This paper presents a 3D image processing method that is based on the analysis of Hessian matrix eigenvalues combined with a multiscale image analysis approach. The method, originally developed for blood vessels detection in medical images, can also be used in other areas, where finding line-like structures in the image is required. Theoretical background, advantages and disadvantages of the method are described. Possible modifications required to allow the method to detect structures of different character (airway tree) are mentioned. An implementation of the method was tested on synthetic images containing airway-like structures as well as on real medical images from chest CT scan. Results show that the method in general can be used to airway detection in 3D medical images, however it requires improvements and some adaptation to this specific purpose.