Segmentation of Meningeal Contrast Enhancement in Post-Contrast T1-Weighted Images Using the Deep Learning Method

Authors : Nevin Aydın, Suzan Şaylısoy, Uğur Toprak, Burcu Mert, Özer Çelik

Pages : 600-605

Doi:10.20515/otd.1641306

View : 43 | Download : 9

Publication Date : 2025-06-18

Article Type : Research Paper

Abstract :To evaluate the success of segmentation of meningeal contrast enhancement on post-contrast T1-weighted images using the deep learning method. The study retrospectively included 313 sections obtained from post-contrast T1-weighted sequences of 83 patients with meningeal enhancement. The dataset was divided into three groups. A total of 300 epochs of training were performed using PyTorch U-Net, and the best model was identified. The results were calculated by selecting 50% as the threshold for the intersection over union statistics. In total, images of 83 patients were evaluated, of whom 36 (43.4%) were female and 47 (56.6%) were male. The mean ± standard deviation of the patients’ age was 57.06 ± 16.73 years. Of the 313 sections obtained, 251 were allocated in the training group, 31 to the validation group, and 31 to the test group. The results of the test group were as follows: 35 true positives, 12 false positives, and 12 false negatives. The precision, sensitivity, and F1 score values were all calculated to be 74%. This is one of the pioneering studies in the literature on the segmentation of meningeal contrast-enhanced areas using the deep learning-based U-net architecture. Further studies are needed in this area
Keywords : Derin Öğrenme, Meningeal Kontrastlanma, , Pakimeningeal Kontrastlanma, Leptomeningeal Kontrastlanma, Manyetik Rezonans Görüntüleme

ORIGINAL ARTICLE URL