BicephNet: A Hybrid CNN–ViT–SE Architecture with FFT-Head and Bidirectional Cross-Attention for Hierarchical Multi-Disease and Multi-Stage Brain MRI Classification with Web Application
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Accurate diagnosis of neurological abnormalities such as Alzheimer’s disease (AD), Parkinson’s disease (PD), and brain tumors remains a major obstacle because of significant inter-patient variability and overlapping structural patterns. The majority of current architectures take a limited, single-disease approach and fall short of fully utilizing the complementary nature of spatial, contextual, and frequency-domain information, despite the fact that deep learning techniques have shown great promise in medical image analysis. In order to overcome these constraints, this paper presents BicephNet, a novel hierarchical MRI diagnostic architecture that combines spectral FFT-head for frequency-aware learning, Squeeze-and-Excitation (SE) channel re-weighting, Vision Transformer (ViT)-based global attention modeling, convolutional spatial encoding, and a bidirectional cross-attention mechanism for multi-branch feature fusion.The suggested framework first determines the disease category at Level-1, then performs a more detailed Level-2 subclassification in order to mimic clinical diagnostic workflows including Alzheimer’s disease stages and tumor subtype identification.
Fields of Study & Subjects
Brain Tumor Detection and Classification
Medical Image Segmentation Techniques
Dementia and Cognitive Impairment Research
Publication details
- Journal / source
- International Conference on Computing, Communication, Control and Cyber-Physical Systems (I5CPS)
- Pages
- 1–6
- Publication date
- 26 February 2026
- Language
- EN