Author(s) :

A. M. Asbelshiny, M.Nabeela

Conference Name :

International Conference on Modern Trends in Engineering and Management (ICMTEM-25)

Abstract :

Liver fibrosis, a precursor to cirrhosis and liver cancer, requires early and accurate classification to improve patient outcomes. Traditional image analysis techniques and conventional deep learning models like U Net face challenges in capturing fine grained features essential for fibrosis staging. To address these limitations, this study introduces an advanced deep learning architecture Attention Enhanced Residual Network (Attention ResNet) designed specifically for liver fibrosis classification. The proposed model integrates improved residual modules within both encoder and decoder paths to boost feature extraction and ensure stable gradient flow. A Kernal based Fuzzy C Means (KFCM) is incorporated before skip connections to enrich segmentation through simultaneous channel and spatial attention. Additionally, a Boundary Descriptors layer at the bottleneck extracts contextual information across scales, enhancing the model’s representation power of feature extraction. Evaluation on medical imaging datasets done by Python shows that AER Net delivers superior classification specificity and accuracy of 93% and accuracy of 93% respectively.

No. of Downloads :

1