IDEAL: Artificial Intelligence and Big Data for Early Lung Cancer Diagnosis Prospective Study (Phase 2)

As the commonest cancer with 1.8 million cases diagnosed each year worldwide, early diagnosis of lung cancer is important to reduce mortality. The early diagnosis of lung cancer is contingent on both the detection of a small lung nodule and determining whether it is malignant. Whilst computed tomography (CT) has proven to be a robust way of detecting lung nodules, they are often discovered on routine scanning as an incidental finding or as part of a lung cancer screening program. Hence, determining whether they are benign or malignant is challenging.

Up to 75% of smokers scanned either as part of their clinical care or in lung cancer screening trials have subcentimetre pulmonary nodules detected. This places a substantial burden on scanning facilities, staff and patients. Current methods of determining if lung nodules are benign or malignant are not standardised and unproven. The US National Lung Screening Trial (NLST) showed that up to 95% of lung nodules detected on CT scans of the chest were false positives i.e. they were not malignant. Detection of non-malignant nodules have the unwanted consequences of unnecessary cost as they require follow-up scanning or alternative methods of investigation, cause patient anxiety, may result in increased morbidity potentially by biopsy or resection, and result in increased patient radiation exposure due to follow-up CT scans or from PET-CT scans.

As part of standard care at present, patients with lung nodules greater than 4mm and sub-centimetre are followed up with CT scan(s), up to 5 scans, for up to 24 months, according to the internationally accepted Fleischner guidelines. Additional investigations, such as a positron-emission tomography (PET-CT) scan and biopsy or resection may also be performed based on the size and clinical risk profile of the patient.

Recent studies have shown that incorporating lung nodule characteristics such as size, texture, growth rate, contrast enhancement can improve the accuracy of predicting the risk of malignancy. This allows the stratification of lung nodules into different investigations and/or follow-up pathways based on the predicted risk of malignancy.

This study aims to test the use of novel CT image analysis techniques incorporated into a clinical risk model to characterise small pulmonary nodules. The study will incorporate solid and predominantly solid nodules of 5-15mm scanned using a variety of scanner types, imaging protocols and patient populations.