•  
  •  
 

Corresponding Author

Karthick Palanisamy

Subject Area

Biomedical Engineering

Article Type

Special Issue Original Study

Abstract

The process of accurately segmenting Urolithiasis and pulmonary nodules from medical images presents significant difficulties because of three main factors which include nonuniform intensity distribution and the presence of weak object borders and the uncertainty caused by anatomical structure overlaps. The study introduces F-Zinhger as a solution to these problems through its application of the Forest of Walks technology. The method uses superpixels as fundamental elements which form a multi-scale hypergraph that establishes hyperedges through three criteria: intensity similarity, spatial proximity, and Z-intuitionistic fuzzy hesitation to capture complex contextual connections. The system incorporates a random walk-based label propagation system. The proposed framework was evaluated on two benchmark medical imaging datasets which included a renal stone dataset and a pulmonary nodule dataset and it was tested against current advanced hypergraph-based and fuzzy clustering techniques. The proposed method for Urolithiasis segmentation achieved a Dice coefficient of 0.8862, Jaccard index of 0.7987, precision of 0.8641, recall of 0.9155, and accuracy of 0.9884. The method for pulmonary nodule segmentation produced a Dice coefficient of 0.9518, Jaccard index of 0.9080, precision of 0.9193, recall of 0.9267, and accuracy of 0.9582. F-Zinhger achieves better results than other methods which compete with it because it shows higher performance across both datasets while using less processing power. The method shows its capability to produce accurate segmentation results through two different medical tasks. Future work will focus on extending the F-Zinhger framework to 3D volumetric medical imaging and validating its clinical applicability across larger, multi-institutional datasets.

Keywords

Tumor diagnosis, Medical applications, Urolithiasis Segmentation, Fuzzy Hypergraph cut, Pulmonary Nodule Segmentation, Image Processing.

Creative Commons License

Creative Commons Attribution 4.0 License
This work is licensed under a Creative Commons Attribution 4.0 License.

Download

Share

COinS