asco@alo33:170642
|
Background: We have previously shon the importance of utilizing the intact tissue section and quantitative multiplex immunofluorescence MIF for developing accurate prognostic risk models in prostate cancer. Our earlier biopsy-based surgical outcome model utilized the clinical biopsy Gleason score ith Hx0026;E and MIF images. We no have utilized next-gen MIF image analysis to deconstruct the histologic attributes of the current Gleason grading system. This resulted in the generation of highly complex IF morphometric features that improve upon traditional histologic assessment Methods: We utilized a prior MIF image training cohort of 628 patients ith a median 8 years follo-up post-radical prostatectomy. The analytic group constituted 2400 MIF images from biopsy tissue probed ith CK-18 epithelial cells , CK 56 basal cells , androgen receptor AR , Ki67, AMACR and DAPI nuclear marker . Novel image analysis features that reflect biopsy tumor grade and biology ere included to discriminate Gleason score GS 6 + organ confined disease, lt = pT2 Favorable Pathology vs.any pattern 4 and or non-organ confined = pT3a UFP . AUC as used to assess predictive performance. Results: We have been able to replicate our original models ith ne image features that more accurately reflect Gleason grading in a quantitative, non-subjective manner. These ne models do not contain any Hx0026;E features nor is the clinical Gleason utilized. The AUC of the ne training models ranged from 0.75-0.78 original FP model had an AUC = 0.78 , sensitivty, 70 percent and specificity 70-74 percent. In addition to pre-surgical PSA, the MIF image features selected by the model included gland ring structures hich characterize the architecture of the prostate cancer gland combined ith Androgen Receptor and Ki67 distribution and levels throughout the stroma and epithelium, respectively. Conclusions: The ability to more accurately characterize the Gleason grading system and refine risk classification has the potential to impact on future treatment decisions.,J Clin Oncol 34, 2016 suppl; abstr e16610 ,Publication Only Genitourinary Prostate Cancer
|