Editorial: Pre-clinical Models of PTSD

社论:创伤后应激障碍的临床前模型

阅读:1

Abstract

Glioblastomas (GBMs) are notoriously heterogeneous tumors, both between and within tumors. Non-invasive magnetic resonance images (MRIs) provide some information about tumor characteristics, but understanding of the underlying tumor composition remains incomplete. We propose combining a machine-learning (ML) model of tumor cell density with a model of MRI physics that takes into account signal contributions from the intracellular (ICS) and extracellular (ECS) space to predict the volumetric proportions of tumor cells, non-tumor cells, and extracellular space/edema. T2-weighted and T1Gd MRIs of 18 GBM patients were acquired. In total, 82 image-localized biopsies were collected. A neuropathologist scored the biopsies for the percentage of tumor cells. Biopsies were analyzed for TP53 amplification using array CGH. Machine learning utilizing MRIs was performed to predict the percentage of tumor cells in each voxel. A multi-exponential model of the MRI signal equation was utilized to estimate the ECS and ICS in each voxel. The predicted ICS was then modulated by the ML model to estimate the proportion of tumor and non-tumor cells. Further, indices representing relative edema within the tumor (RAI) and relative amount of tumor compared to all cells (RTI) were calculated. Pathology scores were negatively correlated with the predicted proportion of non-tumor cells (p<0.001) and positively correlated with the RTI predicted by the model (p<0.001). Mean non-tumor cell proportion was significantly lower for biopsies with deleted TP53 (p=0.024). Additionally, RTI was significantly higher for biopsies with deleted TP53 (p=0.006). Biopsy samples lacking the TP53 tumor suppressor gene were predicted to have a lower amount of non-tumor cells and a higher relative amount of tumor cells suggestive of the recruitment of less non-tumor cells (e.g. inflammatory) in the presence of TP53. Combining machine learning models of tumor cell density and mechanistic models of MRI physics can elucidate the underlying microenvironment from non-invasive imaging.

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。