An Enhanced Preprocessed Techniques Based 3-Level Mean Based Enhanced Histogram Equalization with A Self-Attention Segnet Based Segmentation and Fuzzy Aware Classification

Purpose: The accurate detection and classification of bone fractures in medical imaging remain significant challenges in the field of healthcare. Existing techniques often fall short of effectively identifying bone fractures due to suboptimal preprocessing methods. Moreover, the purpose of this study is to enhance the detection and classification accuracy.Methods: We introduce a novel approach incorporating a Three Level Multi-Model (TLMM)based histogram equalization to improve the accuracy. This approach encompasses Normalized Gamma-Corrected Contrast-Limited Adaptive Histogram Equalization (NGC-CLAHE) with Tanzanian Devil Optimization (TDO), CLAHE with Self-Adaptive-TDO (SA-TDO), and Adaptive Weighted Brightness Preserving Dynamic Fuzzy Histogram Equalization (AW-BPDFHE) with Giant-Devil Optimization (GDO), derived from the Giant Trevally Optimizer (GTO) and Tanzanian Devil Optimization (TDO) algorithms. Once Pre-processing is complete, the refined dataset is subjected to segmentation. For the segmentation, the SegNet model is used. Then the Gray-Level Co-occurrence Matrix (GLCM) model is used for extracting the optimal features. The obtained features are then directed to the classification layer, where a Fuzzy Radial Basis Function network (FRBF) efficiently classifies the bone fracture regions. Result: The performance of the proposed design can be validated by calculating different metrics. Moreover, the performance of the proposed model is compared with the existing models such as CNN, SVM, DT and RN. Conclusion: This integrated approach demonstrates promising results in improving the identification and classification of bone fractures in medical images, offering potential benefits to the field of medical diagnosis and patient care. On analyzing the performance of the proposed design as well as the existing model, the proposed model provides better detection and classification accuracy.