We propose an alternative, inexpensive means for identifying the subset of women with ER+ breast cancer for whom hormonal therapy alone is sufficient, and adjuvant chemotherapy is not required. Our novel approach uses computer vision and machine learning techniques to extract information from digitized histological sections of breast tissue (e.g. graphical and morphological arrangement of nuclei, lymphocytes, textural appearance, and tubule density), yielding a continuous image-based risk score (IbRiS) from zero to one that predicts the risk of recurrence. By choosing the appropriate cutoff value, IbRiS can be used to stratify ER+ patients into two classes: low and high risk. A low risk identifies those women who would respond well to hormonal therapy alone. Of the 120,000 women annually diagnosed with estrogen receptor positive (ER+) breast cancer in the US, the vast majority will be considered at high risk of having distant recurrence (metastasis) within 10 years. Under current National Comprehensive Cancer Network (NCCN) guidelines, these women will be advised to receive adjuvant chemotherapy in addition to standard hormonal treatment (e.g. tamoxifen). However, 85% of these women will not benefit from chemotherapy, and yet will still incur its deleterious side-effects. The only non-investigational tool to help better determine which women should receive tamoxifen alone is the Oncotype Dx molecular assay, which is now widely used by medical oncologists. Oncotype Dx is able to correctly reclassify (as low-risk) 50% of those women with ER+ cancers that have been classified as high-risk (or intermediate-risk) by the NCCN guidelines, obviating their need for chemotherapy. However, Oncotype Dx is inaccessible to the majority of women in the US (and worldwide) because of its high cost ($4500), and requirement that the tissue samples be sent to specialized remote facilities. Consequently, there is clearly a market need for a lower-priced assay capable of reaching a wider audience.
The specific aims of this proposal are as follows: 1) develop an image-based risk score (IbRiS) for predicting the subset of women with ER+ breast cancer that wil respond wel without chemotherapy and 2) evaluate IbRiS performance in predicting recurrence over an independent set of ER+ breast cancers treated with tamoxifen alone. IbRiS has several key advantages over molecular assays. First, it requires no disruption of the current clinical protocol since the necessary tissue samples are already collected during routine pathological examinations. Second, IbRiS has a zero cost-of-goods sold, and thus could serve as either a lower-priced alternative to a molecular assay or as a quantitative triage, determining which patients should be administered the more expensive molecular test. Finally, since IbRiS only requires a digital slide scanner (i.e. no specialized facility is necessary), its footprint could extend worldwide (via the internet).
Every year tens of thousands of women in the US with estrogen receptor positive (ER+) breast cancer are treated with chemotherapy, though only a few thousand will benefit from it. In this proposal we will develop an image-based risk score (IbRiS) to predict which women with ER+ breast cancer do not require chemotherapy. This test will provide an economical alternative to the far more expensive gene-expression based assays currently in use.
|Cruz-Roa, Angel; Gilmore, Hannah; Basavanhally, Ajay et al. (2017) Accurate and reproducible invasive breast cancer detection in whole-slide images: A Deep Learning approach for quantifying tumor extent. Sci Rep 7:46450|
|Rusu, Mirabela; Rajiah, Prabhakar; Gilkeson, Robert et al. (2017) Co-registration of pre-operative CT with ex vivo surgically excised ground glass nodules to define spatial extent of invasive adenocarcinoma on in vivo imaging: a proof-of-concept study. Eur Radiol 27:4209-4217|
|Wan, Tao; Bloch, B Nicolas; Plecha, Donna et al. (2016) A Radio-genomics Approach for Identifying High Risk Estrogen Receptor-positive Breast Cancers on DCE-MRI: Preliminary Results in Predicting OncotypeDX Risk Scores. Sci Rep 6:21394|
|Madabhushi, Anant; Lee, George (2016) Image analysis and machine learning in digital pathology: Challenges and opportunities. Med Image Anal 33:170-175|
|Xu, Jun; Xiang, Lei; Liu, Qingshan et al. (2016) Stacked Sparse Autoencoder (SSAE) for Nuclei Detection on Breast Cancer Histopathology Images. IEEE Trans Med Imaging 35:119-30|
|Tiwari, P; Prasanna, P; Wolansky, L et al. (2016) Computer-Extracted Texture Features to Distinguish Cerebral Radionecrosis from Recurrent Brain Tumors on Multiparametric MRI: A Feasibility Study. AJNR Am J Neuroradiol 37:2231-2236|
|Ginsburg, Shoshana B; Taimen, Pekka; Merisaari, Harri et al. (2016) Patient-specific pharmacokinetic parameter estimation on dynamic contrast-enhanced MRI of prostate: Preliminary evaluation of a novel AIF-free estimation method. J Magn Reson Imaging 44:1405-1414|
|Xu, Jun; Luo, Xiaofei; Wang, Guanhao et al. (2016) A Deep Convolutional Neural Network for segmenting and classifying epithelial and stromal regions in histopathological images. Neurocomputing 191:214-223|
|Sparks, Rachel; Bloch, B Nicolas; Feleppa, Ernest et al. (2015) Multiattribute probabilistic prostate elastic registration (MAPPER): application to fusion of ultrasound and magnetic resonance imaging. Med Phys 42:1153-63|
|Ginsburg, Shoshana B; Viswanath, Satish E; Bloch, B Nicolas et al. (2015) Novel PCA-VIP scheme for ranking MRI protocols and identifying computer-extracted MRI measurements associated with central gland and peripheral zone prostate tumors. J Magn Reson Imaging 41:1383-93|
Showing the most recent 10 out of 25 publications