Institutional Repository [SANDBOX]
Technical University of Crete
EN  |  EL

Search

Browse

My Space

Recursive-mode K-means clustering for self-organization of dynamic imaging data

Vourlaki Ioanna-Theoni, Livanos Georgios, Giakoumakis Theodoros-Marios, Zervakis Michail, Giakos George C., Balas Costas

Full record


URI: http://purl.tuc.gr/dl/dias/E14EAFBC-9659-4279-B0EC-B19B2DEF118A
Year 2016
Type of Item Conference Full Paper
License
Details
Bibliographic Citation I. Vourlaki, G. Livanos, T. Giakoumakis, M. Zervakis, G. Giakos and C. Balas, "Recursive-mode K-means clustering for self-organization of dynamic imaging data," in IEEE International Conference on Imaging Systems and Techniques, 2016, pp. 54-59. doi: 10.1109/IST.2016.7738197 https://doi.org/10.1109/IST.2016.7738197
Appears in Collections

Summary

The aim of this study was to develop a novel algorithmic scheme for self-organizing data, adopting an recursive-mode k-means clustering approach. The proposed methodology attempts to refine and improve the clustering result by sequentially updating centers on the basis of their present and previous positions, exploiting both prior expert knowledge and posterior data information from the statistical distribution of the examined population. The performance of the implemented algorithm is evaluated on Dynamic Imaging data from cervical tissue, examining numerous sample curves representing the temporal response of tissue areas under the aceto-whitening effect. In comparison to the performance of classical k-means approach, the preliminary results of this study indicate the outperformance of the proposed iterative scheme. The primary benefit is attributed to the center improvement strategy against the center replacement methodology enforced in the classical approach. The performance and conceptual integration of knowledge justify the proposed update strategy as an efficient data grouping and classification tool, revealing a proposing potential to tissue evaluation and disease characterization applications.

Services

Statistics