Automated Sickle Cell Identification in Digital Blood Smears Using FSEF-MobileNet with Wolf Bird Optimization Algorithm

Resumo

Sickle Cell Disease (SCD), a genetic blood disorder characterised by abnormal red blood cell shape, is sometimes difficult, time-consuming, and susceptible to individual interpretation when inspected under a microscope. To get over these limitations, this study proposes an automated computer-aided diagnostic approach for sickle cell diagnosis from digital blood smear images. The technique begins with preprocessing, which involves resizing and improving images using Adaptive Contrast Limited Adaptive Histogram Equalization (CLAHE) in order to increase the visibility of cellular structures. Red blood cells sometimes be accurately separated from complex smear backgrounds through segmentation using Hierarchical K-Means Clustering. Next, Multi-Scale Gray Level Co-occurrence Matrix (MSGLCM) features are obtained to capture discriminative textural and morphological patterns across multiple scales. The enhanced features are classified using a special Focal-SE-ECA-Fused MobileNet (FSEF-MobileNet) architecture. This architecture reduces class imbalance and highlights useful features by combining Efficient Channel Attention (ECA) and Squeeze-and-Excitation (SE) techniques with focal loss. Wolf Bird Optimization (WBO), which ensures reduced redundancy and enhanced separation between normal and sickle cells to enhance classification reliability, is used to tune hyperparameter. The proposed approach is analyzed using Python, and the findings indicate that it has a great deal of promise for providing prompt and precise sickle cell disease assistance and achieves an accuracy, precision, recall and FI-score of 97%.