RSNA 2021 Imaging Informatics Preview

CT radiomics help stratify risk in NSCLC patients

By Erik L. Ridley, staff writer

Tuesday, November 30 | 3:00 p.m.-4:00 p.m. | SSIN05-1 | Room S401
In this session, researchers will share how CT radiomics can enable prognostic stratification in patients with non-small cell lung cancer (NSCLC).

Presenter Mitchell Chen, PhD, of Imperial College London in the U.K. and colleagues assessed the utility of CT radiomics in a retrospective study involving 292 NSCLC patients diagnosed at their institution over a four-year period. Training was performed on two-thirds of the patients; the remaining cases were set aside for validation.

The radiologists double-reviewed the scans and performed semiautomated segmentation of the tumor, peritumoral penumbra, and a spherical parenchymal patch in the surrounding pulmonary lobe, according to the researchers. The imaging data were preprocessed and then analyzed for radiomics features using internally developed software.

After the researchers performed unsupervised hierarchical clustering analysis for each histological subtype on the normalized radiomics profile, they found significant difference in survival for both squamous cell carcinomas (p = 0.0013) and adenocarcinoma (p = 0.0080). In addition, a composite radiomics prognostic vector (RPV) -- derived from least absolute shrinkage and selection operator (LASSO)-Cox regression feature selection -- showed significant difference (p < 0.001) in survival between the risk groups.

"In this work, we present clear evidence supporting the clinical utility of CT-based radiomic analysis in NSCLC," the authors wrote.

What else did they find? Check out this talk on Tuesday to get all of the results.

Last Updated er 11/18/2021 4:12:43 PM