Enhancing Multi-Modal MRI Brain Tumor Detection using Hybrid Convolutional Based Deep Learning Model
Fragment książki (Rozdział monografii pokonferencyjnej)
MNiSW
20
Poziom I
| Status: | |
| Autorzy: | Ranpara Nirav, Lathigara Amit, Shingadiya Chetan, Dzierżak Róża |
| Dyscypliny: | |
| Aby zobaczyć szczegóły należy się zalogować. | |
| Wersja dokumentu: | Drukowana | Elektroniczna |
| Język: | angielski |
| Strony: | 1 - 5 |
| Efekt badań statutowych | NIE |
| Materiał konferencyjny: | TAK |
| Nazwa konferencji: | International Conference on Computational Intelligence, Emerging Technologies, and Smart Systems 2025 |
| Skrócona nazwa konferencji: | ICCIETSS 2025 |
| URL serii konferencji: | LINK |
| Termin konferencji: | 28 marca 2025 do 29 marca 2025 |
| Miasto konferencji: | Rajkot |
| Państwo konferencji: | INDIE |
| Publikacja OA: | NIE |
| Abstrakty: | angielski |
| A critical yet difficult job in medical imaging is the detection and segmentation of brain cancers from multi-modal magnetic resonance imaging (MRI), primarily because brain tumors are complicated and heterogeneous. Clinical practice is greatly limited by the time-consuming, heavily reliant on radiologist expertise, and inter-observer variability of traditional manual segmentation procedures. The low generalizability of models across varied datasets, the challenge of segmenting small tumor sub-regions, and the underutilization of multi-modal MRI data are some of the major problems that still exist despite recent advances in deep learning showing promise in automating brain tumor segmentation. In order to overcome these inadequacies, this study presents a new deep learning framework that maximizes segmentation accuracy and model robustness by merging a transformer-based architecture with a hybrid Convolutional neural network (CNN). The recommended model works remarkably well in segmenting tumor locations across a range of MRI sequences, obtaining a significant accuracy of 98.85% and a low loss of 0.02. This work’s key innovation is its capacity to use hybrid deep learning architectures to harness multi-modal MRI data. This is a considerable improvement over standard CNN-based models, which frequently fall short of properly exploiting the rich, complimentary information supplied by diverse imaging modalitie |
