Radiological images contain hidden information about tumor phenotype, such as intratumor heterogeneity or tissue invasiveness, according to presenter Patrick Grossmann of Harvard Medical School.
"Radiomic analyses comprehensively quantify this tumor phenotype by defining hundreds of imaging features describing a tumor's shape, size, texture, and intensity," Grossmann explained. "Previously, strong associations between radiomics and clinical outcomes have been reported, but the underlying biology of these associations remains largely unknown."
To understand the relationship between radiomics, genomics, and clinical outcomes in lung cancer, the research team analyzed two independent cohorts of 262 North American and 89 European patients. After applying radiomics analysis to patient CT scans and measuring gene expression profiles, the researchers discovered association "modules" of strongly correlated imaging features and molecular pathways for cancer.
The molecular pathways were also correlated with known clinical factors of these pathways, including survival, tumor stage, and histologies. The team was also able to identify radiomic features that could actually predict whether a given cancer pathway was activated or deleted in a patient, Grossmann said.
By combining imaging features with gene expression signatures and clinical predictors, the researchers discovered that radiomics can increase the prognostic power of models that predict patient survival, he said.
"Because we show that radiomics complements genomic and clinical information, we address the critical and previously unanswered question of whether radiomics contributes to traditional approaches," Grossmann told AuntMinnie.com. "Together, our study shows that radiomic approaches permit a noninvasive assessment of both molecular and clinical characteristics of tumors, and could lead to the development of molecular biomarkers on the basis of standard-of-care imaging that doctors can use to personalize treatment options in a data-driven fashion."
