Academy for Engineering and Technology, Fudan University, Shanghai, China.
Suzhou Institute of Biomedical Engineering and Technology, Chinese Academy of Sciences, Suzhou, China.
J Magn Reson Imaging. 2021 Sep;54(3):880-887. doi: 10.1002/jmri.27592. Epub 2021 Mar 11.
Differential diagnosis of primary central nervous system lymphoma (PCNSL) and glioblastoma (GBM) is useful to guide treatment strategies.
To investigate the use of a convolutional neural network (CNN) model for differentiation of PCNSL and GBM without tumor delineation.
Retrospective.
A total of 289 patients with PCNSL (136) or GBM (153) were included, the average age of the cohort was 54 years, and there were 173 men and 116 women.
FIELD STRENGTH/SEQUENCE: 3.0 T Axial contrast-enhanced T -weighted spin-echo inversion recovery sequence (CE-T WI), T -weighted fluid-attenuation inversion recovery sequence (FLAIR), and diffusion weighted imaging (DWI, b = 0 second/mm , 1000 seconds/mm ).
A single-parametric CNN model was built using CE-T WI, FLAIR, and the apparent diffusion coefficient (ADC) map derived from DWI, respectively. A decision-level fusion based multi-parametric CNN model (DF-CNN) was built by combining the predictions of single-parametric CNN models through logistic regression. An image-level fusion based multi-parametric CNN model (IF-CNN) was built using the integrated multi-parametric MR images. The radiomics models were developed. The diagnoses by three radiologists with 6 years (junior radiologist Y.Y.), 11 years (intermediate-level radiologist Y.T.), and 21 years (senior radiologist Y.L.) of experience were obtained.
The 5-fold cross validation was used for model evaluation. The Pearson's chi-squared test was used to compare the accuracies. U-test and Fisher's exact test were used to compare clinical characteristics.
The CE-T WI, FLAIR, and ADC based single-parametric CNN model had accuracy of 0.884, 0.782, and 0.700, respectively. The DF-CNN model had an accuracy of 0.899 which was higher than the IF-CNN model (0.830, P = 0.021), but had no significant difference in accuracy compared to the radiomics model (0.865, P = 0.255), and the senior radiologist (0.906, P = 0.886).
A CNN model can differentiate PCNSL from GBM without tumor delineation, and comparable to the radiomics models and radiologists.
4 TECHNICAL EFFICACY: Stage 2.
原发性中枢神经系统淋巴瘤（PCNSL）和胶质母细胞瘤（GBM）的鉴别诊断有助于指导治疗策略。
研究卷积神经网络（CNN）模型在不勾画肿瘤的情况下区分 PCNSL 和 GBM 的应用。
回顾性。
共纳入 289 例 PCNSL（136 例）或 GBM（153 例）患者，队列平均年龄为 54 岁，其中 173 例为男性，116 例为女性。
磁场强度/序列：3.0T 轴向对比增强 T1 加权自旋回波反转恢复序列（CE-T1WI）、T1 加权液体衰减反转恢复序列（FLAIR）和扩散加权成像（DWI，b=0 秒/mm2、1000 秒/mm2）。
使用 CE-T1WI、FLAIR 和 DWI 衍生的表观扩散系数（ADC）图分别构建单参数 CNN 模型。通过逻辑回归将单参数 CNN 模型的预测值组合，构建基于决策级融合的多参数 CNN 模型（DF-CNN）。使用集成的多参数 MR 图像构建基于图像级融合的多参数 CNN 模型（IF-CNN）。构建了放射组学模型。由 3 位具有 6 年（初级放射科医生 Y.Y.）、11 年（中级放射科医生 Y.T.）和 21 年（高级放射科医生 Y.L.）经验的放射科医生进行诊断。
采用 5 折交叉验证进行模型评估。采用 Pearson 卡方检验比较准确率。采用 U 检验和 Fisher 确切检验比较临床特征。
CE-T1WI、FLAIR 和 ADC 单参数 CNN 模型的准确率分别为 0.884、0.782 和 0.700。DF-CNN 模型的准确率为 0.899，高于 IF-CNN 模型（0.830，P=0.021），但与放射组学模型（0.865，P=0.255）和高级放射科医生（0.906，P=0.886）的准确率无显著差异。
CNN 模型无需勾画肿瘤即可区分 PCNSL 和 GBM，与放射组学模型和放射科医生相当。
4 级
2 级
