Breast cancer is a heterogeneous disease. Around 20% to 30% of women diagnosed with invasive breast cancer will have a recurrence and may eventually die of their disease. Currently, there are no reliable methods to identify which cancers will recur on an individual basis. Because of this, adjuvant therapies are given to nearly all patients with breast cancer, but benefit only a small proportion. A similar dilemma exists for neoadjuvant treatment, many patients fail to pathologically response to chemotherapy, and yet suffer from the associated toxicity. The conventional one-size-fits-all approach causes overtreatment, leading to morbidities and mortalities. To avoid these side effects, biomarkers that stratify patients with clinical relevance are critically needed for precision medicine in breast cancer. Molecular profiling is currently used to stratify breast cancer, but is limited by the requirement for invasive biopsy and confounded by intra-tumor genetic heterogeneity. Conversely, imaging provides a unique opportunity for the noninvasive interrogation of the tumor, its microenvironment, and invasion to surrounding normal tissues. We hypothesize that imaging characteristics reflect underlying tumor biology, and quantitative imaging features can provide independent valuable information, which are synergistic to known clinical, histologic, and genetic predictors. Accordingly, we have planned three specific aims to develop new quantitative imaging biomarkers for breast cancer, as well as clinically and biologically validate them.
In Aim 1 we plan to develop automated computational tools to robustly quantify whole tumor, intratumor subregions, and parenchyma phenotypes from multimodal MRI. The curated breast cancer cohort (n=504) from our preliminary study will be analyzed, with available MRI scans and manually-delineated contours of tumor and parenchyma by board-certified radiologists.
In Aim 2 we will build imaging feature-based models to predict recurrence-free survival and treatment response separately. By integrating with clinicopathologic and genomic predictors, the comprehensive models can predict clinical outcomes more accurately. The internal cohort (n=450) will be used for discovery, and the multi-center prospective cohort from I-SPY (n=186) will be used for validation.
In Aim 3 we will elucidate the biological underpinnings behind our newly identified prognostic and predictive imaging biomarkers, by correlating them with biospecimen-derived phenotypes from the same tumor. In particular, we will investigate multi-omics molecular data as well as tumor morphology from H&E stained pathology slides. Three cohorts will be analyzed, including our internal cohort (n=450), the I-SPY cohort (n=186), and the TCGA cohort (n=1095). For three proposed aims, we have carried preliminary studies to prove the feasibility. By leveraging the richness of available well-annotated data and advanced artificial intelligence algorithms, it will increase the likelihood of success. Our proposed research will point new biomarkers of high value to better predict recurrence and treatment response at the individual level, and lead to better treatment decisions for women with breast cancer.
This project aims to augment the imaging role in the personalized management of breast cancer, by developing automated tools to extract quantitative imaging biomarkers from both tumor and parenchyma, as well as evaluating the biological and clinical relevance of these imaging biomarkers. The new imaging biomarkers will work synergistically with the existing clinicopathologic and genomic predictors, to better stratify breast cancer patients and guide individualized therapy. Together, these will promote a major paradigm shift of MR imaging role and allow it serve as an important approach in personalized cancer management.
| 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 |
| Liu, Lei; Li, Kai; Qin, Wenjian et al. (2018) Automated breast tumor detection and segmentation with a novel computational framework of whole ultrasound images. Med Biol Eng Comput 56:183-199 |
| Wu, Jia; Li, Xuejie; Teng, Xiaodong et al. (2018) Magnetic resonance imaging and molecular features associated with tumor-infiltrating lymphocytes in breast cancer. Breast Cancer Res 20:101 |
| Qin, Wenjian; Wu, Jia; Han, Fei et al. (2018) Superpixel-based and boundary-sensitive convolutional neural network for automated liver segmentation. Phys Med Biol 63:095017 |
