A paper published in Lung Cancer demonstrates the ability of Optellum’s Lung Cancer Prediction artificial intelligence (AI) to classify lung nodule malignancy.
As discussed in earlier news stories, there has been a growing interest in the use of computed tomography (CT) scans to screen patients for lung cancer. An EIT-Health funded project has seen Optellum collaborate with Oxford University Hospitals NHS Foundation Trust (OUH), one of the largest NHS teaching trusts in the UK; University Medical Centre Groningen (UMCG), core members of the NELSON European lung screening trial team; and the Thoraxklinik at the Heidelberg University Hospital, one of the centers which performs the most lung cancer surgeries in Europe.
The Early Lung Cancer Diagnosis Using Artificial Intelligence and Big Data (LUCINDA) project leverages the expertise of top radiologists to help develop and validate the Optellum Lung Cancer Prediction (LCP) AI as a clinical tool that has robust performance across various patient populations, clinical pathways and CT scanner types.
Patients were recruited from the three hospitals to form the LUCINDA dataset of 2106 unique nodules (205 malignant, 9.7%) from 1650 unique patients (201 lung cancer patients, 12.2%). Each reported lung nodule and lung cancer was located, contoured and labeled benign or malignant by clinicians, who had access to rigorously verified ground truth diagnosis (e.g. the results of histology for patients with cancerous nodules, and two-year stability on imaging follow-ups for benign nodules that have not been biopsied, as per the current clinical management guidelines).
Standard nodule guidelines for both incidentally detected and screening detected nodules recommend follow-up based on nodule type, size, and growth. Despite their widespread adoption, these nodule management protocols still result in a rather high false-positive rate, requiring healthy people to undergo unnecessary procedures such as biopsies, PET and CT scans.
The Optellum Lung Cancer Prediction AI generated a malignancy score for each nodule in the dataset demonstrating excellent performance on identification of benign nodules (sensitivity 99%) in an independent multi-center dataset from three European countries not included in the training data. The key result is that malignancy could be ruled out in about 20% of patients with 5−15mm benign nodules using the Optellum Lung Cancer Prediction, who might otherwise undergo stressful and potentially harmful diagnostic workups
The successful performance of the Optellum AI could have a huge impact on the diagnostic workup in lung cancer screening, leading to fewer false positives and directing medical resources to those with the greatest need, and so facilitate a resource-efficient roll-out of CT lung cancer screening programs across Europe
Jan-Phillip Beck, Chief Executive Officer at EIT Health, commented:
“EIT Health is delighted to see the result of this consortium of leading university hospitals in Oxford, Groningen and Heidelberg, and lung health software company Optellum. We’re proud to have supported Optellum’s journey through our accelerator programme and several grant awards including the LUCINDA project. This illustrates how EIT Health is working hard to facilitate the rollout of artificial intelligence solutions in healthcare across Europe, transforming the lives of doctors and cancer patients.”
Read the paper here: https://www.lungcancerjournal.info/article/S0169-5002(21)00045-3/fulltext