Lung cancer screening is proven to save lives through early detection. But, countries in Europe still differ on strategies surrounding its implementation.
A recent article on screening policies in ten European nations found that, while some have committed to nationwide programmes, most are investigating their viability or are in doubt. Yet there is clear common ground, too: they all face similar challenges.
At the Annual Meeting of the European Society of Thoracic Imaging (ESTI 2022), three renowned radiology experts discussed the current status of lung cancer screening and the use of AI, based on their experience in Belgium, France, and the UK. This article is a summary of their conversation.
Hopeful changes at EU level
Prof Dr Annemiek Snoeckx, Associate Professor at the University of Antwerp and Head of Radiology at Antwerp University Hospital, talked about the recent changes in the European Union policy on cancer care:
“As chest radiologists, we remember 2021 not only for a virus but also for the publication of Europe’s Beating Cancer Plan in February. Its focus is on four aspects of cancer: prevention, early detection, diagnosis, treatment and quality of life for lung cancer patients.”
Following intense lobbying from European societies and the chest community, the final publication included a mention of lung cancer screening.
Another milestone was reached in March 2022, when the Science Advice for Policy by European Academies (SAPEA) published its positive advice to the European Commission. The report states that:
“There is a strong scientific basis for introducing lung screening for current and ex-smokers using the latest technologies, such as low-dose CT scanning.”
In the coming months, the European Commission will pass this advice to the European Parliament, which will vote for or against implementing lung cancer screening. This vote is likely to be a game-changer on these initiatives.
The main challenges to implementation
The panel participants confirmed that challenges are similar across their countries. In the words of Dr Arjun Nair, Consultant Radiologist at University College London Hospital and Joint Clinical Lead for the National Targeted Lung Health Checks (TLHC) programme (presenting in a personal capacity, views his own):
“It’s reassuring to hear from both colleagues in Europe that they are having the same conversations. We may not have all the answers, but at least we agree on the problems, which is the first step.”
The first major challenge is the radiology workload. In the UK, the latest census from the Royal College of Radiologists showed a shortfall of 29%, with immense pressure on physicians. Reading and reporting CT scans is time-consuming. Within a workforce already under strain, introducing a new CT-screening programme seems like a tall order.
Secondly, radiologists are one piece of the puzzle. Lung screening requires input from several healthcare professionals involved in the end-to-end pathway. Prof Dr Marie-Pierre Revel, Head of Radiology at Cochin Hospital and Full Professor at the University of Paris, highlighted the likely need for additional training for surgeons who may not be sufficiently familiar with video-assisted thoracoscopic resections.
Another challenge is participant uptake. For instance, a recent update from the NHS showed that only a third (35%) of invited patients attended their consult. Dr Snoeckx believes that one of the reasons is that the high-risk population is hard to reach:
“The beliefs of current or former smokers on health differ from those of, for example, women who get screened for breast cancer. Every country needs to research how to address this population.”
Finally, screening requires countries’ long-term commitment to achieving the intended population health benefits. Organising such a programme is, in the end, also a matter of willpower.
Novel research into the role of AI
Artificial intelligence (AI) is well-suited for the precise detection of elusive lesions and automatic volume measurements. It can play a key role in lung cancer screening. The use of AI has the potential to increase the productivity of CT reporting, freeing time for clinical decision-making and complex cases. It may also boost the radiologist’s confidence in their assessment.
Before the large-scale implementation of lung cancer screening, there is a need to prospectively evaluate the role of AI. Should programmes be based on a double reading? Or a single reading aided by AI? The CASCADE study led by Prof Marie-Pierre Revel is set to find out:
“The main objective of the CASCADE study is to compare the performance of a single reader trained on lung cancer screening and using AI (Saige Lung formerly Veye Lung Nodules) with the standard of reference (double reading by expert thoracic radiologists) in a campaign of low-dose CT screening in high-risk women.”
CASCADE targets 2,400 asymptomatic women aged 50 to 74 with a smoking history. It will take place over two years across four regions in France: Paris, Rennes, Bethune, and Grenoble. Prof Revel explains the research background:
“The epidemiology of lung cancer in European women is worrisome. The French public hospitals 2020 report showed that the proportion of women among lung cancer patients increased from 16% in 2000 to 34.6% in 2020. The female gender is associated with a lower participation rate, whilst the benefit of screening is greater.”
The study has had an encouraging start:
“We’ve seen strong adherence so far. Over 1,000 women contacted the CASCADE team within the first month.”
The performance of the AI tool has also been satisfying. No false negatives were detected, and only one false positive, a rare case in which the first reader had also incorrectly flagged.
The lung health checks example
Dr Arjun Nair enriched the discussion with a real-world, practical example of a lung cancer screening programme: NHS England’s Targeted Lung Health Checks (TLHC). 600 people have already been diagnosed earlier through this service. The programme is hoping to become a nationwide initiative.
This is also the first large-scale lung screening project substantially supported by AI. Most reporting sites use our solutions, Saige Lung and Saige Lung Reporting, to find, manage, and report on pulmonary nodules.
Quality assurance in this context is not limited to the radiologist; it is a consideration for the entire lung cancer pathway. Dr Nair made three recommendations, based on lessons learned from leading the TLHCs:
“Standards are everything.”
Arjun Nair compared standards to a compass that everyone in the programme follows. Within the TLHCs, protocols are in place for both nodule and incidental findings follow-up. The training of radiologists is also standardised (through the regular BSTI workshops).
Yet, although essential, he emphasised, these standards are a ‘living document’ that should be subject to constant evaluation.
“Engage early and continuously.”
When setting up a screening programme, the team should align with all the software providers (Computer-Aided Diagnosis and AI) supporting the scan acquisition, analysis, and reporting. They should also engage with the radiology and project teams on the ground. They are the ones to hear the participants’ feedback and concerns first-hand.
“Automate as far as possible.”
Whilst the radiologist’s role remains unquantifiable, automated reporting contributes to an efficient programme. With the currently available technology, it is possible to automatically generate an AI-supported analysis for radiologists to use as a second or concurrent scan read. Adding these findings to a report can also be done in one click and in line with the programme’s protocols.
An invigorated commitment
Marie-Pierre Revel, Arjun Nair, and Annemiek Snoecks agreed that the next step in lung screening implementation is to run pilot programmes. These should gradually scale up to national levels. The discussion ended on a cautiously optimistic note, with a clear determination to overcome the challenges and improve lung cancer prognosis for patients.
The role of AI, in this context, will only become more important. Dr Snoeckx added when reflecting on the future:
“What if we had an AI system so good that it could report, with confidence, that particular scans do not contain any nodules? If radiologists solely focused on examinations with nodules, their workload would be drastically reduced.”