Fundus image registration for vestibularis research

In research on vestibular nerve disorders, fundus images of both left and right eyes are acquired systematically to precisely assess the rotation of the eye ball that is induced by the rotation of entire head. The measurement is still carried out manually. Although various methods have been proposed for medical image registration, robust detection of rotation especially in images with varied quality in terms of illumination, aberrations, blur and noise still is challenging. This paper evaluates registration algorithms operating on different levels of semantics: (i) data-based using Fourier transform and log polar maps; (ii) point-based using scaled image feature transform (SIFT); (iii) edge-based using Canny edge maps; (iv) object-based using matched filters for vessel detection; (v) scene-based detecting papilla and macula automatically and (vi) manually by two independent medical experts. For evaluation, a database of 22 patients is used, where each of left and right eye images is captured in upright head position and in lateral tilt of ±200. For 66 pairs of images (132 in total), the results are compared with ground truth, and the performance measures are tabulated. Best correctness of 89.3% were obtained using the pixel-based method and allowing 2.5° deviation from the manual measures. However, the evaluation shows that for applications in computer-aided diagnosis involving a large set of images with varied quality, like in vestibularis research, registration methods based on a single level of semantics are not sufficiently robust. A multi-level semantics approach will improve the results since failure occur on different images.

[1]  M. Gresty,et al.  Ocular tilt reaction with peripheral vestibular lesion , 1979, Annals of neurology.

[2]  Konstantina S. Nikita,et al.  Automatic retinal image registration scheme using global optimization techniques , 1999, IEEE Transactions on Information Technology in Biomedicine.

[3]  Matthijs C. Dorst Distinctive Image Features from Scale-Invariant Keypoints , 2011 .

[4]  Hanqing Lu,et al.  A Fast Mean Shift Procedure with New Iteration Strategy and Re-sampling , 2006, 2006 IEEE International Conference on Systems, Man and Cybernetics.

[5]  Robert C. Bolles,et al.  Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography , 1981, CACM.

[6]  Houjin Chen,et al.  A Registration Method of Fundus Images Based on Edge Detection and Phase-Correlation , 2006, First International Conference on Innovative Computing, Information and Control - Volume I (ICICIC'06).

[7]  Purang Abolmaesumi,et al.  Deformable registration using scale space keypoints , 2006, SPIE Medical Imaging.

[8]  George Wolberg,et al.  Robust image registration using log-polar transform , 2000, Proceedings 2000 International Conference on Image Processing (Cat. No.00CH37101).

[9]  P. Abolmaesumi,et al.  Medical image registration based on distinctive image features from scale-invariant (SIFT) key-points , 2005 .

[10]  Leiguang Gong,et al.  Focused atlas-based image registration for recognition , 2008, 2008 15th IEEE International Conference on Image Processing.

[11]  M. Dieterich,et al.  Cyclorotation of the Eyes and Subjective Visual Vertical in Vestibular Brain Stem Lesions a , 1992, Annals of the New York Academy of Sciences.

[12]  Juan Xu,et al.  Automated Optic Disk Boundary Detection by Modified Active Contour Model , 2007, IEEE Transactions on Biomedical Engineering.

[13]  Yulong Shen,et al.  Registration and fusion of retinal images-an evaluation study , 2003, IEEE Transactions on Medical Imaging.

[14]  B. N. Chatterji,et al.  An FFT-based technique for translation, rotation, and scale-invariant image registration , 1996, IEEE Trans. Image Process..

[15]  Kai-Kuang Ma,et al.  Fast object-based image registration using principal component analysis for super-resolution imaging , 2008 .

[16]  Kuo-Chin Fan,et al.  Image Registration Using a New Edge-Based Approach , 1997, Comput. Vis. Image Underst..

[17]  S G Diamond,et al.  Binocular Counterrolling during Sustained Body Tilt in Normal Humans and in a Patient with Unilateral Vestibular Nerve Section , 1982, The Annals of otology, rhinology, and laryngology.

[18]  M. Goldbaum,et al.  Detection of blood vessels in retinal images using two-dimensional matched filters. , 1989, IEEE transactions on medical imaging.