“Radiomics” is the process of extracting structured and mineable data from biomedical images, and then using these data to provide more power for cancer diagnosis and prognosis, as well as prediction and monitoring response to anti-cancer therapies (1). Lung “nodules” are commonly observed in chest CT scans, which are coherent structures that may or may not be cancerous. These can be observed in a screening setting (for those at high risk defined by the US Preventive Service as Adults 55-80 with a history of smoking), or they can be detected incidentally (during a CT scan for another concern). In either case, there is a significant problem of over-detection and over-diagnosis, as 96% of indeterminate screening nodules and over 60% of incidental nodules are not cancers, but nonetheless require follow up. Radiomics has shown exceptional power in being able to better define the future progression of these nodules and will eventually reduce the burden of over-detection and over-diagnosis (2-5).
Radiomics features distinguish cancer from non-cancer in incidentally detected IPNs
Hawkins et al., JTO, 2016
Robert J Gillies, PhD
PI, Chair of Department