Pittsburgh Compound B (PiB) is a C11 PET tracer designed to bind to amyloid plaques, one of the hallmark of Alzheimer's disease. The potential of PiB as an early marker of Alzheimer's disease has lead to an increasing use of PiB and the development of several F18 equivalents. Quantitative analysis of PiB images requires an accurate normalisation, parcellation and estimation of retention in the brain's gray matter. Typically this relies on co-registered MRI to extract the cerebellum, compute the standardized uptake value ratio (SUVR) and provide parcellation and segmentation for quantification of neocortical SUVR. However, not all subjects undergo MRI. In this paper we propose a highly accurate MR-less parcellation, SUVR normalisation and quantification method for PiB images. This involves rigidly registering the raw PiB images to a PiB atlas, computing pair-wise normalised mutual information, and constructing a 2D manifold. Each new scan is mapped on the manifold and its k nearest neighbours are selected as atlases in a segmentation propagation scheme with their associated MRI segmentations and parcellation used as priors to estimate the SUVR normalisation and quantification. Comparison of our MRless approach to an MR-based approach showed a coefficient of correlation of neocortical PiB SUVR of R2=0.94 and an absolute mean error of 5.9%.