Harnessing the power of AI in the pandemic
Industry experts discuss the role of artificial intelligence in the fight against COVID-19, the lessons learned and its future promise in healthcare
26 Aug 2021
While artificial intelligence (AI) is hardly a new concept, it is still a nascent technology looking for new applications. One area in which it is showing great promise is the healthcare sector, particularly in the field of molecular diagnostics. And it is here that the global COVID-19 pandemic has presented the ideal opportunity to highlight what AI can offer, exemplified by its use in the rapid and reliable interpretation of data from hundreds of thousands of molecular diagnostic assays and images of damaged lung tissue.
There are challenges, however, not least in the trust of AI interpretation, and technology companies across the globe are linking up with assay and imaging kit providers to step into this brave new world. To find out more, we speak with James Grayson, lead applications and technology specialist at UgenTec, about how he has been working with LGC during the pandemic to harness LGC’s existing high-throughput platforms in clinical diagnostic settings. We also hear from Dirk Smeets, chief technology officer at Icometrix, about how the application of AI to medical imaging is helping to tackle the ongoing pandemic, and how he sees it impacting healthcare in the future.
The benefits of incorporating AI in healthcare
The AI technology Grayson uses at UgenTec is called FastFinder — a hosted, cloud-based molecular diagnostics interpretation platform which is ‘instrument-agnostic’. In other words, it can take the raw data from almost any real-time PCR cycler on the market and evaluate curve morphology. It has come into its own during the pandemic where the unprecedented numbers of COVID-19 samples and differences between diagnostic assays have presented unique challenges. “You have all these different assays from all these different providers,” Grayson explains. “Our analysis module is designed to reduce the time-to-result for a patient sample and to standardize and automate that process.”
Grayson also alludes to the elimination of human error. “At peak, we were reviewing data from the UK at about 120,000 samples per day. The risk of human error is high – you have too few people looking at too much data that needs to be moved through the process at a certain rate to meet real demands, often political,” he observes. “When you start applying AI, you have a way of standardizing and automating to make those results more reliable and at least consistent.”
A further benefit of applying AI to molecular diagnostics has been to alleviate the potentially unbearable demand for resources placed upon individual countries or institutions that have fixed staff, equipment, and ‘bandwidth’. “You now have the ability to help these organizations increase throughput whilst maintaining quality,” Grayson points out.
Challenges to AI acceptance
The key challenge to an AI technology provider such as UgenTec is getting to the point of trust in the interpretation that the FastFinder platform generates. When it comes to assessing PCR curves from COVID samples, for example, clear positives and flat-line negatives are no problem to a human (or an AI). It is the seemingly anomalous curves that present the greatest challenge, such as those generated by a very low concentration of virus in a sample.
As Grayson questions rhetorically, “Is that an artefact of an instrument, or human error, or a true positive? That’s truly where AI comes in and can correctly identify or predict that the sample is positive,” he says. “It’s trying to find a way to address the needs of throughput, of automation, of managing resources to get an outcome that the user is willing to accept as true.” This sometimes includes the use of AI as a filter, in which any anomalous curves are drawn to the attention of the clinician for a final human approval of outcome.
The interface between AI and LGC technology
At UgenTec, Grayson has been working to support LGC’s existing high-throughput platforms to bring them to a level where they can be used in a clinical diagnostic environment. “You have all these amazing instruments that were designed to work at ultra-high-throughput settings producing incredible amounts of results per day, but relatively cheap compared to other molecular diagnostic technologies on the market,” he enthuses.
Pairing UgenTec’s diagnostic software with Biosearch Technologies’ high-throughput extraction instrumentation has been particularly powerful in response to COVID-19 mass testing demand. “Because Biosearch Technologies’ instruments work with low sample volumes, it’s a very good way to support high-volume testing, during situations like a pandemic, cost-effectively,” Grayson notes.
A more personalized future
Grayson expresses a view of the future both specific to his company UgenTec and more generally for the healthcare sector: “To date, UgenTec has been focused on a limited number of molecular diagnostic modalities, and mainly PCR reactions. We’re starting to expand into sequencing assays, sequencing tests, and other modalities as well,” he says. “In a world increasingly awash with data and metadata, I essentially see AI facilitating better, more personalized medicine outcomes.”
AI lung image analysis in the fight against COVID-19
Dirk Smeets is the chief technology officer at Icometrix, a company that has been leveraging artificial intelligence to support radiologists with the assessment of lung CT images of COVID-19 patients throughout the pandemic.
AI is being used within Icometrix’s Icolung technology to interpret images of lung tissue damage and provide quantitative data to aid both the diagnosis and the severity determination of COVID-19. For this application, the benefits of AI are clear, asserts Smeets. “Lung damage is hard to assess visually by a radiologist,” he says. “With Icolung, this information is at their fingertips, and we have been able to help about 50,000 lung damage patients using AI assessment.”
Speaking more broadly about the advantages of AI, he continues: “It allows the latest research insights to be available faster than ever to patients in the clinical setting. As a result, we can expect AI to improve the quality of care at lower cost.”
“The future for AI looks bright,” Smeets concludes. “Many tasks in healthcare will be taken over by AI, improving the working conditions for clinicians and improving the lives of patients.”