Minimally invasive partial nephrectomy surgery is a common procedure required to remove kidney tumors and to treat a variety of renal malfunctions. It presents many challenges to the surgeon due to the kidney anatomy and the limited sight of the minimal invasive technique. Preoperative planning based on Computed Tomography (CT) scans is a crucial step, since it can reduce the risk and duration of the procedures. We have developed a semi-automatic method for the creation of patient-specific kidney models from four-phase CT studies. The kidney models include the kidney outer contour, kidney arteries, veins, urine collecting system and tumors when present. The segmentation methods rely on the mutual intensity distribution between the CT scan phases, region-growing, and morphological operations. The models are incorporated into a preoperative planning system that allows the surgeon to visualize the kidney and its internal components and to position a resection plane. A validation study on three four-phase CT sequences comparing automatic vs. manually segmented models of the kidney and its internal components on three patients yields a mean overlap error 1.7% for the kidney contour, 14.1% for the ureter, and 25.6% for the blood vessels.