The cochlea, the auditory part of the inner ear, is a spiral-shaped organ with large morphological variability. An individualizedassessment of its shape is essential for clinical applications related totonotopy and cochlear implantation. To unambiguously reference mor-phological parameters, reliable recognition of the cochlear modiolar axisin computed tomography (CT) images is required. The conventionalmethod introduces measurement uncertainties, as it is based on man-ually selected and difficult to identify landmarks. Herein, we present analgorithm for robust modiolar axis detection in clinical CT images. Wedefine the modiolar axis as the rotation component of the kinematic spi-ral motion inherent in the cochlear shape. For surface fitting, we use acompact shape representation in a 7-dimensional kinematic parameterspace based on extended Pl ̈ucker coordinates. It is the first time such akinematic representation is used for shape analysis in medical images.Robust surface fitting is achieved with an adapted approximate maxi-mum likelihood method assuming a Student-t distribution, enabling axisdetection even in partially available surface data. We verify the algo-rithm performance on a synthetic data set with cochlear surface subsets.In addition, we perform an experimental study with four experts in 23human cochlea CT data sets to compare the automated detection withthe manually found axes. Axes found from co-registered high resolutionμCT scans are used for reference. Our experiments show that the algo-rithm reduces the alignment error providing more reliable modiolar axisdetection for clinical and research applications.