A Hybrid Framework for Brain Tumor Segmentation: Integrating Cellular Neural Networks with Firefly Optimization

Early and accurate diagnosis of brain tumors is critical for effective treatment planning and improving patient prognosis. Magnetic Resonance Imaging (MRI) is a cornerstone in this process, providing detailed anatomical information of soft tissues such as white matter, grey matter, and cerebrospinal fluid, which is essential for localizing pathologies. This paper proposes a novel automated framework for the detection and segmentation of brain tumors from MRI scans. The methodology comprises four key stages: (۱) preprocessing for noise reduction and image equalization; (۲) feature extraction using morphological operators to identify edges and edema; (۳) initial tumor localization with a cellular neural network; and (۴) segmentation optimization using the Firefly Algorithm (FA) to precisely delineate the tumor boundaries. Evaluated on the benchmark BraTS dataset, the proposed framework, termed BTD-CFA, demonstrates superior performance, achieving an accuracy of ۹۶.۳۰% and a sensitivity of ۹۰.۵۱%. These results significantly outperform several baseline techniques and underscore the potential of bio-inspired optimization algorithms in enhancing the accuracy and reliability of computer-aided diagnosis systems for early and precise brain tumor analysis.

کلیدواژه ها:

Brain Tumor ، MRI ، Firefly Algorithm ، Cellular Neural Network ، Image Segmentation