Intratumoral heterogeneity is a hallmark of diffuse gliomas and inhibits effective prediction of clinical outcome. To devise a three-dimensional reconstruction of tumor lineage, we used neuronavigation to acquire eight to twelve image-guided and spatially separated stereotactic biopsy samples from 16 adult patients with a diffuse glioma, which we characterized using DNA methylation arrays. A total of 133 samples were obtained from regions with and without imaging abnormalities. Methylation profiles were analyzed to construct phylogenetic trees and subsequently projected on image-derived tumor maps. Lineage analysis of these evolutionary trees indicated that the sampled gliomas largely evolved stochastically, suggesting that critical tumor drivers were acquired early in time. These results were further validated using 102 multi-region samples from 24 independent patients. Patristic (evolutionary) and cartesian (spatial) distances between pairs of tumor samples from the same patient demonstrated strong correlations, suggesting that this information could be used to determine trajectories of tumor evolution. Evolutionary and spatial distance metrics were combined with histologically obtained and computationally quantified cellularity and proliferation rates to infer vectors representing the direction and magnitude of tumor growth. Using the resulting vector field we determined the minimum and maximum rates of change in order infer the tumor’s evolutionary trajectory. Using this metric we identified three distinct growth patterns: (1) tumors with a trajectory oriented towards the tumor core, (2) outward growing tumors with a linear trajectory pointing outside tumor boundaries, and (3) outward growing tumors with a branching trajectory directed outside tumor boundaries. Association of evolutionary growth patterns with survival demonstrated distinct impacts on outcome, suggesting that growth patterns are an important determinant of tumor aggressiveness. Taken together, although our analyses indicated that the observed glioma heterogeneity is small and largely stochastic, when spatially mapped these changes can be used to track tumor lineage and identify clinically relevant growth patterns.