Skip to Content
COVID-19 Resources
Cancer Imaging Program (CIP)
Contact CIP
Show menu
Search this site
Last Updated: 10/15/18
Quantitative Imaging Network (QIN)

Dana Farber Cancer Institute

Genotype and Imaging Phenotype Biomarkers in Lung Cancer

Hugo Aerts, PhD
Grant Number: U01 CA190234

Advances in genomics have led us to recognize that tumors are characterized by distinct molecular events that drive development and progression of disease. But the need for repeated sampling of heterogeneous tumors and the relatively high cost of the assays provides limited opportunities to monitor the disease and its response to treatment. New quantitative imaging techniques and the emerging field of “radiomics” provides opportunities to search for predictive biomarkers using non-invasive imaging assays that can be used throughout the course of treatment.

Radiomics is based on application of Artificial Intelligence (AI) technologies, either based on predefined engineered algorithms or deep learning methods, to automatically quantify radiographic characteristics of the tumor phenotype1. Indeed, the Harvard-DFCI team have recently demonstrated that radiomic biomarkers have strong prognostic performance in large cohorts of cancer patients2, are associated with underlying gene-expression3 and somatic mutation patterns4,5, as well as being associated with treatment response6. Furthermore, working together with the CRICK institute in London, they found that imaging characteristics of the primary tumor are associated with circulation tumor DNA (ctDNA)7. The transformative hypothesis is that radiomic analysis, either alone or in combination with genomic mutational profile data obtained from biopsies, can provide a detailed characterization of the tumor phenotype.

In this current QIN project, the Harvard-DFCI team will develop a radiomics system that will be shared with the public8; develop a rigorous statistics platform specific for analyzing radiomic and genomics data, and apply their developments on a large cohort of non-small cell lung cancer (NSCLC) using tumor samples for which both non-invasive CT (PET) imaging data and mutational profiling data are available. The team will also explore whether the radiomic image features quantifying the tumor phenotype are related to genomic mutational profiles, providing a means to monitor non-invasively the molecular state of the disease throughout therapy. This proposal takes advantage of the Profile study at DFCI, a comprehensive personalized cancer medicine initiative generating mutational data on majority of patients undergoing therapy. The team will also leverage other public and private databases for discoveries.

Website Laboratory:

Selected Publications:

  1. Parmar C, Barry JD, Hosny A, Quackenbush J, Aerts HJWL, “Data Analysis Strategies in Medical Imaging”, Clinical Cancer Research, 2018 Aug 1;24(15):3492-3499.

  2. Hosny A, Parmar C, Quackenbush J, Schwartz LH, Aerts HJWL, “Artificial intelligence in radiology”, Nature Reviews Cancer, 2018 Aug;18(8):500-510.

  3. Barry JD, Fagny M, Paulson JN, Aerts HJWL, Platig J, Quackenbush J, “Histopathological Image QTL Discovery of Immune Infiltration Variants”, iScience. 2018 Jul 27;5:80-89.

  4. Aerts HJWL, “Data Science in Radiology: A Path Forward”, Clinical Cancer Research. 2018 Feb 1;24(3):532-534. doi: 10.1158/1078-0432.CCR-17-2804.