identity “Eddy Merckx” SUBJ1001 Patient Developer Investigator What is the correct identity of this subject? It depends on who you ask.
- 15. Multiple Identities • Name, birthdate, hospital history numberPersonal Identity • Name or code wished to be used at scan timeStudy Identity • Name and history actually used at scan timeScan Identity A person enrolled in a study and who receives a scan has three identities:
- 16. Eliminating Error in Software Development Code Test Input Run Output ✔ ✘ Lightweight Iterative Version control Truth Replicable Correct Deterministic
- 17. Accommodating Error in Identity Mapping Test Images Process Output ✔ ✘ Lightweight Iterative Individual Version control Truth Replicable Correct Deterministic