Integration of Architectural and Cytologic Driven Image Algorithms for Prostate Adenocarcinoma Identification
Jason D Hipp, James Monaco, Priya L Kunju, Jerome Cheng, Yukako Yagi, Jaime Rodriguez-Canales, Michael R Emmert-Buck, Stephen Hewitt, Michael D Feldman, John E Tomaszewski, Mehmet Toner, Ronald G Tompkins, Thomas Flotte, David Lucas, John R Gilbertson, Anant Madabhushi, Ulysses J Balis. University of Michigan, Ann Arbor; Rutgers The State University of New Jersey, Piscataway, NJ; Harvard, Boston; National Cancer Institute, NIH, Bethesda; Perlman School of Medicine at the University of Pennsylvania, Philadelphia; Mayo Clinic, Rochester, MN; Massachusetts General Hospital, Boston, MA
Background: The advent of digital slides offers new opportunities including the use of image analysis techniques to facilitate computer aided diagnosis (CAD) solutions. However, the development and testing of prostate cancer CAD solutions requires a ground truth map of the cancer which in turn requires a pathologist to annotate, or paint, each of the malignant glands in prostate cancer with an image editor software - a time consuming and exhaustive process.
Recently, two CAD algorithms have been described: probabilistic pairwise Markov models (PPMM) and spatially-invariant vector quantization (SIVQ). Briefly, SIVQ operates as a highly sensitive and specific pattern matching algorithm, making it optimal for the identification of epithelial morphology, whereas PPMM operates as a highly sensitive detector of malignant perturbations in glandular architecture.
Design: By recapitulating algorithmically how a pathologist reviews prostate tissue sections, we created an algorithmic cascade of PPMM and SIVQ as previously described by Doyle el al. (2011) where PPMM identifies the glands with abnormal lumenal architecture, and this area is then screened by SIVQ to identify the epithelium.
Results: The performance of this algorithm cascade was assessed qualitatively (using heatmaps) and quantitatively (using ROC curves). The PPMM-SIVQ analysis had approximately 90% sensitivity, 90% specificity (Sample #1), 85% sensitivity, 94% specificity (Sample #2), and 90% sensitivity, 88% specificity (Sample #3). This data demonstrates greater performance in the identification of only the prostatic adenocarcionoma than PPMM or SIVQ alone.
Conclusions: This ability to semi-autonomously paint nearly all the malignant epithelium of prostate cancer has immediate applications to future prostate cancer CAD development as a validated ground truth generator. In addition, such an approach has potential applications as a pre-screening/quality assurance tool.
Category: Genitourinary (including renal tumors)
Wednesday, March 21, 2012 9:30 AM
Poster Session V # 110, Wednesday Morning