Quantitative assessment of disease progression in multiplesclerosis (MS) using MRI is an important issue for therapeuticmonitoring and understanding the development ofdisability. In particular, the relationship between spinalcord atrophy and development of disability has been thesubject of recent interest (5,7,13,16,17) and a correlationbetween atrophy and disability has been demonstrated.Although imaging techniques have improved signifi-cantly and provide convenient images to study the spinalcord, reliable and reproducible postprocessing and imageanalysis techniques to measure spinal cord atrophy arestill limited. Atrophy is mostly assessed by measuringcross-sectional area at specific levels, typically C2 and C5(5,17), along the cervical cord. This creates several problems,including the choice of the level at which the measurementsare performed, the cord orientation, and thecord segmentation process. To provide nonbiased areameasurements, the image, or part of the image, needs to bereformatted (17), or acquired (5), with slices perpendicularto the cord. The concept of “perpendicular to the cord” issomewhat imprecise and operator-dependent for a structurewhich is not a strict cylinder or a surface of revolution.It has not been formally defined in the above-mentionedpublications. Moreover, spinal cord cross-sectionalarea has often been measured either manually or using intensity-based 2D processing techniques. The limitationsof such methods are various: because the orientation of thecord changes along its axis, measurements are restricted toa predefined level at which cord and slices are orthogonal;intensity-based segmentation is hindered by the signifi-cant intensity variations caused by surface coils typicallyused during acquisition; 2D measurements are more proneto being biased by partial volume effect than 3D measurements;manual analysis are more time-consuming andmore sensitive to intra- and interoperator variability.We present here a method that aims to solve those problems.The method is based on a 3D extraction of the cervicalcord surface and the computation of a medial axis,used to define orthogonal cross-sections anywhere alongthe cord. The segmentation method is semiautomated withvery few interventions by the operator, based on the optimizationof a parameterized active surface. Cross-sectionalarea measurements can be provided globally as well aslocally at any point along the cord. We show that themethod provides accurate and reproducible measurements.