An automated cascade framework for Glioma prognosis via segmentation, multi-feature fusion and classification techniques

Glioma is the most lethal type of brain tumor, comprising about 33% of diagnosed cases. Segmentation and classification are crucial for precise glioma characterization, emphasizing early detection of malignancy for effective treatment and to stop tumor growth. MRI is a non-invasive technique for examining gliomas without ionizing radiation. Manual diagnosis is impractical and requires the expertise of radiologists. Hence, computer-aided diagnosis (CAD) systems have greatly evolved to assist neuro-oncologists in the screening process of brain cancer. Our glioma classification strategy is based on 3D multi-model MRI segmentation using the CNN models SegResNet and Swin UnetR, with a transformer mechanism for better segmentation. The MRI images were pre-processed using a Gaussian filter and brain skull stripping. Key textural features are extracted for the classification of Low Grade Glioma (LGG) or High-Grade Glioma (HGG) by SVM. Experiments on benchmarks such as BRATS2020 and BRATS2023 prove that our proposed Swin UnetR-based framework performs very well in grading brain tumors as LGG or HGG. The proposed framework has the advantage of providing strong support to radiologists in early detection and classification of gliomas. .