The goal of this research is to clinically translate software tools we developed through the Quantitative Imaging Network and validate their ability to assess the response of cancer in clinical trials. Current RECIST response criteria are inadequate to detect tumor changes in targeted molecular therapy and immunotherapies, two of the most promising avenues for drug discovery. We hypothesize that innovative volumetric and radiomics signatures of response and progression, identified using our quantitative CT imaging tools, can be integrated into clinical trial workflow to meet the urgent need for alternatives to RECIST criteria. Two large multi-site trials present a unique opportunity to test this hypothesis in one disease treated with multiple therapeutic options driven by tissue biomarkers. S0819 is a completed Phase III trial with 1300+ patients and Lung- MAP (S1400) is an ongoing first-of-its-kind Phase II/III model projected to enroll up to 5,000 patients using a multi-drug, targeted screening approach to match patients with sub-studies testing investigational treatments based on their unique tumor profiles.
Aim 1 tests whether change in tumor volume over time, measured by our advanced volumetric segmentation algorithms, outperforms unidimensional RECIST 1.1 response criteria.
Aim 2 correlates genomic mutations identified in S0819 and Lung-MAP with radiomics signatures constructed by our machine learning models, with the goal of developing a non-invasive, easily repeatable virtual biopsy through CT imaging.
Aim 3 validates the prediction of clinical outcomes using early biomarkers of response and progression based on quantitative CT-based radiomic features, hypothesized to outperform both RECIST and volumetrics alone across therapeutic options including chemotherapies, targeted molecular agents, and immune checkpoint blockade. Our work has substantial health significance because validation of volume and radiomic changes as early biomarkers of response or progression will guide clinical trials for drug discovery and help match patients to personalized treatment. Response criteria developed through this study will be widely applicable to clinical practice because CT is the most common cancer imaging modality and the quantitative image analysis tools can easily be incorporated into existing popular imaging platforms and clinical workflow, reducing the time required by radiologists. Data from this project, including anonymized imaging data (CT for all patients and PET for a large subset), clinical meta- data, and lesion mark-ups by independent radiologists, will be shared for use by other researchers through the TCGA Cancer Imaging Archive, continuing an extensive history of data sharing by this team.
This work will improve drug discovery for promising new classes of therapies by creating innovative response criteria based on volumetric and radiomics signatures developed with prior support by the QIN and now validated in two large multi-site clinical trials studying multiple therapeutic options in the most important histologic types of lung cancer and most commonly used imaging modality, CT. Results are expected to be applied to other tumor types and help match patients to the treatment that works best for them.
