Biomedical image segmentation using fuzzy multilevel soft thresholding system coupled modified cuckoo search

The automated computer-aided biomedical image analysis tools help in achieving precise and accurate analysis of disease with less manual intervention and facilitate quick and accurate treatment. Computer vision and machine learning are two important technologies used frequently as a tool for automated biomedical image analysis. Automated segmentation of digital images is always challenging and has different applications in diagnosis procedures. This work is focused to address this challenge by a hybrid approach that takes the advantage of the modified cuckoo search approach and fuzzy system. This combined approach is applied to determine the multiple threshold values by optimizing different objective functions separately. The proposed approach is evaluated by using both qualitative and quantitative approaches. Standard evaluation metrics like MSE, PSNR, SD, Mean, SSIM, and running time quantify the outcome. Average quantitative outcomes are tabulated and compared with some standard approaches for a different number of clusters and three objective functions separately. It is observed that on most occasions, the proposed approach outperforms its competitors and achieves significant improvements. On average, the proposed approach achieves 0.8076, 0.5361, 0.7155, and 0.6594 values for the SSIM by optimizing the fuzzy Tsallis entropy for 3, 5, 7, and 9 clusters respectively. These encouraging results motivate deploying the proposed approach in real-life scenarios.

Automated summary

The study focuses on addressing the challenge of automated segmentation of digital images using a hybrid approach that combines the modified cuckoo search approach and fuzzy system. The proposed method is evaluated using both qualitative and quantitative approaches and outperforms competitors, achieving significant improvements. The results show that the proposed approach achieves higher SSIM values for different numbers of clusters by optimizing the fuzzy Tsallis entropy, motivating its deployment in real-life scenarios.