Özet:

Abstract In the technological world, health care systems play essential role in extending the human age through early detection. The grooming of Artificial Intelligence technology and deep learning enlighten many smart surgical devices and aid knowledge in disease diagnosis and planning. In particular, the brain disease diagnosis process embraces several devices, such as Magnetic Resonance (MR) machines etc. and software. Most of the brain disease diagnosis includes brain extraction (skull stripping) as the pre-processing step. Brain portion extraction from its non-brain tissues from MR image is a tedious process and takes 40 to 60 minutes per patient in manual process and the earlier model lacked accuracy. In the age of deep learning, accurate brain tissue detection using deep learning is essential and accounts for good results. This paper proposes a modified Unit for skull stripping from 2D MR images and comparing it with other deformable and deep learning models such as BET2, RoBEX and UNet3D. The proposed model provides good competency results, such as a 98 % Dice score and 1% higher than the UNet3D model.

Anahtar Kelimeler: