Proceedings of OrthancCon 2019

“Which organs are damaged ?“ and “How serious is the patient condition ?” are the two main questions to answer for any healthcare professional. These answers save lives. Current diagnosis tools are expensive, results long to get, they don’t help provide quick answers.

echOpen is a handheld low-cost ultrasound imagery tool enabling diagnosis orientation. It is connected to any mobile device allowing to see inside the body and immediately identify patients painpoints.

Medical imaging is a complex and diverse field that depends upon software. Many software projects are begun in this area, ranging from small special-purpose utilities from an individual, to packages developed by a funded group. We might measure the success of a free software project by its widespread and long-lasting adoption. If the project is open-source, we might further include the involvement of, and contributions from, the developer community. By these standards, not all projects are successful. What are the common attributes of those that are?

In this talk we will look at the record of some free and open-source medical imaging software from the past 20 years. Some projects, addressing a small and stable problem, have succeeded despite modest use and infrequent updates. Other teams have focused on larger or rapidly-changing domains, with varying results. We will discuss definitions of success, and what makes a software project stand out in this crowded field.

The Cytomine software is a virtual microscope that allows you to securely share your whole slide images over internet, and fosters their collaborative annotation and analysis with distant users.

Cytomine is designed to be used in education, research or diagnostics, in every domain where collaborative analysis of large amounts of images is needed, as in Digital Histoloy and Pathology.

As every full web platform, Cytomine is available anywhere, at anytime, and on any device having an internet connection and a web browser. Additionally, Cytomine is able to run AI algorithms for semi-automated image analysis.

Providing a documented API and several clients, Cytomine can be easily linked to other software and integrated into any work environment.

Released under an Apache license, Cytomine is vendor free as it support various whole slide images formats.

Using Docker, Cytomine can be deployed on infrastructures adapted to your needs, ranging from personal desktops to platforms of thousand of users.

The deployment of EHRs in the cloud offer huge advantages for the sharing of data among healthcare parties but faces increasing scrutiny in terms of data governance. The highly sensitive nature of medical data makes it a first class target for undesired data collection or unsanctioned use. We propose a distributed, trust based and open source platform that allows the sharing of data among previously known actors and compare it to existing initiative in Belgium for the sharing of information among healthcare professionals.

The website medfloss.org provides a platform for retrieving medical FLOSS projects. Currently, it lists 361 projects that are described by certain characteristics like application type, license, and enterprise functions supported.

The vocabulary used to describe the FLOSS projects does not describe the functionalities of FLOSS projects sufficiently in many cases. An ontology enabling systematic descriptions of medical software could help developers to clearly specify the software’s functionalities and make FLOSS projects on medfloss.org even better retrievable and comparable.

HITO, the health IT ontology, is developed by an Austrian-German research team in order to support systematic descriptions of medical software. HITO integrates different existing taxonomies (e.g. parts of SNOMED CT, WHO classification of digital interventions) and software characteristics found in literature.

To describe FLOSS projects on medfloss.org using HITO, the support of the FLOSS community is needed.

The objective is to discuss how to help to solve health issues using information technology.

The first step is to analyse the current methodology of reasoning of experienced doctors. The approach is to try to represent natural intelligence in a very objective way, as visual graphs. Therefore graphs are useful because they can simulate the neural network of the human mind. Indeed a large number of about 10**15 neurons are connected in many ways as an N dimensional space. Relationships between neurons or groups of neurons should be quantified among as being relatively weak or stronger. An usual strategy is that the current set of available observations can raise or exclude hypotheses. Likelyhood are then futher evaluated in order to decide what to do next. Or where to look next in case of step by step image analysis. Moreover visual graphs can help specialized Care Team members to share information about a patient. A kind of synchronization between their minds, using very natural visual presentations.

New software tools working on graphs can provide recommendations. Some typical approaches will be discussed, as cluster detection, similarity, multifactorial decision, etc...

From the point of social and educational objectives, the challenge in developing regions is to provide useful recommendations. If local nurses can record their observations and own hypotheses as a graph, recommendations could be very well explained in a visual way. Cases could also become easier to review accross telemedicine. Seeking partners for the Open Source experimental project Interactive Medical Mind Maps.

This presentation will explain why health organizations around the world benefit from using Tryton.

We will start by explaining the origin of Tryton and the reasons behind the creation of the Tryton Foundation. Those initial choices have a huge impact on how the project lives and is organized. Tryton is thus naturally suited of health organizations inf emerging economies.

Next we will cover that it is not its only key advantage. By relying on sound engineering principles about accounting and stock management, Tryton has offered to GNU Health a solid base on which building its suite of health-dedicated modules.

Medical imaging has a long history of free and open source software, benefiting from the ethos of sharing knowledge which is central to the software and academic worlds. Open source software consistently leads the adoption of new trends in medical imaging, software development, and computing practices.

We will look at these changing trends, and the software projects that address them. Imaging itself evolves, with ever-increasing resolution and new modalities such as digital pathology. Software development changes much more rapidly and we see projects using new, or newly-popular, languages. Modern computing platforms - cloud, web services, containers - lead to new or repackaged projects.

Yet there are top software projects that are decades old and have kept their lead through continued refinement. We will look at these and the wide world outside of clinical DICOM systems: research pipelines, image analysis, batch processing and lesser-served imaging modalities and formats.

Teledermoscopy: Skin cancer has risen sharply in recent years. In 2015, it is the 4th most frequent among women. Having an appointment at the specialist for a consultation can take 8 to 12 weeks. This projet TeleSPOT (Teledermoscopy Smartphone-based Pigmented lesion diagnosis Online Taskforce) aims to propose a rapid and reliable diagnostic aid in the face of a suspicious pigment lesion, accelerate the care of very suspicious malignant lesions and revalorize the first line of care in screening and early diagnosis.

Teleanapathology: CHU Liège has multiple remote sites where frozen sections (extemporanés) exams are realized and for which anapathologists give their diagnosis. Some of these sites are too far to be directly connected to the CHU network. Anapathologists have sometimes to drive several hours to give their diagnosis. The aims of this project are to make use of the new digital slide scanner and orthanc API to save time to the anapathologist and to make available the whole slide images to all laboratories.

We developed a free and open source (AGPLv3) web app for clinical trial imaging centralization including, DICOM deidentification, storage, quality control, expert's interpretation and clinical trial management.

The web app using Orthanc as backend is built to manage multiple trials with users having one or multiple roles (Investigator, Controller, Monitor, Supervisor, Reviewer) in each trial.

At the study creation the administrators define the number of imaging procedure in the trial, the expected patients and define the trial specific forms to be filled.

Investigators are allowed to send DICOM through the web app after local deidentification.

Reviewers can download DICOM from the web interface with the software of their choice and could also benefit from a direct viewer integration such as FIJI (using RestAPIs).

Supervisor's dedicated interface shows the trial progress, reviews answers and integrates tools such as upload reminder, audit trail and images download interface.

Improvement of radiation safety culture in medicine is a challenge as it is in other critical industrial fields. However, in order to ensure of a certain level of process quality achievement, checklists have only been introduced during the last two decades in hospitals, especially in the operating rooms (i.e. : WHO Surgical safety checklist).

In the way to improve safety and efficiency in our Radiation Oncology department, we initiated an inhouse software development to help us on having an overview of the workload.

iTP is a web-based open-source tool which is able to display the list of all the patients being supported in a department. Workload per role is displayed through a colour code. Entering the patient chart will bring you to a detailed checklist of the different items to be covered and finalized at a specific stage of the whole process.

We have now an experience covering thousands of patients for which the treatment plans process preparation has been driven through this software.

Purpose: To describe the open-source software platform CAPTAIN dedicated to research in radiation oncology and more specifically in proton therapy. The software provides online QA and decisionmaking tools and serve as a platform for the development of automatic adaptive re-planning tools. It offers the possibility to recompute proton doses maps with an independent Monte-Carlo dose-engine (MC-square), estimate clinical goals, DVH and gamma analysis and normal tissue complication probability (NTCP) model computation using the MATLAB library OPEN-REGGUI and python script. The architecture of the tool is very modular to allow researchers to use each piece independently.

Orthanc is one of the most important building blocks in the development of these workflows. It’s the DICOM interface between the CAPTAIN platform and the external medical software: treatment planning system, the imaging system, Oncology Information System,…. For each new DICOM object received by Orthanc, it notifies CAPTAIN, which then triggers the appropriate worklow among those implemented. The results are accessible to the researchers, physician and physicist remotely via a web-browser on any device. Available workflows are (1) indications for adaptive therapy, machine log QA, and NTCP computation.

Results: Daily anatomical changes are evaluated using the in-room CT or virtual CT (vCT) computed by deforming the pCT onto the daily CBCT. The estimated fractional dose expected to be delivered by the plan on the daily image is computed and used to determine the need for adaptive replanning. The same kind of indicators could also be evaluated by performing a dose reconstruction on the CT image using information recorded in the irradiation logs, i.e. using information about the actually delivered beam instead of the planned beam. in order to accelerate the process of patient specific QA. This last workflow in the CAPTAIN platform was used as the research prototype for the development of the medical software myQAion. A third workflow automatically computes the NTCP when CT, structure set, treatment plan and dose map are available on Orthanc. The computation is repeated for every new dose map. It provides a demonstrator of implementation of the model-based approach of patient selection.

Orthanc also offer the possibility of 3D dose map visualization, overlaid onto the CT scan and the RTstructure within stone-of-Orthanc web-viewer to support the user in his research.

Conclusion: CAPTAIN is an open source platform for translational research research in PT and RT, facilitating the development and evaluation of algorithms for clinical research studies. The application is of interest for clinical research in adaptive proton-therapy and it helps the clinical user to test, validate and evaluate the workflows developed by other researchers in proton therapy. Several workflows are already implemented, and the modularity of the tool offers possibility to anyone to implement his own application. In that platform Orthanc is a major piece of the architecture. The open-source approach allows research groups to develop their own workflow to share with the community to speed up research in PT.

Hospital Taiping is a 608-bed hospital in Malaysia which was established in 1880. Prior to the deployment of Orthanc, radiology images and reports were printed on film to be distributed to the wards and clinics.

We introduced Orthanc supported with PostgreSQL in August 2017 with a custom front-end for search and reporting, on a desktop computer (Intel Core i5 with 4Gb RAM), which functioned as a mini-PACS for the radiology department. Multiple Orthanc instances were connected to the same PostgreSQL database to improve performance. In addition, Orthanc also serves as the backend for a custom mobile app for remote image viewing. Through negotiations, we managed to integrate all available modalities to Orthanc. Clinician acceptance was high, and problems were centered around our custom front-end.

With a successful proof-of-concept, we obtained funding and in July 2019, migrated to a small server (Intel Xeon with 8Gb RAM) for expansion of PACS services to the entire hospital. Our database is currently at 2.2Tb, consisting of approximately 80,000 studies. Future plans include load-balancing and an improved disaster recovery solution.

We present a case study of a long-term research imaging study involving multiple academic PET imaging centres. A cloud-based Orthanc installation is used to provide a central image repository to receive and store PET, CT, and MR images. We use the DIMSE, REST API, and graphical interfaces of Orthanc as appropriate for image collection and management. A range of cloud services provide for easy data archiving and provide secure and validated access.

Research imaging studies pose different challenges than clinical practice, especially in the variability of scanners, imaging protocols, and image header contents between centres and over time. We describe how the Orthanc API allows us to query the database of image metadata to monitor the study and assist in designing a flexible analysis workflow to accommodate differences between scans. A future expansion of the system will be built upon the API to allow study-wide image processing, accommodate errors in subject names, and study-specific image de-identification.

Case-based learning (CBL) is an efficient method for radiologist education. RadioLogic is a system to create clinical cases from real DICOM files and provides a self-assessment tool to view the studies, submit a diagnosis and compare the performance with peers. The main components are a progressive web application and an Orthanc plugin to create and serve the teaching cases. The user selects a learning module, views the cases, submits his diagnosis and checks the results. The teacher selects a patient, study, series or instance inside an Orthanc server to start the creation of a clinical case in a webpage. He enters the name of the clinical case, the choices for a possible diagnosis, the answer and two images showing clinical data and explanations about the correct diagnosis. An asynchronous job is started in the Orthanc job engine to customize and compress the case and to generate two DICOM files from these images including private tags to hold the submitted data.

In 1988, the first brain activation studies, relying on Positron Emission Tomography, were reported. Back then, regional differences between brain images were simply characterised from hand-drawn regions of interest. The idea of producing a Statistical Parametric Map (SPM), i.e. providing valid inferences about signals across the entire brain, was subsequently formally introduced in 1990 by Friston et al. The SPM software, implementing these concepts in a complete analysis package, was first released in 1991 and made freely available to the neuroimaging community, to promote collaboration and a common analysis scheme across laboratories.

Nowadays SPM remains the most used software for the analysis of neuroimaging data and has been extended over the years to handle functional and structural MRI, along with EEG/MEG data. An ecosystem of third-party extensions also flourished to address specific issues (over 60 at the time of writing). Some of them constitute toolboxes on their own, for example on the application on machine learning techniques in neuromaging or the processing of quantitative MRI.

Orthanc Tools is a standalone application written in Java to provide high level services using the power of Orthanc Rest APIs.This software could be described into 3 axes :

• An Orthanc GUI where most of the Rest APIs can be used through an user friendly GUI (display patients/studies/series, tag edition, deletion, Query/Retrieve/Send).
• High level services built on the top of Orthanc such as :
  • Automatic Query, to retrieve a batch of studies from a PACS with one predefined list of queries and a fully automatic retrieve
  • Anonymize a batch of patient with management of different anonymization profile.
  • CD/DVD burner management, automatically trigger CD/DVD burning from a simple DICOM transfer on Epson or Primera burning DiscProducer, with queue management, portable DICOM viewer, label printing
  • Auto Fetch, retrieve automatically history from a PACS when a patient is received by Orthanc
• Some Orthanc utils, such as Orthanc config json edition and generation through a GUI.

According to the TIOBE August 2019 index, Java is the most popular language, while C++ is fourth. Knowledge of the later is required to write Orthanc plugins. Therefore, the availability of an Orthanc Java API could prove to be a valuable asset in making Orthanc even more popular.

The purpose of jOrthanc is to allow Java developers to extend Orthanc's capabilities. By harnessing the power of frameworks such as Spring Security, teams with no knowledge of C++ will be able to adopt Orthanc in their projects.

This presentation aims to compare demo plugins written in both C++ and Java in regards to their total lines of code, testability and overall simplicity. It will follow that Java is a viable plugin technology just as C++ already is.

It will also be shown that the cost of developing Java-enabled Orthanc plugins can be considerably smaller than that of their C++ counterparts. This makes jOrthanc an interesting proposition to further extend Orthanc's adoption.

Participants will learn how to write an Orthanc C++ plugin. The goal is to fight against the FUD that arises when the topic of writing a C++ plugin arises. The workshop will be designed to be as easy to grasp as possible, even for people without any C++ experience.

The workshop will adapt to the systems of the participants to demonstrate it is easy to set up Orthanc on many types of systems, as well as to interconnect them. We will then 'normalize' the setups on containers and explore a couple of typical production configurations and scenarios. Finally, we'll get our hands off the keyboards and exchange tips and operational considerations when deploying in production for the time we have left (open discussion).