Our team had an incredible three days at UKIO 2024, marked by some really engaging sessions and insightful discussions. One of the key highlights for us was the chance to talk about autonomous reporting. The session was chaired by Dr Sarim Ather, with presentations from David Lowe and our chief medical officer, Naglis Ramanauskas. Naglis highlighted that while autonomous AI is here, health providers need support for evaluating and implementing these solutions.
Globally, radiology imaging acquisition is outpacing radiologist reporting resulting in backlogs of unreported examinations. Our study found that up to 68% of practices had unreported radiology examinations. Reporting backlogs vary significantly across imaging modalities and were highest for chest radiographs. Almost half of the facilities surveyed report radiologist vacancies.
This is where ChestLink comes in. It can reduce radiologist workload with automation and is the first radiological class IIB AI device certified for standalone use (CE). It identifies chest X-rays with no abnormality and produces finalised patient reports without any intervention from the radiologist.
It works by analysing every chest X-ray study in real-time and producing a binary output (=healthy with high confidence; or cannot rule out pathologies with high confidence). If ChestLink identifies the study to be normal with high confidence, it generates a normal study report, which is intended to be used as a final report for the chest X-ray study. The radiologist does not need to review the study. The study is removed from the radiologist reporting workflow. The report is automatically submitted to the information system of the healthcare institution. If ChestLink cannot rule out the presence of pathologies with high confidence, it leaves the study for normal radiologist reporting in his usual reporting workflow. This includes both the studies where abnormalities are evident, or where the software cannot rule out the presence of minute abnormalities with absolute confidence.
Reducing workload
The NHS is currently facing an exceptional strain on its radiology services. As Dr Katharine Halliday, president of The Royal College of Radiologists, explained in her session at UKIO, there are huge workforce challenges in the sector at the moment and we need to address the radiologist shortage.
The most recent Royal College of Radiologists census showed there is a 30% deficit in clinical radiologists within the NHS. A staggering 97% of clinical directors are reporting that the shortage of radiologists is directly leading to bottlenecks and delays in their trust or health board, while 91% highlighted that these workforce deficiencies are impacting patient safety. In 2023, over four weeks of waiting for imaging test results post-scan were experienced by 745,000 patients in England.
ChestLink aims to reduce radiologist workload by autonomously reporting on cases with no abnormalities. The software is especially useful for primary care centres, where up to 80% of studies may feature no abnormalities. It is also useful for large scale screening projects (such as the global TB effort), where healthy study populations can be filtered out of human workload. The platform helps to shift limited healthcare resources on reporting on cases where pathologies are present.
Autonomous AI is like nuclear fusion – it’s always 10 years away
Creating AI for normals is not easy. In fact, it is very hard. ChestLink is the first step towards the autonomous future of medical imaging. The software can autonomously report on healthy patient chest X-ray studies without any involvement from a human radiologist, automating up to 40% of normal CXR studies, reducing the workload and strain for the radiologists.
ChestLink’s performance is measured by two key metrics for evaluating both the accuracy and efficiency of the solution. The first metric is the normal reporting fraction – this represents what fraction of all normal chest X-ray studies at the institution can be automated (reported on autonomously with high confidence). It estimates real-world productivity gains of ChestLink deployment. The second metric is accuracy – which represents the probability of a study being classified as normal by ChestLink, while containing detectable clinically significant radiological findings.
Every day, we analyse studies from various institutions around the world and pinpoint missed radiological findings that may indicate early signs of pneumonia, lung cancer, or inaccurately described device positions. We promptly return these findings for immediate review and correction when necessary.
The time for automation is here – there is a workforce crisis and we need to be able to find innovative ways to reduce delays and improve patient outcomes and safety, as these tasks take up valuable clinical and administrative time.
Contact the team to book a demo and see how our products work in action.