Possibilistic picture fuzzy product partition C-means clustering incorporating rich local information for medical image segmentation

Liu, Tairong
DOI: https://doi.org/10.1007/s11042-024-19080-4
IF: 2.577
2024-04-26
Multimedia Tools and Applications
Abstract:Picture fuzzy C-means clustering is a new computational intelligence method that has more significant potential advantages than fuzzy clustering in medical image interpretation. However, the practical application of picture fuzzy clustering in medical image segmentation is severely limited by its sensitivity to noise or outliers. Therefore, this paper proposes a medical image segmentation method called possibilistic picture fuzzy product partition C-means clustering with local information. This method combines picture fuzzy clustering with fuzzy possibilistic product partition C-means clustering to solve the segmentation problem of noisy medical image. Firstly, this paper extends picture fuzzy clustering to construct a possibilistic picture C-means clustering, enhancing the robustness of picture fuzzy clustering to noise or outliers. Secondly, by combining picture fuzzy clustering with possibilistic picture C-means clustering, a novel possibilistic picture fuzzy product partition C-means clustering is proposed, further enhancing the adaptability of picture fuzzy clustering in medical image analysis. Finally, a weighted squared Euclidean distance with complement spatial information and a novel possibilistic picture fuzzy local information factor are constructed, and they are introduced into the possibilistic picture fuzzy product partition clustering to enhance the robustness of this method in noisy image segmentation. The experimental results on medical images indicate that the proposed method not only has good segmentation performance and strong anti-noise robustness, but also improves segmentation accuracy by 1.66% to 17.39% compared with FRFCM method.
computer science, information systems, theory & methods,engineering, electrical & electronic, software engineering
What problem does this paper attempt to address?