Imaging biomarkers have been widely used for screening, diagnosing, and measuring the progression of disease in neuroradiology. However, quantifying these imaging biomarkers is costly, time-consuming, and often not reproducible. Here, we propose statistical segmentation models for neuroimaging data to address these issues and resolve the bottleneck in imaging research. Chapter 2 introduces a fast, free, consistent, and unsupervised beta-mixture oligodendrocyte segmentation system (FAST) that can segment and track oligodendrocytes in three-dimensional images over time with minimal human input. We demonstrate that the FAST model can segment and track oligodendrocytes similarly to a blinded human observer. For non-invasive neuroimaging data, Chapter 3 proposes an atlas-based normalization method that can improve the signal-to-noise ratio while reducing the inhomogeneous intensities of the thalamus in magnetic resonance imaging. With this normalization method, chapter 4 introduces an error-adjusting segmentation model that provides error-adjusted covariate effect estimates, as well as estimates for false-positive and false-negative rates of the training data. Through applications to multiple sclerosis thalamic lesion segmentation in magnetic resonance imaging, we showed that our proposed model offers better assistance in detecting thalamic lesions. These new statistical segmentation models can facilitate neuroradiology research and assist in diagnostic and disease monitoring application by providing robust, reliable, and reproducible results.