Segmentation and Volumetric Analysis of Medical Images
Image segmentation refers to the decomposition of a scene into its components. In medical imaging, different tissues and organs can be recognized and segmented. For this purpose specific segmentation software Anatomatic [1] was developed. The software operates semi-automatically and utilizes the IARD [2] segmentation algorithm. The source data can be any 3D medical images, such as Magnetic Resonance (MR) images and Computed Tomography (CT) images. Because the physical dimensions of MR/CT images are known, it is possible to apply segmentation in volumetric analysis. The volumetric accuracy of Anatomatic and IARD algorithm was tested and the results suggested relatively high accuracy (98.5%) and reproducibility [3].
Anatomatic Segmentation Software
Anatomatic software was developed to PC/Windows95 environment using C++ language (and also to NeXTstep environment using Objective-C language). The software enables segmentation of arbitrary organs and structures and has been tested in numerous research projects.
Figure 1: The user interface of the Anatomatic software.
Pilot Projects
Anatomatic has been applied in volumetric analysis of brain infarcts, multiple sclerosis lesions (hyper- and hypo-intense plaques), cerebral atrophy (associated with MS, brain infarct, and vascular dementia patients), gynecological tumors, fetuses, nasal cavities and paranasal sinuses. It has also been utilized in the reconstruction of resistive head and thorax models. The IARD algorithm enables relatively accurate segmentation of the skull from T1-weighted MR images and can therefore be applied excellently in the generation of resistive head models. Some of the segmented structures are presented in Fig 2. The pilot projects proved Anatomatic to be efficient. The advantages of the software are functionality, intuitivity, and versatility.
Figure 2: (A) Segmented brain infarct, (B) Multiple sclerosis plaques, original slice of the Visible Human Man cryosection (C) together with the segmented slice (D), Segmented MR images of the head (EF) together with 3D presentations (G-J).
REFERENCES
[1] Heinonen, T., Dastidar, P., Kauppinen, P., Malmivuo, J. & Eskola, H. 1998. Semi-automatic Tool for Segmentation and Volumetric Analysis of Medical Images. Medical & Biological Engineering & Computing 36, 3, p. 291-296.
[2] Heinonen, T., Eskola, H., Dastidar, P., Laarne, P. & Malmivuo, J. 1997. Segmentation of T1 MR Scans for Reconstruction of Resistive Head Models. Computer Methods and Programs in Biomedicine 54, Ireland. Elsevier Science Ltd.. p. 173-181.
[3] Heinonen T, Dastidar P, Eskola H, Frey H, Ryymin P, Laasonen E. 1998. Applicability of semi-automatic segmentation for volumetric analysis of brain lesions, Journal of Medical Engineering & Technology, 33(4): 173-178.
For further information, please contact:
Tomi Heinonen, Tampere University of Technology, P.O.Box 692, FIN-33101 TAMPERE, Finland
Email: tomi@ee.tut.fi, fax: +358 3 247 4013, tel: +358 40 547 1451