Intelligent System Based on GAN Model for Decision Support in Brain Tumor Segmentation
Résumé
The most prevalent malignant brain tumors are gliomas, with a variety of grades, and each grade has a significant impact on a patient's chances of survival. Low-grade gliomas are usually found in the human brain and spinal cord. Low-grade glioma may be accurately diagnosed and detected early, lowering the risk of mortality for patients. In the examination gliomas of low grade, segmentation of MRI images is critical. The result, manual of Segmentation Techniques takes a long time and require a lot of pathology knowledge. in our study, we provide a unique generative adversarial network-based approach for segmenting images of tumors in the brain. The network is a structure between two neurons the generator and the discriminator. The generator is taught to construct an input mask of a take original image, The discriminator can tell the difference between the original and created masks, the end goal is to create masks for the input. The suggested model achieves a dice result of 0.97 in generalized experimental results from the TCGA LGG dataset, with a loss coefficient of 0.030, which is more effective and efficient than the compared approaches.