Jaruenpunyasak Jermphiphut, Duangsoithong Rakkrit, Tunthanathip Thara
Department of Biomedical Sciences and Biomedical Engineering, Faculty of Medicine, Prince of Songkla University Songkhla, Songkhla, Thailand.
Department of Electrical Engineering, Faculty of Engineering, Prince of Songkla University, Songkhla, Thailand.
J Neurosci Rural Pract. 2023 Jul-Sep;14(3):470-476. doi: 10.25259/JNRP_50_2022. Epub 2023 Jun 15.
It can be challenging in some situations to distinguish primary central nervous system lymphoma (PCNSL) from glioblastoma (GBM) based on magnetic resonance imaging (MRI) scans, especially those involving the corpus callosum. The objective of this study was to assess the diagnostic performance of deep learning (DL) models between PCNSLs and GBMs in corpus callosal tumors.
The axial T1-weighted gadolinium-enhanced MRI scans of 274 individuals with pathologically confirmed PCNSL ( = 94) and GBM ( = 180) were examined. After image pooling, pre-operative MRI scans were randomly split with an 80/20 procedure into a training dataset ( = 709) and a testing dataset ( = 177) for DL model development. Therefore, the DL model was deployed as a web application and validated with the unseen images ( = 114) and area under the receiver operating characteristic curve (AUC); other outcomes were calculated to assess the discrimination performance.
The first baseline DL model had an AUC of 0.77 for PCNSL when evaluated with unseen images. The 2 model with ridge regression regularization and the 3 model with drop-out regularization increased an AUC of 0.83 and 0.84. In addition, the last model with data augmentation yielded an AUC of 0.57.
DL with regularization may provide useful diagnostic information to help doctors distinguish PCNSL from GBM.
在某些情况下，基于磁共振成像（MRI）扫描区分原发性中枢神经系统淋巴瘤（PCNSL）和胶质母细胞瘤（GBM）可能具有挑战性，尤其是那些累及胼胝体的病变。本研究的目的是评估深度学习（DL）模型在胼胝体肿瘤中区分PCNSL和GBM的诊断性能。
对274例经病理证实为PCNSL（n = 94）和GBM（n = 180）患者的轴位T1加权钆增强MRI扫描进行检查。图像汇总后，术前MRI扫描按照80/20的比例随机分为训练数据集（n = 709）和测试数据集（n = 177），用于开发DL模型。因此，DL模型作为一个网络应用程序进行部署，并用未见过的图像（n = 114）和受试者操作特征曲线下面积（AUC）进行验证；计算其他结果以评估判别性能。
当用未见过的图像进行评估时，第一个基线DL模型对PCNSL的AUC为0.77。采用岭回归正则化的2模型和采用随机失活正则化的3模型使AUC分别提高到0.83和0.84。此外，最后一个采用数据增强的模型的AUC为0.57。
具有正则化的DL可能提供有用的诊断信息，以帮助医生区分PCNSL和GBM。
