A Hybrid Deep Learning Framework for Liver Cancer Detection using Capsule Networks, Vision Transformers, Graph Neural Networks, and Quantum-Inspired Optimization

Liver cancer is one of the major causes of death in
most parts of the world; this is mainly caused by late-stage
diagnosis and complicated morphology of tumor. In this paper,
a hybrid framework based on Python is suggested with
combining Capsule Networks (CapsNet), Graph Neural
Networks (GNN), Vision Transformer (ViT), and a Quantuminspired Evolutionary Algorithm (QEA) to detect liver cancer
early and accurately. Actually, prepared medical imaging data
such as CT and MRI images are processed by adaptive
histogram equalization and noise reduction to improve quality.
CapsNet preserves hierarchical spatial variants and ViT learns
long range relationships, GNN depicts structural relationships
in visuals in the form of graphs. QEA identifies the best feature
sets, enhancing the accuracy of classification and decreasing the
computation costs. The results of the experiments have been
shown to be highly robust and scaled, with better performance
than traditional CNN-based techniques. The method has great
potential of real-time clinical decision support, augmenting the
earlier diagnosis and tailor-made treatment plans that will be
applied in the liver cancer treatment.