Fused Biomarker-Based Prediction of Aggressive from Indolent Prostate Cancer
Abstract
Prostate cancer is often detected as localized tumors that can range from low malignant potential (which, if left untreated, are still unlikely to result in morbidity or reduce life expectancy) to those that are curable with a single modality directed exclusively to the gland itself or those destined to recur locally or systemically despite optimal local therapy. The last category that encompasses tumors that are broadly classified as "high-risk" or alternatively "locally advanced." Although the literature on "high-risk" prostate cancer is extensive and is increasing, no classification scheme exists that enables outcomes for patients with high-risk prostate cancer to be determined reliably and consistently to optimize patient management. The situation is further hindered by the wide range of diagnostic methods used to classify patients and by variations in the treatment itself from studies based primarily on a surgical or radiotherapy approach. We have assembled an exceptional research team consisting of a biomedical engineer, expert in medical image analysis and computational decision support, basic urologist, clinician, oncologist, biostatistician, and pathologist to generate highly accurate predictive models for a complex disease like prostate cancer. To differentiate between indolent versus aggressive disease and predict the recurrence of prostate cancer, we will (i) utilize a novel suite of computerized image analysis and computer vision tools to quantify histomorphometric features of tumor on H&E slides indicative of more aggressive prostate cancer to identify the cases that are likely to undergo biochemical failure at the time of diagnosis, and (ii) quantify cell type-specific and nuclear expression of two key biomarkers viz. NF-kappaB/p65/RelA (a marker for prostate cancer aggressiveness) and phosphorylated form of Akt (Ser473) (a marker for prostate cancer recurrence) that contribute to biochemical failure/recurrence and tumor progression. Under the proposed Specific Aim 1, we plan to optimize and qualify novel image segmentation, feature extraction, and classification tools to quantitatively characterize prostate cancer appearance on digitized histopathology using prostatectomy specimens. Studies in Specific Aim 2 will construct and validate computerized image-based histologic classifiers ("Signatures") that combine quantitative histomorphometric and molecular-based tissue biomarker measurements of the tumor using needle biopsy specimens. Successful development of this robust platform will have broad applications in patient diagnosis, treatment management, and prognostication.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Mar 29, 2016
- Source ID
- W81XWH1510558
Entities
People
- Sanjay Gupta
Organizations
- Case Western Reserve University
- United States Army