Advances in genomics have led to the recognition that tumors are characterized by distinct molecular events that drive their development and progression. However, the need for repeated sampling of heterogeneous tumors, together with the relatively high cost of assays, limits their use in monitoring the disease and its response to treatment. New medical imaging technologies and the emerging field of radiomics quantifies the tumor phenotype at a macroscopic level, allowing identification predictive phenotypic biomarkers using non- invasive imaging assays that is routinely collected throughout the course of treatment. We recently demonstrated that radiomic biomarkers have strong prognostic performance in large cohorts of lung and head and neck cancer patients, and are associated with underlying mutation and gene-expression patterns. A critical barrier hampering the widespread use of such quantitative features in clinical practice is the lack of robust software tools for the identification of imaging biomarkers and a collection of validated markers that have been shown to work across sites. Part of the reason for the relatively slow progress is that technical developments in quantitative imaging are often isolated; radiomics feature definitions are non-standardized; implementations occur in proprietary environments that make scientific exchange difficult; and analyses are focused on a single disease site or imaging modality. Here we propose to construct a publicly available computational radiomics system for the objective and automated extraction of quantitative imaging features that we believe will yield biomarkers of greater prognostic value compared with routinely extracted descriptors of tumor size. In this proposal, we will outlines research and development plans focused on creating a generalized, open, portable, and extensible radiomics platform that is widely applicable across cancer types and imaging modalities and describe how we will use lung and head and neck cancers as models to validate our developments. To achieve our goals we will identify and implement a large array of quantitative imaging features, develop a flexible radiomics platform usable by both image analysis experts (such as engineering scientists) and imaging non-experts (such as bioinformatics scientists or physicians) alike, and validate these developments by integrating radiomics, genomics, and clinical data to evaluate prognostic performance and examine associations. We will take advantage of The Cancer Imaging Archive (TCIA) with imaging data, and The Cancer Genome Atlas (TCGA), with corresponding genomic and clinical data. Throughout the project all software, tools, and other resources will be made freely available to ensure community building. We have assembled an interdisciplinary team including experts in imaging, computational biology, molecular biology, oncology, and bioinformatics that we believe uniquely positions us to substantially advance the field of radiomics and provide tools that will allow its translational use in the clinic.
One of the most difficult yet important tasks in providing cancer care is predicting whether a patient's tumor is likely to respond to a particular therapy. We now recognize that outcome in cancer depends on molecular changes in the tumor cell that activate particular genetic programs and have developed biopsy-based tests to search for genetic 'biomarkers' that can predict likely drug response. In this application, we present our plans to develop advanced non-invasive imaging technologies (that can be used throughout treatment) that can be used to identify morphological biomarkers that can both predict and monitor response to treatment.
| 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 |
| Hosny, Ahmed; Parmar, Chintan; Quackenbush, John et al. (2018) Artificial intelligence in radiology. Nat Rev Cancer 18:500-510 |
| Dou, Tai H; Coroller, Thibaud P; van Griethuysen, Joost J M et al. (2018) Peritumoral radiomics features predict distant metastasis in locally advanced NSCLC. PLoS One 13:e0206108 |
| 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 |
| Barry, Joseph D; Fagny, Maud; Paulson, Joseph N et al. (2018) Histopathological Image QTL Discovery of Immune Infiltration Variants. iScience 5:80-89 |
| Parmar, Chintan; Barry, Joseph D; Hosny, Ahmed et al. (2018) Data Analysis Strategies in Medical Imaging. Clin Cancer Res 24:3492-3499 |
| Aerts, Hugo J W L (2018) Data Science in Radiology: A Path Forward. Clin Cancer Res 24:532-534 |
| Rios Velazquez, Emmanuel; Parmar, Chintan; Liu, Ying et al. (2017) Somatic Mutations Drive Distinct Imaging Phenotypes in Lung Cancer. Cancer Res 77:3922-3930 |
| Grossmann, Patrick; Stringfield, Olya; El-Hachem, Nehme et al. (2017) Defining the biological basis of radiomic phenotypes in lung cancer. Elife 6: |
| Trebeschi, Stefano; van Griethuysen, Joost J M; Lambregts, Doenja M J et al. (2017) Deep Learning for Fully-Automated Localization and Segmentation of Rectal Cancer on Multiparametric MR. Sci Rep 7:5301 |
Showing the most recent 10 out of 30 publications
