?skemik ?nme Lezyon Segmentasyonunda Segmentasyon A?lar?n?n ve Kay?p Fonksiyonlar?n?n Etkilerinin Analizi

Stroke was the cause of one out every six deaths from cerebrovascular disease in 2020. A stroke occurs United States (US) 40 seconds. Every 3.5 minutes, people die a stroke. More than total 795,000 cases occur yearly US. This study aims to detect ischemic lesion that brain. The Ischemic Lesion Segmentation (ISLES) 2017 data set, which includes 82 Magnetic Resonance images 43 patients, used. UNet, Attention Residual and UNet++ segmentation networks were tested. Moreover, Cross Entropy, Dice, IoU, Tversky, Focal their compound forms analyzed. IoU loss function tested on UNet achieved best performance with dice score 0.766, 0.621, sensitivity 0.730, specificity 0.997, precision 0.805, accuracy 0.993.