Accurate Patient-Specific Machine Learning Models of Glioblastoma Invasion Using Transfer Learning.

BACKGROUND AND PURPOSE: MR imaging-based modeling of tumor cell density can substantially improve targeted treatment of glioblastoma. Unfortunately, interpatient variability limits the predictive ability of many modeling approaches. We present a transfer learning method that generates individualized patient models, grounded in the wealth of population data, while also detecting and adjusting for interpatient variabilities based on each patient's own histologic data. MATERIALS AND METHODS: We recruited patients with primary glioblastoma undergoing image-guided biopsies and preoperative imaging, including contrast-enhanced MR imaging, dynamic susceptibility contrast MR imaging, and diffusion tensor imaging. We calculated relative cerebral blood volume from DSC-MR imaging and mean diffusivity and fractional anisotropy from DTI. Following image coregistration, we assessed tumor cell density for each biopsy and identified corresponding localized MR imaging measurements. We then explored a range of univariate and multivariate predictive models of tumor cell density based on MR imaging measurements in a generalized one-model-fits-all approach. We then implemented both univariate and multivariate individualized transfer learning predictive models, which harness the available population-level data but allow individual variability in their predictions. Finally, we compared Pearson correlation coefficients and mean absolute error between the individualized transfer learning and generalized one-model-fits-all models. RESULTS: Tumor cell density significantly correlated with relative CBV (r = 0.33, P < .001), and T1-weighted postcontrast (r = 0.36, P < .001) on univariate analysis after correcting for multiple comparisons. With single-variable modeling (using relative CBV), transfer learning increased predictive performance (r = 0.53, mean absolute error = 15.19%) compared with one-model-fits-all (r = 0.27, mean absolute error = 17.79%). With multivariate modeling, transfer learning further improved performance (r = 0.88, mean absolute error = 5.66%) compared with one-model-fits-all (r = 0.39, mean absolute error = 16.55%). CONCLUSIONS: Transfer learning significantly improves predictive modeling performance for quantifying tumor cell density in glioblastoma.

The biology and mathematical modelling of glioma invasion: a review.

Adult gliomas are aggressive brain tumours associated with low patient survival rates and limited life expectancy. The most important hallmark of this type of tumour is its invasive behaviour, characterized by a markedly phenotypic plasticity, infiltrative tumour morphologies and the ability of malignant progression from low- to high-grade tumour types. Indeed, the widespread infiltration of healthy brain tissue by glioma cells is largely responsible for poor prognosis and the difficulty of finding curative therapies. Meanwhile, mathematical models have been established to analyse potential mechanisms of glioma invasion. In this review, we start with a brief introduction to current biological knowledge about glioma invasion, and then critically review and highlight future challenges for mathematical models of glioma invasion.

Analysis of glioblastoma tumor coverage by oncolytic virus-loaded neural stem cells using MRI-based tracking and histological reconstruction.

In preclinical studies, neural stem cell (NSC)-based delivery of oncolytic virus has shown great promise in the treatment of malignant glioma. Ensuring the success of this therapy will require critical evaluation of the spatial distribution of virus after NSC transplantation. In this study, the patient-derived GBM43 human glioma line was established in the brain of athymic nude mice, followed by the administration of NSCs loaded with conditionally replicating oncolytic adenovirus (NSC-CRAd-S-pk7). We determined the tumor coverage potential of oncolytic adenovirus by examining NSC distribution using magnetic resonance (MR) imaging and by three-dimensional reconstruction from ex vivo tissue specimens. We demonstrate that unmodified NSCs and NSC-CRAd-S-pk7 exhibit a similar distribution pattern with most prominent localization occurring at the tumor margins. We were further able to visualize the accumulation of these cells at tumor sites via T2-weighted MR imaging as well as the spread of viral particles using immunofluorescence. Our analyses reveal that a single administration of oncolytic virus-loaded NSCs allows for up to 31% coverage of intracranial tumors. Such results provide valuable insights into the therapeutic potential of this novel viral delivery platform.

From patient-specific mathematical neuro-oncology to precision medicine.

Gliomas are notoriously aggressive, malignant brain tumors that have variable response to treatment. These patients often have poor prognosis, informed primarily by histopathology. Mathematical neuro-oncology (MNO) is a young and burgeoning field that leverages mathematical models to predict and quantify response to therapies. These mathematical models can form the basis of modern "precision medicine" approaches to tailor therapy in a patient-specific manner. Patient-specific models (PSMs) can be used to overcome imaging limitations, improve prognostic predictions, stratify patients, and assess treatment response in silico. The information gleaned from such models can aid in the construction and efficacy of clinical trials and treatment protocols, accelerating the pace of clinical research in the war on cancer. This review focuses on the growing translation of PSM to clinical neuro-oncology. It will also provide a forward-looking view on a new era of patient-specific MNO.