Top 10 Lessons Learned when deploying radiology and open source PACS Solutions in LMIC Regions

It is sometimes hard to realize that as we have timely and easy access to radiology imaging, two thirds of the world lacks this capability according to the WHO. The good news is that “basic” X-ray combined with a handheld Ultrasound is becoming very affordable which will provide access to the most basic diagnostic tools and can make a major impact on patient health outcomes. Regarding X-ray, digital plate technology has come down in cost and a robust and simple stationary X-ray unit or portable unit can be acquired for about US$50k. These units are equipped with a battery in case the power is unreliable, as is often the case in developing countries. The Butterfly handheld Ultrasound and similar units by major manufacturers such as the Philips Lumify which can be connected to a tablet or a phone has been revolutionary as it can be deployed inexpensively for less than $5k. There is additional cost for training and support but it is nothing like the $100k+ units that were the only options till recently. A clinic or small hospital can be equipped with radiology imaging for much less than $100k on hardware.

What about image access? Commercial PACS systems that connect to these digital modalities to archive, manage the image workflow, and display these images have become mature and rather sophisticated. However, the PACS software with its sophistication has become expensive and is not affordable for most hospitals in LMIC countries. A solution could be to get a PACS company to donate licenses of their software and support for free or at a large discount, but this business model is not scalable, especially if you consider the huge amount of hospitals that are in need of this solution.

The alternative solution is using open source PACS software. There are a couple of factors that make the deployment of open source PACS systems feasible. First, open source PACS software has been around for 20+ years, the most popular software was e-film, followed by ClearCanvas. Last time I talked with the ClearCanvas folks they told me they had seen 50,000+ downloads of their software. I can attest to its popularity because when I asked the students at a PACS training organized by RAD-Aid in Tanzania a few years back what PACS system they were using, 90% answered ClearCanvas (CC). CC is not a bad solution, as a matter of fact, as part of a grant by Rotary International, I installed the CC PACS archive and viewer software in a small children’s hospital in Tanzania a few years back, facilitating a fixed X-ray table with CR, Ultrasound and a dental system. This which worked very well and was the only feasible option at that time.

As of today there are new generation PACS open source providers that provide web-based access to the images using DICOMWeb by open source web-enabled viewers which opens up access to anyone in the hospital with a browser, including mobile devices. For relatively small institutions, Orthanc PACS open-source software is a good choice. Remember that most smaller hospitals only have one or more X-ray units and a few Ultrasounds which can be easily managed by this software. For larger institutions that might even have a CT, the open source DCM4CHEE PACS software is a good solution. Open-source image viewers such as the Stone Web viewer, Weasis, or OHIF can be used for image display. If more sophisticated image processing such as 3-D is desired in case the hospital has a CT, one can get an iOS platform and use Osirix for that.

An example of an open source deployment is shown in the figure below representing a recent installation using a grant from Rotary International. ?The configuration consists of a donated fixed table with CR, a newly purchased portable X-ray with DR plate that can be used either in the department, clinic or at the ICU bedside, and an existing 10 year old Ultrasound. We experienced a very common problem with Ultrasound as it did not have the DICOM option enabled preventing images to be sent to the PACS. Two image servers were installed which were basically high-end desktop computers with a 5TB disk for local storage, and a script was created that would keep the servers in sync providing redundancy. A NAS is connected to one of the servers for image backup. Each server has the Orthanc PACS software installed and a viewer using 2 medical grade monitors donated by a US manufacturer (Double Black Imaging). A donated software license from Idonia provides remote access as it allows the upload of certain studies in the cloud for consultation. The web-based viewer can be launched from the in-house developed EMR with a simple http call or physicians can access the images directly using the web interface.

Here are my top ten lessons learned so far from the first installs and deployment of radiology imaging:

1.????? Open source PACS software is scalable and sustainable. When we needed assistance, I was able to recruit a volunteer to help with the installation and troubleshooting. I learned over the past several years that support by local dealers is very poor due to insufficient training and support from their HQ and found much better and timely support from the many open source afficionados available globally. Of course, this paradigm is unproven till many more similar installs are deployed, but assuming that open-source forums and resources will become more complete and helpful this seems to be a very viable option.

2.????? It is very hard to “wean” hospitals off from film and/or hardcopy. In the above installation a hardcopy printer is connected to the CR system, but when the local disk got full, older studies were deleted. The software on the CR was not able to receive images that were pushed back from the PACS. A hardcopy could not be printed for returning patients that came back after let’s say 6 months. In this case, a physician would have access to the prior study on a monitor but that required a major mindset change.

3.????? It is advisable to use local suppliers for commodity hardware as much as possible. Our medical grade monitors required remote support, which we received from the manufacturer, but it would have been easier if these were sourced from a local dealer. Just make sure you use medical grade monitors for the diagnosis.

4.????? The open source PACS and viewers provide “bare bones” capabilities which means that expectations should be set accordingly. This is especially important when being used by visiting non-local physicians, which often the case. These foreign radiologists provide a critical resource for training the local users, remember that most of these hospitals rely on a Technologist and/or non-radiology physician for the image interpretation due to the lack of locally trained radiologists. However, these volunteers are used to a full blown PASC with sophisticated worklist and reporting capabilities and therefore will need to get used to the limited capabilities.

5.????? Unique patient Identification is a major challenge. Most hospitals in LMIC countries don’t have an EMR that they can rely on to issue a unique identifier, and they enter patient information manually on the imaging modality. In addition, birthdates are often approximated as exact dates are not reported and known by the patients. Non-uniqueness of patient information makes retrieving prior studies hard.

6.????? Diagnostic imaging reporting is very primitive. Most institutions I have seen create a handwritten note that is sent back to the ordering physician with one or two lines containing the finding. The next step in sophistication uses Word templates and cut/paste capabilities to create the findings. Speech recognition that has access to medical libraries is unaffordable and not supported in LMOC countries. The use of ChatCPT has major potential as soon as it becomes more mature and can facilitate medical terminology.

7.????? AI for image detection can make a major impact. Right now, AI applications are evaluated based on their clinical efficacy, in particular if it can improve a diagnosis of a well trained and experienced radiologist. However, in LMIC countries the images are rarely interpreted by a trained radiologist as they are often not available, therefore the potential improvement of the diagnosis and detection by an AI algorithm over a non-imaging trained physician and/or technologist is huge.

8.????? Low earth Orbit Satellites (LEOS) will make a major impact in both clinical and technical support. There are several LMIC countries that do not allow this technology to be deployed (yet), for example, South Africa, Zimbabwe, Tanzania and several others, but I assume that will be a matter of time. Poor internet connectivity, especially in remote areas have been a big barrier till now. One of the hospitals in Tanzania I talked with at the RAD-aid conference was unable to use the remote clinical support capability on their donated Ultrasound units due to bandwidth issues, which could have been solved by using Starlink, the most popular LEOS solution

9.????? A cookie-cutter approach is highly desired for the PACS deployment, i.e. using standard configurations. It is recommended to avoid customizations and local changes, even although the fact that open source provides this capability. A good analogy is the airline industry, with Southwest airlines using only one type of plane (being 737) making training and support much simpler than for example American Airlines who has a mix of various Airbus and Boeing types of planes. Three configurations, i.e. one for small, one for midsized and one for larger image institutions should be sufficient. The smaller and mid size could use Orthanc and the larger one using DCM4CHEE.

10.? Keep your deployments low-tech. There is a tendency for LMIC countries to want the latest and greatest technology, for example a sophisticated mammography 3-D tomo-synthesis X-ray unit (aka DBT). One of the country managers at the RAD-aid conference reported that their one and only DBT unit in the whole country was down waiting for parts. For the money of that one unit one could have purchased a few dozen Point of care Ultrasounds for breast screening. Recent reports from Africa and China using the Ultrasound as a primary breast screening device are very encouraging.

In conclusion, I firmly believe that open source PACS software is the only scalable solution for the two thirds of the world that does not have access to radiology imaging. The modalities themselves have become so affordable that a large-scale deployment is feasible, especially when combined with a “cookie-cutter” approach to the PACS software configuration. There is still a lot to be learned when deploying these systems, and I am sure some of it will be “two steps forward and one step back” due to many non-technical issues such as infrastructure, regulatory and people issues but there is a huge potential to make a big difference.

Herman Oosterwijk