Against the background of increasing numbers of indications for Cochlea implants (CIs), there is an increasing need for a CI outcome prediction tool to assist the process of deciding on the best possible treatment solution for each individual patient prior to intervention. The hearing outcome depends on several features in cochlear structure, the influence of which is not entirely known as yet. In preparation for surgical planning a preoperative CT scan is recorded. The overall goal is the feature extraction and prediction of the hearing outcome only based on this conventional CT data. Therefore, the aim of our research work for this paper is the preprocessing of the conventional CT data and a following segmentation of the human cochlea. The great challenge is the very small size of the cochlea in combination with a fairly bad resolution. For a better distinction between cochlea and surrounding tissue, the data has to be rotated in a way the typical cochlea shape is observable. Afterwards, a segmentation can be performed which enables a feature detection. We can show the effectiveness of our method compared to results in literature which were based on CT data with a much higher resolution. A further study with a much larger amount of data is planned.