The Quantitative Imaging Network (QIN) is a consortium of centers developing quantitative image features, which are proving to be valuable biomarkers of the underlying cancer biology and that can be used for assessing response to treatment and predicting clinical outcome. It is now important to discover the best quantitative imaging features for detection of response to therapeutics, to identify subtypes of cancer, and to correlate with cancer genomics. However, progress is thwarted by the lack of shared software algorithms, architectures, and resources required to compute, compare, evaluate, and disseminate these quantitative imaging features within the QIN and the broader community. We propose to develop the Quantitative Imaging Feature Pipeline (QIFP), a cloud-based, open source platform that will give researchers free access to these capabilities and hasten the introduction of quantitative image biomarkers into single- and multi-center clinical trials. The QIFP will facilitate assessment of the incremental value of new vs. existing image feature sets. It will also allow researchers to add their own algorithms to compute novel quantitative image features in their own studies and to disseminate them to the greater research community. To accomplish this: (1) We will create an expandable library of quantitative imaging feature algorithms capable of comprehensive characterization of the imaging phenotype of cancer. It will support a broad set of imaging modalities and algorithms implemented in a variety of languages, including algorithms that provide volumetric and time-varying assessment of lesion size, shape, edge sharpness, and pixel statistics. (2) We will build a cloud-based software architecture for creating, executing, and comparing quantitative image feature-generating pipelines, including algorithms in the library and/or those supplied by QIN or other researchers as plug-ins. QIFP will also have (a) a machine learning engine that lets users specify a dependent variable (e.g., progression-free survival) that the quantitative image features can used to predict, and (b) an evaluation engine that compares the utility of particular features for predicting the dependent variable. (3) We will assess the QIFP in four ways: (a) by its ability to recapitulate the role of known biomarkers in a related clinical trial, (b) by comparing linear measurement, metabolic tumor burden and novel combinations of the features in our library for predicting one-year progression-free survival, (c) by merging imaging features with known host-, drug- and tumor-based follicular lymphoma biomarkers in order to develop the most robust and integrative predictive model for patient outcomes, and (d) by using the QIFP to combine and to evaluate image feature algorithms developed by another QIN team and our own NCI- funded team in the study of radiogenomics of non-small cell lung cancer. The QIFP will fill a substantial gap in the science currently being carried out in the QIN and in the community by providing the tools and infrastructure to assess the value of novel quantitative imaging features of cancer, and will thereby accelerate incorporating new imaging biomarkers into single and multi-center clinical trials and into oncology practice.
We propose to develop and evaluate a software platform that has major relevance for human health. Many investigators are pursuing image-based surrogates for response to therapy that could be used in clinical trials to predict their success/failure earlier and that are more accurate than existing surrogates. Our developments will facilitate sharing, assessing, and comparing combinations of image feature-generating software algorithms for predicting treatment response, survival, and tissue genomics, which will, in turn, greatly accelerate the development and acceptance of new and more relevant imaging surrogates for assessing cancer treatments.
|Banerjee, Imon; Malladi, Sadhika; Lee, Daniela et al. (2018) Assessing treatment response in triple-negative breast cancer from quantitative image analysis in perfusion magnetic resonance imaging. J Med Imaging (Bellingham) 5:011008|
|Napel, Sandy; Mu, Wei; Jardim-Perassi, Bruna V et al. (2018) Quantitative imaging of cancer in the postgenomic era: Radio(geno)mics, deep learning, and habitats. Cancer 124:4633-4649|
|Bakr, Shaimaa; Gevaert, Olivier; Echegaray, Sebastian et al. (2018) A radiogenomic dataset of non-small cell lung cancer. Sci Data 5:180202|
|El Naqa, Issam; Napel, Sandy; Zaidi, Habib (2018) Radiogenomics is the future of treatment response assessment in clinical oncology. Med Phys 45:4325-4328|
|Balagurunathan, Yoganand; Beers, Andrew; Kalpathy-Cramer, Jayashree et al. (2018) Erratum: Semi-automated pulmonary nodule interval segmentation using the NLST data. Med Phys 45:2689-2690|
|Wu, Jia; Cao, Guohong; Sun, Xiaoli et al. (2018) Intratumoral Spatial Heterogeneity at Perfusion MR Imaging Predicts Recurrence-free Survival in Locally Advanced Breast Cancer Treated with Neoadjuvant Chemotherapy. Radiology 288:26-35|
|Balagurunathan, Yoganand; Beers, Andrew; Kalpathy-Cramer, Jayashree et al. (2018) Semi-automated pulmonary nodule interval segmentation using the NLST data. Med Phys 45:1093-1107|
|Zhou, Mu; Leung, Ann; Echegaray, Sebastian et al. (2018) Non-Small Cell Lung Cancer Radiogenomics Map Identifies Relationships between Molecular and Imaging Phenotypes with Prognostic Implications. Radiology 286:307-315|
|Banerjee, Imon; Chen, Matthew C; Lungren, Matthew P et al. (2018) Radiology report annotation using intelligent word embeddings: Applied to multi-institutional chest CT cohort. J Biomed Inform 77:11-20|
|Echegaray, Sebastian; Bakr, Shaimaa; Rubin, Daniel L et al. (2018) Quantitative Image Feature Engine (QIFE): an Open-Source, Modular Engine for 3D Quantitative Feature Extraction from Volumetric Medical Images. J Digit Imaging 31:403-414|
Showing the most recent 10 out of 29 publications