Local hemodynamics are an important factor in atherosclerosis, from the development of early lesions, to the assessment of stroke risk, to determining the ultimate fate of a mature plaque. Until recently, our understanding of arterial fluid dynamics and their relationship to atherosclerosis was limited by the use of idealized or averaged artery models. Recent advances in medical imaging, computerized image processing, and computational fluid dynamics (CFD) now make it possible to computationally reconstruct the time-varying, three-dimensional blood flow patterns in anatomically realistic models. In this paper we review progress, made largely within the last five years, towards the routine use of anatomically realistic CFD models, derived from in vivo medical imaging, to elucidate the role of local hemodynamics in the development and progression of atherosclerosis in large arteries. In addition to describing various image-based CFD studies carried out to date, we review the medical imaging and image processing techniques available to acquire the necessary geometric and functional boundary conditions. Issues related to accuracy, precision, and modeling assumptions are also discussed.