Skip to Content
Cancer Imaging Program (CIP)
Contact CIP
Show menu
Search this site
Last Updated: 09/21/17
Quantitative Imaging Network (QIN)

The H. Lee Moffitt Cancer Center & Research Institute

Radiomics of NSCLC

Robert Gillies, Ph.D.
Grant Number: U01 CA143062

The H. Lee Moffitt Cancer Center & Research Institute will address the issue of non-small cell lung cancer, NSCLC, through support from the Quantitative Imaging Network. NSCLC is the most prevalent of cancers and has one of the highest mortality rates. Thus, any advancement in the ability to predict response to individualize treatment will have great impact on the treatment of this disease. NSCLC patients are routinely imaged with PET and CT for staging and monitoring, respectively. The major hypothesis of the current work is that quantitative analysis of these clinical images can be prognostic and predictive of response to specific therapies. If true, these results would have medical significance through improved care and outcomes. This would also have socioeconomic significance as it would allow advanced, evidence based medicine to be practiced using standard-of-care images. To test this hypothesis, the project will extract mineable imaging data from two powerful patient databases at the Moffitt Cancer Center in Tampa, FL and the MAASTRO clinic in Maastricht, the Netherlands. These databases contain images, gene expression profiling and outcomes data from hundreds of stage III and IV NSCLC patients. Over 100 features will be extracted from each image using developmental commercial software. Features extracted retrospectively from the Moffitt dataset will be quantitatively analyzed to generate predictive models for gene expression patterns and progression-free survival. These models will be tested in the MAASTRO data set and re-tested using prospective data from Moffitt acquired under rigorous conditions. An important outcome of this work will define the rigor and resolution needed for images to be useful in predictive models. With the right combination of features, the needed rigor and resolution may be readily achievable in a clinical setting. A capstone experiment will add image feature extraction to a theragnostic trial that matches therapy to individual patient expression patterns for two proteins that predict response to specific therapies. The hypothesis to be tested is that image features can segment patients to specific therapy regimens without the molecular biopsy data.

This work will determine if quantitative analysis of images obtained during clinical standards of care can be used to determine outcome or predict response to specific therapies in lung cancer. If true, this would increase the utility of clinical imaging in this disease and potentially improve the care for up to 215,000 patients annually without necessarily increasing in the cost.

The specific aims of the research effort include:

Aim 1. Retrospectively compare of features from pre-therapy CT and PET/CT images to gene expression and clinical outcomes data in patients with NSCLC.
Aim 2. Prospectively evaluate sources of variance within the quantitative components of the image feature set.
Aim 3. A prospective trial in patients in an ongoing randomized phase III trial that customizes chemotherapy based on the molecular characteristics in the tumors of individual patients.