Mu Zhou
Postdoctoral Scholar
BMIR, Stanford University

ABSTRACT:

Growing amounts of molecular data and clinical imaging arrays offer opportunities to gain insights into translational cancer research. Exploring the interplay between these heterogeneous cancer data would deepen our understanding for identifying imaging biomarkers of cancer. In this seminar, I will present two projects in the context of lung cancer research to highlight advances of radiogenomics, an emerging area that is defined to find underlying image-to-genomic associations in oncology. First, I will present our work on integrating transcriptome and CT image data in non-small cell lung cancer (NSCLC). We developed a radiogenomics map linking CT-based semantic features with gene expression clusters. More specifically, we proposed a radiogenomics strategy and identified non-invasive image biomarkers with prognostic implications by leveraging 17 independent, public gene expression cohorts with survival outcomes. Second, we extended to develop and validate preoperative high-dimensional image signatures that were predictive in differentiating EGFR mutations patients with lung adenocarcinoma. These methods allow identification of imaging biomarkers to better characterize molecular profiles, creating novel possibilities of non-invasive lung cancer diagnosis at very early stages. Finally, several challenges and ongoing research directions will be addressed in related areas.