TY - JOUR N2 - The paper is focused on automatic segmentation task of bone structures out of CT data series of pelvic region. The authors trained and compared four different models of deep neural networks (FCN, PSPNet, U-net and Segnet) to perform the segmentation task of three following classes: background, patient outline and bones. The mean and class-wise Intersection over Union (IoU), Dice coefficient and pixel accuracy measures were evaluated for each network outcome. In the initial phase all of the networks were trained for 10 epochs. The most exact segmentation results were obtained with the use of U-net model, with mean IoU value equal to 93.2%. The results where further outperformed with the U-net model modification with ResNet50 model used as the encoder, trained by 30 epochs, which obtained following result: mIoU measure – 96.92%, “bone” class IoU – 92.87%, mDice coefficient – 98.41%, mDice coefficient for “bone” – 96.31%, mAccuracy – 99.85% and Accuracy for “bone” class – 99.92%. L1 - http://www.czasopisma.pan.pl/Content/119435/PDF/04_01998_Bpast.No.69(3)_23.06.21_Druk.pdf L2 - http://www.czasopisma.pan.pl/Content/119435 PY - 2021 IS - 3 EP - e136751 DO - 10.24425/bpasts.2021.136751 KW - deep learning KW - semantic segmentation KW - U-net KW - FCN KW - ResNet KW - computed tomography A1 - Krawczyk, Zuzanna A1 - Starzyński, Jacek VL - 69 DA - 10.03.2021 T1 - Segmentation of bone structures with the use of deep learning techniques SP - e136751 UR - http://www.czasopisma.pan.pl/dlibra/publication/edition/119435 T2 - Bulletin of the Polish Academy of Sciences Technical Sciences ER -