MRG Veri Taban?nda U-Net ile Otomatik Kafatas? Ç?kartma ve Beyin Segmentasyonu

Skull stripping has an important in neuroimaging workflow. is a time-consuming process the Magnetic resonance imaging (MRI). For this reason, skull and brain segmentation are aimed study. purpose, U-NET architecture design, which one of frequently used models field medical image segmentation, was used. Also, different loss functions such as Cross Entropy (CE), Dice, IoU, Tversky, Focal Tversky their compound forms were tested on U-Net design. The function CE Dice achieved best performace with average dice score 0.976, IoU 0.964, sensitivity 0.972, specificity 0.985, precision 0.960 accuracy 0.981. As result, performed to facilitate detection diseases.