By Miriam Mirza
Artificial intelligence (AI) was the dominant conversation at ECR2024, with numerous sessions focusing on its impact on current and future radiology practice.
Speakers pondered whether AI could change the game in cancer screening and detection in a dedicated session chaired by Prof. Luís Martí Bonmatí, Director of the Medical Imaging Department and Chairman of Radiology at La Fe University and Polytechnic Hospital in Valencia, Spain.
Prof. Luis Marti Bonmatí
In his introduction, Martí Bonmatí emphasized that the aim of screening is to identify subclinical diseases for more effective treatment of small lesions. ‘The goal of screening is to detect,’ he said unequivocally.
He also highlighted the challenges faced in implementing population-based cancer screening programs, particularly due to the growing shortage of radiologists. Nonetheless, AI presents with new opportunities for more precise risk prediction, improved cancer detection and segmentation, and optimization of workflow and resource allocation.
‘AI is an effective tool for enhancing accuracy and efficiency in cancer screening while simultaneously relieving medical staff,’ he said.
A Solution for the Skills Shortage?
The challenges and potentials of AI in breast cancer screening were assessed by Prof. Sarah Vinnicombe, Consultant Radiologist at Cheltenham General Hospital, UK, in the following presentation.
Building on Martí Bonmatí’s reference to the radiologist shortage, she underscored the urgency of leveraging AI to address the impending crisis in radiology, especially given the forecasted decline in specialist radiologists. ‘The number of breast radiologists at consultant level is set to decrease by 40 percent,’ she said.
Prof. Sarah Vinnicombe
With the ongoing boom in AI technology, an increasing number of imaging AI products are entering the market. However, some of these products still have to show their efficiency, she nuanced.
‘In 2021, there were 200 AI products primarily used in imaging, but only 50 percent had independent evidence of their effectiveness,’ she said. ‘Of 100 CE-marked products, only 36 had peer-reviewed evidence of their efficacy.’
She was adamant that radiology should focus on using AI to enhance early detection and screening, but also emphasized the need to carefully assess and validate the quality and effectiveness of these technologies.
Potential for Cost Reduction
Prof. Olivier Rouvière from Hopital Edouard Herriot in Lyon, France, echoed her comments in his talk on AI and prostate cancer detection, in which he discussed various AI approaches for detecting prostate cancer, particularly in conjunction with MRI scans.
His presentation covered both deep learning methods and simpler quantitative techniques, as well as the strengths and weaknesses of current AI systems, notably their high sensitivity and lower specificity.
Prof. Olivier Rouvière
Rouvière stressed that AI algorithms should complement human diagnoses and emphasized the necessity of external validation to assess the robustness and suitability of these algorithms for widespread clinical use.
Bram van Ginneken, Professor of Medical Image Analysis at Radboud University Medical Centre in the Netherlands, highlighted another potential advantage of AI in the following talk, in which he reviewed the current status and requirements of early lung cancer detection.
Van Ginneken referred to studies primarily focused on reducing false positives, and explained how AI would help bring costs down in healthcare.
The integration of AI in radiology demonstrates considerable progress and potential, and the scarcity of skilled labor and the importance of precise diagnostics highlight the need for such innovative solutions, the panel observed.
Experts underlined the necessity for external validation and quality assurance of AI technologies to ensure their efficacy and reliability. When utilized correctly, AI not only promises to increase efficiency in cancer screening but also contribute to costs reduction in healthcare, they concluded.
