The authors have been developing a surgical navigation system for the graft replacement of thoracoabdominal aortic aneurysm and applied clinically 30 times since July 2006. This system supports surgeons to figure out the patient's intercostal level before thoracotomy. However, in the current navigation system, there is a problem that patient's postures are different between preoperative CT scanned data and intraoperative real data. In this paper, it is aimed to improve current navigation system to follow global deformation arising from difference of patient's postures. And the mathematical method is discussed that matches image space with real space globally by correcting the image. It is assumed that the deformation of thoracic cage is affected by the weight of trunk, and that of lumbar is affected by the torsion. To match both spaces, the algorithm that makes image slanted at the thoracic region and rotated at the lumbar region was developed. The angle of inclination for slanting and that of torsion for rotating were calculated from body weight, width of costale and elastic coefficient of bone. This algorithm was evaluated by using MRI images of two volunteers those were scanned in dorsal position and right decubitus position. As a result, fitting between image space and real space decreased by 20% by operating this algorithm. It was concluded that this algorithm is effective to match both spaces. Foreseeable future, this algorithm will use in navigation used in clinical.