Overall Survival Prediction of Glioma Patients with Genomics
Glioma derives from glial cells and is the most common primary central nervous system (CNS) malignant tumor. This poor prognosis of Glioma is a result of intra-tumor heterogeneity, which can be seen in levels of protein expression, metabolic, or bioenergetic behavior, besides their micro-environment biochemistry and structural composition. Mostly brain tumor in the MRI is segmented and decisions are taken for treatments and overall survival calculations, based on the radiomic biomarkers. Therefore, the need of genomics, for glioma prognosis analysis occurs.
In previous works statistical analysis have been done to find genomic biomarkers such as gene expression, methylation, mutations and copy number variations that can impact on treatment planning, subtype classification, overall survival etc. Thus, there is a research direction for applying deep learning and machine learning approaches to predict the overall survival of glioma patients with genomics. This research is mainly focused on obtaining an explainable algorithm, based on both machine learning, deep learning and statistical analysis tools, to predict overall survival of Glioma patients, overcoming the current limitations in state-of-art overall survival prediction with radiomics.
This work is important in clinical applications such as treatment planning of glioma patients. The clinicians can decide what kind of treatment (temozolomide chemotherapy, radiation therapy etc) should be given by looking at the overall survival prediction given by the model. This can also be done using imaging biomarkers and clinical data. But it won’t be able to capture the intra tumor heterogeneity, which occurs due to the variations in the gene sequences and have a direct impact on tumor progression. These gene sequence changes can happen due to expression level variations, methylation variations, mutations and copy number variations. Therefore, we focus on identifying, out of these what genomic features are most significant in developing an overall survival prediction model.
