Primary Central Nervous System Lymphoma: Clinical Evaluation of Automated Segmentation on Multiparametric MRI Using Deep Learning

Abstract

Background: Precise volumetric assessment of brain tumors is relevant for treatment planning and monitoring. However, manual segmentations are time-consuming and impeded by intra- and interrater variabilities. Purpose: To investigate the performance of a deep-learning model (DLM) to automatically detect and segment primary central nervous system lymphoma (PCNSL) on clinical MRI. Study Type: Retrospective. Population: Sixty-nine scans (at initial and/or follow-up imaging) from 43 patients with PCNSL referred for clinical MRI tumor assessment. Field Strength/Sequence: T1−/T2-weighted, T1-weighted contrast-enhanced (T1 CE), and FLAIR at 1.0, 1.5, and 3.0T from different vendors and study centers. Assessment: Fully automated voxelwise segmentation of tumor components was performed using a 3D convolutional neural network (DeepMedic) trained on gliomas (n = 220). DLM segmentations were compared to manual segmentations performed in a 3D voxelwise manner by two readers (radiologist and neurosurgeon; consensus reading) from T1 CE and FLAIR, which served as the reference standard. Statistical Tests: Dice similarity coefficient (DSC) for comparison of spatial overlap with the reference standard, Pearson's correlation coefficient (r) to assess the relationship between volumetric measurements of segmentations, and Wilcoxon rank-sum test for comparison of DSCs obtained in initial and follow-up imaging. Results: The DLM detected 66 of 69 PCNSL, representing a sensitivity of 95.7%. Compared to the reference standard, DLM achieved good spatial overlap for total tumor volume (TTV, union of tumor volume in T1 CE and FLAIR; average size 77.16 ± 62.4 cm3, median DSC: 0.76) and tumor core (contrast enhancing tumor in T1 CE; average size: 11.67 ± 13.88 cm3, median DSC: 0.73). High volumetric correlation between automated and manual segmentations was observed (TTV: r = 0.88, P < 0.0001; core: r = 0.86, P < 0.0001). Performance of automated segmentations was comparable between pretreatment and follow-up scans without significant differences (TTV: P = 0.242, core: P = 0.177). Data Conclusion: In clinical MRI scans, a DLM initially trained on gliomas provides segmentation of PCNSL comparable to manual segmentation, despite its complex and multifaceted appearance. Segmentation performance was high in both initial and follow-up scans, suggesting its potential for application in longitudinal tumor imaging. Level of Evidence: 3. Technical Efficacy Stage: 2.