The SIIM 2019 convention took place at the Gaylord Rockies Resort and Convention Center in Aurora (near Denver), Colorado. It’s a brand new large-size hotel with the allures of a Rocky-Mountain ski-resort, with a (by European standards) huge convention center attached to it. From the imposing entrance hall, a spectacular view of the snowy mountains was offered through a gigantic window, with in the foreground the vague outline of the Denver skyline.
In between the lectures at the SIIM meeting I was able to visit the Technical Exhibit several times a day. The SIIM is the American counterpart of the society of which I’m currently the president, EuSoMII. In the exhibitors’ hall a mix was presented of booths from the classical vendors (Agfa, GE, Carestream etc) and startups. The latter had been assigned their own “Startup Lane”. At the end of the hall a theatre had been constructed with a stage. It was surrounded by transparent plastic walls, probably with the idea of protecting it from the sound coming out of the exhibitor’s hall , which didn’t turn out to be very successful to be honest. In this theater, several interactive sessions were organized on topics related to Economics of AI, the shift to Enterprise PACS, reading from home, AI in practice, etc. The program is still available here, and the Twitter channel gives a good overview of the presentations that were given by each company. On the other side of the end of the hall scientific posters were presented, and in between both several smaller stands were positioned, representing the projects participating in the Innovation Challenge. All these innovative projects could compete during a pitch-session on Thursday afternoon, for a jury representing several hotshots in imaging informatics, such as Elliot Siegel and George Shih. The participants were subjected to the questions of the jury, which resulted in an exciting pitch-session. Some participants showed that they were really well-prepared, this was also decisive in determining the winner.
Although I visited many booths including those of the major vendors, I would like to keep this overview concise and limit myself in providing some thoughts about those companies and products that struck me the most, and which I think were the most innovative and usable from my perspective as radiologist with a background in imaging informatics and AI. To be more concrete: those firms whose brochures I took with me and those booths whose pictures I took. Here we go.
The ScriptSender services focus on the automated integration of orders coming from referring physicians, by enabling direct communication between systems (RIS, EMR/EHR) in different practices or hospitals, using industry-standard encryption technologies. This way all kinds of documents (referrals, clinical notes etc) can be safely transmitted and integrated in each system. In addition to this they also provide the possibility to share images through their IVIS (Image View Image Share) module, facilitating instant clinical review on any device. Knowing that at this moment (2019) in the Netherlands many transmural orders are still sent by fax (“What the hell is that?” my children ask me when telling this), this automated order-integration tool seems to be the missing link for many hospitals and radiological practices.
Nowadays many primary care providers often make pictures of a patient’s symptoms (e.g. rash, skin abnormality) for documentation of diseases, by using the camera in their smartphone, where also the data is left — which is not compliant to GDPR nor HIPAA regulations. Safe integration of those images/videos in the PACS/VNA system is usually not possible, unless they use the Mobile ImageMover. With this tool providers can securely capture pictures and video’s, annotate them and automatically import them in the EHR by using a QR-code scan. Also other unscheduled image captures (from the emergency doctor, dermatologist and other specialists) can use the ImageMoverMD solution to import these files directly into the EHR. In addition — and maybe most important — the Uploader solution make it easier to upload CD’s and DVD’s with images from outside patients, import them in the central archive system and link them to the EHR, a procedure that usually costs a lot (personnel) and involves a rather complicated workflow. As the icing on the cake, the ImageMoverMD also facilitates the exchange of images between health systems, which can significantly avoid repeat imaging. Of course there are also other companies providing this kind of image exchange, but their complete package of services facilitating the sharing and seamless and well-structured integration of images in existing digital patient records is very appealing I think. The tool allowing to annotate these pictures of course will also be very useful in developing additional AI-based algorithms.
In this era of AI almost everybody knows how important and relevant the availability of curated data is. Many AI- developers are looking for secure access to curated medical images. OneMedNet actually is a clinical data broker service allowing researchers to purchase de-identified clinical data stored across a network of participating healthcare institutions. All clinical data are indexed and remain on-premises. A web-based search tool allows the customer to find the most relevant collection of data. From the information displayed on the website (on July 7th, 2019), the company already has 46.6 million labeled images available, equal to 467,418 radiology exams from 117,825 patients. The interesting thing is that you can ask them to calculate how much your imaging archive is worth now. This type of services undoubtedly will become more visible in the next few years, as part of a new business-type of ecosystem developing around the availability and sharing of medical images, with the intention to facilitate the development of new AI-tools.
A similar new startup is Segmed.ai, which is supported by the Stanford School of Business. The company collaborates with an international group of hospitals. It has established collaboration with a global team of radiologists, residents, technicians and medical students to label images, allowing the AI-developers to focus on building their algorithms. On the European side of the ocean data broker type of platform called Medexprim, which is based in France but also offering its services in the US. The representatives of both companies were also at the SIIM meeting but did not have their booth at the exhibition hall.
An interesting South-Korean AI-tool with FDA-approval is offered by this company, and is named ClariCT.AI. It’s an ultra-low-dose CT denoising solution based upon Deep Learning Technology pre-trained with millions of images, allowing to make Chest CT examinations at very low doses (1/20 of regular dose) which provide images similar to scans made with regular doses. The software seems to be compatible with any CT scanner and does real-time processing. Of course this tool allowing to reduce the radiation exposure by 70–95% is a very interesting tool in the context of lung-cancer screening CT scans, although this type of service is not yet available in most countries.
CureMetrix is a UK-based company providing an FDA-cleared and AI-based triage software for mammography called cmTriage. It enables radiologists to prioritize and sort their mammography work list based on cases that might need immediate attention. In other words the software identifies which cases are suspicious and allows radiologists to focus on those first. Interestingly the software only flags 8.2% of the cases, whereas according to data from the BCSC radiologists — at the same sensitivity — recall 9.6% of the cases. So not too many false positive alerts, which of course is essential in this kind of applications. The nice thing about this tool is that it’s really focusing on the implementation of A.I. for improving the radiological workflow in oncological imaging instead of emergency-type of imaging, which — to be honest — is still a quite rare nowadays.
For me being a Belgian citizen it was nice to find a startup booth from my own country. It certainly wasn’t the only company with Belgian roots — also Agfa and Barco belong to this group — but I feel reassured when I can find signs of Belgian entrepreneurship in the AI arena. Nevertheless in this exhibit I was missing promising companies like Icometrix and Robovision, which are both active in Europe and in the US. Osimis is a spin-off of the University of Liège and the whole product line is based upon the Orthanc, which is an open-source server developed by Sébastien Jodogne. Orthanc has a very user-friendly Web interface (API), enabling you to transfer DICOM images in a secure and very convenient way. It is also ranked among the best and most frequently used open source servers for many healthcare facilities, hospitals and research centers around the world. Recently it was also integrated in the Google Cloud Platform (GCP) for facilitating the handling of DICOM images, which of course is a ground-breaking achievement. The Lify solution on the other hand offers a much wider range of services, mainly facilitating the sharing of medical image data for other types of services such as distant reading, patient access to images and integration of AI-tools with the existing PACS. Actually Lify can be seen as a Orthanc platform offering add-on features to those using Orthanc in a hospital environment. The concept certainly looks very promising, but many features are still in the development phase. Nevertheless, there still seems to be room for further growth since the company recently raised its capital with 2,3 million EUR.
Veolity is a reading software developed by MeVis (Germany) for reading Chest CT studies in lung cancer screening programs, as is the case in the US market. Other countries are likely to follow however. This also means dat large amounts of imaging data will have to be read in a standardized and efficient way, to ensure high diagnostic accuracy. The great thing about Veolity is that it improves the workflow by automatic registration and integration of previous studies, and by creating reports based on the Lung-RADS criteria. The software automatically marks regions that are suggestive of solid pulmonary nodules (CAD function). The software allows to read the Chest CT studies more efficiently, and thus is also streamlined for high-throughput scenarios in which the radiologist can concentrate on reading cases as other data processing steps are automated. I’m much in favor if all AI-products that enable radiologists to improve their workflow and increase the accuracy and value of their readings.
Apparently radiologists are doing more and more remote reading, away from the hospital. Also in my own professional environment I also hear more and more colleagues talking about the possibility of setting up a workstation at home. This is probably driven by number of factors, such as the millennials entering the workforce that have a different perspective on work life balance. From a technology perspective, recent advancements have allowed home reading to become more common. Laptops are becoming more powerful and compact and secure remote access is more easily achievable. Another game changer is the Thunderbolt 3 cable, which enables small portable workstations to drive high resolution displays with the graphics performance needed for medical imaging applications. The new Barco solution consists of 3 main components, being a mobile workstation laptop from Lenovo, a diagnostic display system and a Thunderbolt external eGFX graphics box from VisionTek (only available in the US but for the EU market another box is currently being searched after, as was explained at the booth). A major advantage is that this workstation can be maintained by the IT-department from distance, similar to a workstation within premises, thus taking away the hassle of having to visit the doctor’s home or having to transport a heavy desktop. The laptop only needs to be connected to the graphics box, which makes it a very practical tool since the laptop can even be used by all staff members by passing it on to the person on duty. Barco provides all necessary features and workflow tools for optimal use of its 2K displays at home, and the company also guarantees continuous quality monitoring of the screens and workstation. Great practical solution from Barco, which will certainly find its users!
The SIIM 2019 Technical Exhibit showed a lot of innovative new tools, confirming my opinion that it’s always worth visiting such exhibits. I tried to give an overview of those applications and tools that I think would fit well in radiological practices, or that hold a lot of promise and have the potential to add value to radiological services. In this blog I did not yet fully address the Innovation Challenge, but I in my next story I will go further into depth in the projects that were presented there.
Please do not hesitate to send me your questions, comments and thoughts about this report — an applause is always welcome of course.