Free ecosystem for

1 2 Worldwide explosion of medical images

Belgium (2013): 33 millions of imaging studies for 11 millions of people

CT + MRI + PET-CT

Reason: Multimodal and longitudinal imaging (oncology, cardiovascular diseases, surgery, neurology...) 3 The hype: AI for medical imaging 4 The reality: It’s still Jurassic Park! 5

Sharing images (basic need) is painful

inside hospitals hospitals to patients

between hospitals among skilled workers 6 Some terminology: Imaging flows for radiology

Everything is driven by software through the DICOM standard! 7 The real-world difficulties 8 Pain 1: Interoperability and lock-in

Good news: A single worldwide standard in open-access!

• The DICOM standard is very complex, both for users and developers. • Many specialized vendors, with costly, proprietary and monolithic ecosystems ⇒ highriskof high risk of lock-in, few agility. • Interoperability is checked in “Connectathons” where vendors meet (N² complexity) ⇒ highriskof no reference implementation. • Not every PACS comes with teleradiology (remote expertise) ⇒ highriskof need to combine vendors. • Few IT expertise in hospitals about imaging⇒ highriskof need to share knowledge.

Heterogeneous modalities ▶ very problematic in emerging economies! 9 Pain 2: The PACS is focused on radiology

• Hard to retrieve and access the raw DICOM files. • The PACS must be interfaced with specialized software (nuclear medicine, radiotherapy, dentistry, neurosurgery…). • Multitude of files, as images are split slice-by-slice (one typical 3D image = 1000 DICOM files, 500MB) ⇒ highriskof need for standalone server to ensure continuity of care (cancers, move, rare diseases...). • Patients and skilled workers haven’t access to professional tools at home, besides basic viewers.

Especially problematic for patient empowerment and training! 10 Pain 3: Automation of imaging workflow

University Hospital of Liège: 300 modalities

Every hospital redevelops its own scripts ▶ huge cost inefficiency!

12 • Get back in control of imaging workflows • Need for automation • Interfacing specialized software • Academic: Research, teaching, QA 2011

• PACS without a RIS and reporting (VNA) • Focused on simplicity and portability • Built-in support of Web technologies • Industrial grade 2012 • Libre software (GPL/AGPL)

Virtual & Lightweight Rest API cross-platform Extensible 13 in the hospital: Ancillary PACS 14 Web interface in action

• Transparent indexing: Patient → Study → Series → Instance. • Preview 2D images. • Send to other modalities in one click. • Inspect medical data, ZIP, anonymize... 15 Orthanc as a microservice for medical imaging

External applications

REST API

DICOM Modalities Lua scripts

Plugin SDK

Database engines Embedded Web applications (default: SQLite) (servlets) 16 Free and open-source plugins for Orthanc

Whole-slide imaging Basic teleradiology

Advanced teleradiology

DICOMweb 17 Some external applications using the REST API

Diabetes screening (fundus) Quality control (ImageJ)

Pharmaceutical studies (anonymization) Teaching and research (download from PACS) 18 Stone of Orthanc

• Innovative C++ library to render medical images • Fully CPU-based for maximum portability • Compatible with WebAssembly • Support advanced data: PET-CT fusion, MPR, doses, contours… • Goal = quick development: build a new viewer in some days

Will be at the core of our next generation of viewers 19 Reference paper in open-access

21 Positive feedback from the community

2014 Award for the Advancement of

2015 Best e-Health project (from the Belgian industry)

● One of the two main free PACS around the world (with dcm4che) ● > 200,000 downloads ● > 550 members on the forum ● > 1,200 threads on the forum ● > 170,000 lines of C++ code (35 men-years effort)

● Availability on popular NAS (QNAP and Synology) ● Docker-friendly + Standard Base (LSB) binaries ● Official packages for and Fedora (no openSUSE yet ;-) ) 22

First spin-off of University Hospital of Liège (2015) 23 Business model

Support Custom developments Packaged versions of Orthanc (independent centers) (&D for industry) (for e.g. EU/USA hospitals)

2 FTE on Orthanc (virtuous circle) 24 Services (1/3): Automating exchanges between hospitals

Oncology, continuity of care, rationalization of studies, ionizing radiations, clinical research…

Interface for reconciliation 25 Services (2/3): ORU-PACA = Telestrokes in France

(55 emergency departments) 26 Services (3/3): ERN = Tele-expertise for rare diseases 27 References

(80 clients in 12 countries) (17 installations in 4 countries) 28 Conclusions 30 Summary

● “Backbone” software for medical imaging (VNA). ● Open-source, lightweight, scriptable, extensible. ● Originates from real-world, clinical needs. ● Large, worldwide community of users.

● Commercial partner of the Orthanc project. ● First spin-off company of the University Hospital of Liège. ● Open-source business model selling services to the industry. ● Sells packaged versions of Orthanc to hospitals. 31 Interfacing with GNU Health

Access to Orthanc Web viewer from the patient record

Populate DICOM modality worklist (schedule studies)

Drive transfers to/from other centers 32 Perspectives for emerging economies (portable X-rays)

+ =

(1) Low-cost PACS in (2) Collect images, analyze (3) Remote expertise remote locations in hospitals (cf. WWI) with Web viewers (+ open AI algorithms for auto-diagnostics) 3G, satellite Transfers accelerator (4) Transmit images on plugin (cf. BitTorrent) bad connectivity