With increasing detection of early CaP with improved diagnostic methodologies, it has become important to predict biologic behaviors and """"""""aggressivity"""""""" to identify patients who might benefit from a """"""""wait and watch policy"""""""" as opposed to those who need more aggressive strategies. Traditionally, T-stage, amount of cancer in the core biopsy, the Gleason grade, and PSA at diagnosis has been used to evaluate the prognosis in localized CaP. While the Gleason score is currently assumed to be the strongest prognostic marker for CaP, there is often considerably high inter-, intra-observer variability associated with Gleason grade determination by pathologists. While some newer markers have recently shown promise, none of these methods have individually proven to be accurate enough to serve routinely as a prognostic marker for CaP. Recently, there has been a call to combine multiple prognostic markers to create an integrated meta-marker, with potentially greater accuracy in predicting CaP recurrence compared to any individual marker. While it is apparent that prognostic information resides in histopathology imagery in terms of the arrangement of nuclei and glands, sophisticated graph, and computerized image analysis algorithms are required to quantitatively model and characterize the architectural appearance of prostate cancer histopathology and thus provide a marker that is accurate and reproducible (unlike Gleason grade). In addition, while tumor micro-vascular density has been correlated to CaP outcome, prognostic information may also potentially reside in the specific spatial architectural arrangement of the micro-vascular network. The objective of the proposed work is to develop an integrated quantitative prognostic marker that combines information based on architectural arrangement of nuclear, glandular, and micro-vasculature network patterns on whole mount histology sections (WMHS) obtained via radical prostatectomy (RP) to predict prostate cancer recurrence. The proposed work comprises a total of 3 specific aims. For this project we will digitize approximately 100 annonymized WMHS obtained via RP that have been matched for Gleason score, stage, PSA, but with different clinical outcomes (half the patients having undergone cancer recurrence and the other not, following RP).
Under Aim 1, segmentation algorithms will be developed to automatically identify cancerous nuclei, glands and tumor microvasculature (MV), stained immuno-histochemically via CD31.
Under Aim 2 we will apply graph based image analysis algorithms to quantitatively characterize the architectural arrangement of CaP nuclei, glands and the MV network. These graph-based features will be integrated via a computerized machine learning algorithm to yield a numerical image based risk score (IbRiS) reflecting the CaP prognosis (disease recurrence or non-recurrence) of the patient. IbRiS will be evaluated in terms of its ability to distinguish between CaP progressors and non-progressors (matched for stage, Gleason grade, PSA), in a cohort of 50 independent studies (test set) for which survival and outcome data is available. This project will be a collaboration between investigators at Rutgers University (RU) and the University of Pennsylvania (UPENN). Data accrual will be done at UPENN while algorithmic development for computerized image analysis and classification will be carried out at RU.
The broad long term goal of this project is to develop an integrated image based histological biomarker for predicting prostate cancer (CaP) survival and outcome by integrating quantitative image signatures that define tissue architecture and vascular density patterns. Sophisticated image analysis and graph based algorithms will be developed to yield an image based risk score (IbRis) to predict whether or not a CaP patient will have disease recurrence following treatment. Our hypothesis is that IbRis will prove to be a better prognostic marker compared to such traditional markers as Gleason score and PSA.
Showing the most recent 10 out of 48 publications
