By 2020 the number of patients with p16+ (HPV-related) oropharyngeal squamous cell carcinoma (OSCC) is predicted to surpass that for cervical SCC. Epidemiologists have termed this a head and neck cancer "epidemic." At the same time, there is an emerging view that we may be "over-treating" patients with p16+ (HPV-related) OSCC because it is typically more biologically indolent, with tumors having less gross chromosomal abnormalities, ~1/2 the gene mutation rate, and favorable responses to treatment. For these reasons, many speculate that therapies could be "de-escalated" to maintain favorable patient survival while minimizing treatment-related morbidity. However, a significant minority of patients with p16+ OSCC have aggressive disease that will recur, predominantly in the form of distant metastasis, resulting in death. There are currently few clinical-and no molecular-markers to discriminate more from less aggressive p16+ OSCC. The focus of this project is to optimize and evaluate a quantitative histomorphometric (QH)-based image classifier (QuHbIC) to identify which p16+ OSCC are likely to be clinically aggressive and which OSCC patients have cancers that are very unlikely to recur. QuHbIC only requires digitized images of standard hematoxylin and eosin (H&E) stained sections, from which a series of features describing spatial distribution, morphology, texture and arrangement of tumor and stromal cell nuclei will be extracted via advanced computer vision and pattern recognition tools. Thus "histologic biomarkers" for more and less aggressive p16+ OSCC will be identified. Although molecular genetic approaches have become popular for tumor characterization, H&E morphology is still remarkably useful. In reality, tumor morphology reflects the sum of all molecular pathways in tumor cells, thereby providing incredible utility for predicting tumor biology, clinical behavior, and treatment response. While the visual reading of such slides by pathologists can predict behavior, sophisticated histomorphometric analysis with computer-aided quantitation has the potential to "unlock" more revealing information about tumors just from their morphology. The hypotheses underlying this project are that (a) markers for disease aggressiveness are encoded in visual attributes in histological (biopsy or resection) images of cancer, and some of these "histologic biomarkers" (e.g. nuclear anaplasia and/or multi-nucleation) can be correlated with disease recurrence independent of other clinical and pathologic features;and (b) these "histologic biomarkers" can be extracted via computerized image analysis. QuHbIC will be trained and refined via a large cohort of digitized H&E slides with long term clinical follow up data from Washington University in St. Louis (Wash U). Independent evaluation of the classifier will be performed on scanned H&E slides available from both Wash U and Johns Hopkins University. The successful validation of QuHbIC could pave the way for rapid integration of QuHbIC into the clinical workflow as a decision support tool, providing critical information to assist oncologists in making more informed treatment decisions.
In this project we propose to develop and evaluate a quantitative histomorphometric (QH)- based image classifier (QuHbIC) to identify which p16+ (HPV-related) oropharyngeal squamous cell carcinoma patients (OSCC) are likely to have aggressive cancers and, perhaps even more importantly, for identifying which patients have cancers that are very unlikely to recur. The successful validation of QuHbIC could pave the way for its adoption as a decision support tool, allowing for the identification of patients for whom therapies could be de-escalated or specifically targeted in order to maintain favorable survival rates while minimizing the substantial (and long term) treatment-related morbidity and expense from surgery, radiation, and chemotherapy.
|Sparks, Rachel; Madabhushi, Anant (2016) Out-of-Sample Extrapolation utilizing Semi-Supervised Manifold Learning (OSE-SSL): Content Based Image Retrieval for Histopathology Images. Sci Rep 6:27306|
|Janowczyk, Andrew; Basavanhally, Ajay; Madabhushi, Anant (2016) Stain Normalization using Sparse AutoEncoders (StaNoSA): Application to digital pathology. Comput Med Imaging Graph :|
|Antunes, Jacob; Viswanath, Satish; Rusu, Mirabela et al. (2016) Radiomics Analysis on FLT-PET/MRI for Characterization of Early Treatment Response in Renal Cell Carcinoma: A Proof-of-Concept Study. Transl Oncol 9:155-62|
|Ginsburg, Shoshana B; Algohary, Ahmad; Pahwa, Shivani et al. (2016) Radiomic features for prostate cancer detection on MRI differ between the transition and peripheral zones: Preliminary findings from a multi-institutional study. J Magn Reson Imaging :|
|Shiradkar, Rakesh; Podder, Tarun K; Algohary, Ahmad et al. (2016) Radiomics based targeted radiotherapy planning (Rad-TRaP): a computational framework for prostate cancer treatment planning with MRI. Radiat Oncol 11:148|
|Litjens, Geert J S; Elliott, Robin; Shih, Natalie N C et al. (2016) Computer-extracted Features Can Distinguish Noncancerous Confounding Disease from Prostatic Adenocarcinoma at Multiparametric MR Imaging. Radiology 278:135-45|
|Romo-Bucheli, David; Janowczyk, Andrew; Gilmore, Hannah et al. (2016) Automated Tubule Nuclei Quantification and Correlation with Oncotype DX risk categories in ER+ Breast Cancer Whole Slide Images. Sci Rep 6:32706|
|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|
|Singanamalli, Asha; Rusu, Mirabela; Sparks, Rachel E et al. (2016) Identifying in vivo DCE MRI markers associated with microvessel architecture and gleason grades of prostate cancer. J Magn Reson Imaging 43:149-58|
|Leo, Patrick; Lee, George; Shih, Natalie N C et al. (2016) Evaluating stability of histomorphometric features across scanner and staining variations: prostate cancer diagnosis from whole slide images. J Med Imaging (Bellingham) 3:047502|
Showing the most recent 10 out of 27 publications