Parc Taulí kicked off this July a multisite clinical trial to validate artificial intelligence software that detects lung cancer in unreported chest X-rays. This tool, dubbed Optimal XR, is designed to prevent potentially malignant lung nodules, which are not currently diagnosed in time because the X-rays have not been examined by a radiologist, from going undetected.
The innovation is part of the Optimal Lung project, a clinical decision support system tapping artificial intelligence for detecting pulmonary nodules run by Parc Taulí University Hospital and Vall d’Hebron University Hospital in partnership with the Eurecat technology centre.
The study, led by pulmonologist and researcher emeritus at the Parc Taulí Research and Innovation Institute (I3PT) Eduard Monsó, seeks to confirm the effectiveness and safety of the algorithm while also validating its use in a real-world hospital setting and generating the evidence needed to apply for clearance to market it.
The clinical trial will last for sixteen months, ten for recruitment and six for clinical follow-up, with additional passive follow-up for three years after its completion. In the course of the study, nearly 3,000 chest X-rays will be analysed in real time from the database of the Parc Taulí’s Centre for Digital Medical Imaging (CIMD) which are not necessarily related to respiratory diseases but rather may have been taken on other medical grounds. “The study will consist of using the algorithm to identify pulmonary nodules in these X-rays, although not all will have them,” says Monsó. “Nodules detected by the system will be reviewed by a radiologist and, if confirmed, will be referred to the Pneumology Service for further evaluation and follow-up.”
Optimal XR and lung cancer
Lung cancer is the leading cause of cancer death worldwide. In Catalonia alone, in 2023, almost 5,000 new cases and over 3,400 deaths were registered as a result of this disease. However, the late onset of symptoms means that 80% of lung cancers go undetected and are diagnosed in too advanced stages when the five-year survival rate is already less than 20%. Early detection not only significantly increases the survival rate by up to 60% but also enhances the living conditions of people who have the disease.
At present, early detection of lung cancer is a major challenge for primary care in which thousands of chest X-rays are performed every year on a significant proportion of patients as part of standard screening or as diagnostic procedures that may not necessarily be related to suspected lung cancer at the outset. However, the growing shortage of specialised radiologists means that it is not possible to examine these X-rays as efficiently as might be desired, which carries the potential risk of failing to identify cancerous lung nodules in early stages that are subsequently detected when the disease is already in advanced stages.
“We spotted the need to develop and implement a solution that could accurately and massively analyse these chest X-rays from primary care, thereby detecting high-risk cases and steering them into the radiology workflow to significantly improve early identification and intervention,” notes Monsó.
Against this background, Parc Taulí University Hospital and Vall d’Hebron University Hospital in partnership with the Eurecat technology centre have developed Optimal Lung, a clinical decision support system tapping artificial intelligence which includes two algorithms for detecting pulmonary nodules: Optimal XR, which targets X-rays coming from primary care and emergency departments to avoid missing cancerous lung nodules, and Optimal CT, which focuses on CT scans to cover all levels of lung cancer diagnosis.
Optimal XR is artificial intelligence software that leverages deep learning technology to analyse X-rays. The algorithm processes all X-rays taken at the hospital and identifies the ones with a high likelihood of containing pulmonary nodules based on more than a thousand real X-rays that it has previously trained on. Detected X-rays are sent directly to a radiologist for review and confirmation of the presence of nodules. This simplifies the work of radiologists and improves the early detection of potential lung problems that may have been overlooked until now.
Departments from various clinical specialities including radiology, respiratory medicine and oncology together with engineers, information systems experts and innovation staff have been involved in the development of this solution. The ultimate goal will be to roll out Optimal XR in other hospitals, especially in primary care settings where it is now most needed. Monsó points out that “the team is hoping to bring it to other countries where chest X-ray reading and care is not as advanced as it is here.”
