Computer Vision News - February 2023

37 Sybil: Predict Future Lung Cancer Risk “ One of our collaborators is looking at a cohort of firefighters, and in Taiwan, there’s a big concern that cancer could be related to pollution, ” Jeremy points out. “ Ideally, screening should be something everybody does, but of course, that would be expensive. Hopefully, these technologies enable more efficient screening, meaning you can scan more people within the same budget. I don’t think that will be the initial aim, but it’s important if you want to expand to a broader population. ” Computational models that identify cancers inCT have already beendeveloped, but to the best of the group’s knowledge, their model is one of the first to forecast so far ahead. Radiologists are skilled at detecting cancers but predicting what might happen in the future , which is critical when looking at modifications to screening policies, is more complex. The group felt confident their model would achieve its aims, having proven a similar concept in breast screening with mammograms , but were aware that Early detection is key in improving the survival rate for lung cancer . Screening tests can help identify lung cancer early, even before symptoms appear. This work explores whether the low-dose CT scans used for screening today could also be used to forecast the risk of someone getting cancer within a certain number of years . “ Our model predicts whether or not cancer will develop within the next one to six years , ” Peter tells us. “ We achieved this using data from a big trial that happened in the early 2000s, but we also wanted to validate it beyond that one trial, so we had data from our collaborators here in Boston at Massachusetts General Hospital (MGH) and the Chang Gung Memorial Hospital in Taiwan and were able to show that amongst those two populations, the model maintained its performance. ” Low-dose CTs are typically recommended for people aged 50 years and over with a smoking history, but the number of people without any smoking history developing lung cancer is growing.

RkJQdWJsaXNoZXIy NTc3NzU=