Thirona, taking AI from spin-off to 40 countries in 6 years
Thirona, a Dutch start-up company specialising in AI to analyse medical images, is offering one of their products for free to medical imaging specialists in an effort to combat the COVID-19 pandemic. International Hospital speaks to Dr Eva van Rikxoort, the Managing Director and founder of Thirona, about the company and its products.
International Hospital: Thirona is a Dutch company established in 2014. Can you give our readers a bit of background about the company?
Eva van Rikxoort: Thirona started in 2014 as a spin-off from the Radboud University in Nijmegen. We started with two full-time employees and we have built our company to 25 full-time employees and 20 part-time medical analysists. Together, we developed artificial intelligence software to analyse chest CT scans, chest X-ray images and retina images. Healthcare specialists around the world use our AI software for diseases like asthma, COPD, tuberculosis and diabetic retinopathy.
IH: What led you to set it up?
EVR: I was doing research on chest CT analysis at the Radboud University with my co-founder Prof. Bram van Ginneken. We saw that there was a gap between research that was being done on potential clinical solutions and putting those into clinical practice; Thirona was founded with the vision to bridge that gap.
IH: What products did you have at that time?
EVR: We started with two launching customers for one solution – our AI solution for chest CT analysis.
IH: Where does the name Thirona come from?
EVR: The name Thirona comes from the Celtic goddess named Thirona, worshipped for healing. Although more commonly spelled as ‘Sirona’ in the Latin alphabet, the spelling Thirona was chosen to reflect our roots in thoracic image analysis.
IH: Delft Imaging appears to be part of Thirona. Can you tell us a bit more about Delft Imaging and the relationship between Thirona and Delft Imaging?
EVR: Technically, Thirona and Delft Imaging are separate companies, although we collaborate extensively and practically work together as sister companies. Where Thirona specialises in artificial intelligence software for medical imaging, Delft Imaging specialises in diagnostic innovations that can be used in developing countries. For example, our AI solution for chest X-ray analysis (CAD4TB) is developed by Thirona and distributed by Delft Imaging.
IH: Can you tell us briefly about your key products and where they are being used?
EVR: We essentially offer three categories of products. AI software for chest CT analysis focusing on COPD and asthma, called LungQ; AI software for X-ray analysis – focusing on tuberculosis, called CAD4TB; and AI software for the analysis of retinal images – focusing on DR, AMD and Glaucoma, called RetCAD. LungQ allows for the quantification of chest CTs and is used for patient monitoring, treatment planning and clinical trial analysis.It is mostly used across the United States and European Union. CAD4TB is used in 40 countries around the world and has screened more than 6 million people for tuberculosis. RetCAD is being rolled out across Europe and Asia.
IH: If we look at how the company has grown over the past 5-6 years -- can you explain what has been the driving force behind the growth?
EVR: As a spin-off of the Radboud University, our company (and our solutions) is rooted in science. Every software product we develop is thoroughly validated (through 150+ publications to date) and that level of validation drives our growth, I believe; our customers know the type of quality we aim to offer.
IH: What guided your research and development? In other words, why did you take the product development route you have taken?
EVR: We are a demand-driven organisation. Meaning, we develop and work on where our customers and partners have needs to be supported. That has led us on the route we have been on for the past sevearl years.
IH: What are the main challenges have you faced and how have you overcome them?
EVR: One of the main challenges was keeping the same culture in the company during the growth from a small team of a few people with similar backgrounds to a larger more diverse team. We did this by installing a management team structure. By making it a focus for each team allowed us to overcome any difficulties.
IH: Most recently, you are offering a free AI-powered COVID-19 tool – the CAD4COVID-Xray. I understand it has been developed on the back of your successful TB-screening AI tool, CAD4TB. Can you explain how the CAD4TB tool works and how you adapted it for COVID-19 screening.
EVR: Yes, we developed two AI-powered COVID-19 tools actually: one for chest X-ray analysis and one for chest CT. Both have been built on the technical foundation of our existing and proven CAD4TB and LungQ solutions. Because there were underlying algorithms already in place, we were able to rapidly pivot these for the detection of COVID-19.
IH: How will this tool help healthcare facilities and COVID-19 patients?
EVR: CAD4COVID-XRay and CAD4COVID-CT automatically detect COVID-19 related abnormalities and thereby help with triage before any follow-up testing, like RT-PCR. This helps to reduce the workload of healthcare personnel and alleviate the burden on RT-PCR tests. Furthermore, both solutions show the percentage of affected lung tissue, thereby helping to track disease progress and recovery.
IH: How has the AI tool been validated? Has it been approved for use in Europe? In which other countries / regions has it been approved?
EVR: We have done several studies for both solutions (a publication on CAD4COVID-XRay was recently published in Radiology), through which we were able to prove that the software performs on par with expert human readers. For both solutions we have applied for class IIa CE certification, which we expect to receive soon.
IH: Are the CAD4COVID tools specific to certain platforms?
EVR: Both solutions are system agnostic. They process DICOM images which can be from any type of system.
IH: Are you receiving many requests for the CAD4COVID-Xray AI tool?
EVR: The response has been tremendous, which is probably also because we made the software available free-of-charge. CAD4COVIDXRay was launched first (March 31st) and has since been made available to 30+ healthcare facilities across more than 20 countries. We are especially focusing on resource-constrained settings (mostly in developing countries) because in those settings CT often has limited availability, making X-ray all the more important. CAD4COVID-CT was launched a month later and is available at 15+ facilities across 10 countries. We are also integrating it in several platforms through collaborations with partner companies like Smart Reporting.
IH: Why are you offering it for free?
EVR: We knew that for CAD4COVID to have the biggest impact during the pandemic and provide the most support to healthcare specialists globally, we needed to roll it out rapidly. In order to do that, we wanted to avoid as many hurdles as possible that could cause a delay in facilities being able to use the software. We believe cost is a big factor in that. That’s why we, supported by several organisations, have made it available to use free-of-charge.
IH: Can it be shared easily with healthcare facilities that want to use it?
EVR: Yes, people can fill in a form on www.delft.care/cad4covid (for CAD4COVID-XRay) and www.thirona.eu/cad4covid (for CAD4COVID-CT) and our team will reach out to them to help them set it up for their facility.
IH: Is any training required to use the CAD4COVID tools?
EVR: The tool is designed to be very intuitive, but we have developed an onboarding tool to guide new users on how to use the software effectively.
IH: Lastly, what’s in the pipeline for Thirona and how do you envisage development of the field of AI in medical imaging?
EVR: We are quickly expanding into other areas like cystic fibrosis on chest CT analysis, silicosis on chest X-ray analysis and cataract on retinal images. I believe that AI still has to prove itself in many settings, for many different use cases, but that it will become more and more accepted over time, and we already see this happening at incredible speed. In time, AI will leverage the efforts of our healthcare specialists, helping them to diagnose and determine treatment planning quicker, more effectively and reduce their workload in the meantime.