||CADESS is a decision support system for the prognostication of prostate cancer from tissue samples. CADESS bases its algorithms on tissue data from whole mount sections comprising 650 small images, each with one dominant pattern. The 650 images were consensus-graded by a panel of thirteen internationally prominent uropathologists.
At the 2nd Digital Pathology Congress in London we described the consensus-grading process and the resulting intra- and inter-observer grading variations. During a consensus meeting following the grading process, we reached a complete consensus for about 90% of the images, but little or no consensus on the remaining images.
In this talk, which I gave in July at the Digital Pathology Congress in Philadelphia, USA, I use multi-dimensional scaling to illustrate the grading variations using the new five grade group grading system from Johns Hopkins University. I also show results of the grading of the original whole mounts by the consensus-panel relating the results to the grading of the small sub-images, and answer the question: Did context help in the grading?