A New Hybrid Technique for Dermatological Image Registration

Malignant melanoma is nowadays one of the most malignant tumors among white-skinned populations around the world. The key element in malignant melanoma treatment is the detection of melanomas and their changes at an early stage, before they develop irreversible clinically significant and potentially fatal damage to the patients. Computer automatic diagnosis of skin lesions using early symptoms would be particularly useful as an aid in primary care. During such computer systems, there is a significant clinical demand for accurate dermatological image registration. In this paper, we introduce a new algorithm for the registration of melanomas in successive dermatological image. We reduce the melanoma registration problem to a bipartite graph matching problem. The Voronoi cells are used to measure the similarity between melanomas and build the weighted bipartite graph. A minimum weight maximum cardinality matching is employed to find the global correspondences between dermatological images. Distances order and dynamic programming method are applied into the bipartite graph matching to preserve topology of melanoma distribution. The dermatoscopy images are used to validate the effectiveness of our approach. Since our method is a general registration method for melanomas, it can also be used in other dermatological images.

[1]  D A Perednia,et al.  Automatic derivation of initial match points for paired digital images of skin. , 1992, Computerized medical imaging and graphics : the official journal of the Computerized Medical Imaging Society.

[2]  Franjo Pernus,et al.  Segmentation of muscle fibre images using Voronoi diagrams and active contour models , 1996, Proceedings of 13th International Conference on Pattern Recognition.

[3]  Jim Graham,et al.  2-D electrophoresis gel registration using feature matching , 2004, 2004 2nd IEEE International Symposium on Biomedical Imaging: Nano to Macro (IEEE Cat No. 04EX821).

[4]  Jean-François Sadoc,et al.  Voro3D: 3D Voronoi tessellations applied to protein structures , 2005, Bioinform..

[5]  S. Pavlopoulos,et al.  New hybrid stochastic-deterministic technique for fast registration of dermatological images , 2004, Medical and Biological Engineering and Computing.

[6]  Bruce Mcgregor,et al.  Automatic registration of images of pigmented skin lesions , 1998, Pattern Recognit..

[7]  Riccardo Bono,et al.  Early diagnosis of malignant melanoma: Proposal of a working formulation for the management of cutaneous pigmented lesions from the Melanoma Cooperative Group. , 2003, International journal of oncology.

[8]  R. Braun,et al.  Dermoscopy of pigmented lesions: a valuable tool in the diagnosis of melanoma. , 2004, Swiss medical weekly.

[9]  F. Makedon,et al.  A bipartite graph matching framework for finding correspondences between structural elements in two proteins , 2004, The 26th Annual International Conference of the IEEE Engineering in Medicine and Biology Society.

[10]  Juha Röning,et al.  Registration of nevi in successive skin images for early detection of melanoma , 1998, Proceedings. Fourteenth International Conference on Pattern Recognition (Cat. No.98EX170).

[11]  Suchendra M. Bhandarkar,et al.  NOVEL GRAPH THEORETIC ENHANCEMENTS TO ICP-BASED VIRTUAL CRANIOFACIAL RECONSTRUCTION , 2007, 2007 4th IEEE International Symposium on Biomedical Imaging: From Nano to Macro.

[12]  David S. Johnson,et al.  Network Flows and Matching: First DIMACS Implementation Challenge , 1993 .

[13]  D A Perednia,et al.  Automatic registration of multiple skin lesions by use of point pattern matching. , 1992, Computerized medical imaging and graphics : the official journal of the Computerized Medical Imaging Society.

[14]  H. Kuhn.  The Hungarian method for the assignment problem , 1955 .