Applications of a HiBED Model to prognosticate Glioma/Glioblastoma Patients
Gaura Jha
Lay Summary:
I'm developing a new tool to help doctors predict how aggressive a type of brain cancer called glioblastoma might be in each patient. By analyzing different kinds of medical data, this tool aims to provide more accurate and personalized treatment plans, potentially improving outcomes for patients facing this challenging disease.
Abstract:
Glioblastoma, a highly aggressive and lethal form of brain cancer, presents significant challenges in patient prognosis due to its complex and heterogeneous nature. Traditional prognostic methods often fall short in accurately predicting patient outcomes, leading to a critical need for advanced models that can integrate diverse biological data and provide personalized predictions. The HiBED model (Hierarchical Bayesian Evidence Distribution), developed for this purpose, represents a novel approach in prognosticating glioblastoma and glioma patients by leveraging a sophisticated probabilistic framework. This model integrates multi-modal data, including clinical, radiological, and molecular features, to create a comprehensive and individualized risk profile for each patient. In this project, the application of the HiBED model is explored to enhance the accuracy of survival predictions and treatment planning for glioblastoma patients. The model's ability to incorporate hierarchical data structures allows for the assessment of both population-level trends and patient-specific factors, providing a more nuanced understanding of disease progression. Key components of the model include Bayesian inference, which enables the incorporation of prior knowledge and uncertainty into predictions, and evidence distribution, which captures the variability in patient outcomes. The project involves a detailed evaluation of the HiBED model's performance using retrospective patient data, with comparisons to existing prognostic models. Metrics such as accuracy, sensitivity, and specificity are used to assess the model's effectiveness in different clinical scenarios. Furthermore, the potential for the HiBED model to inform personalized treatment strategies, by predicting responses to various therapeutic interventions, is investigated. The results of this study could lead to significant advancements in the field of neuro-oncology, providing clinicians with a powerful tool for improving patient care and outcomes.
Q&A:
Bios: Gaura Jha
Program Track: Advanced Research
GitHub Username:
gaurajha -Gaura Jha
What was your favorite seminar? Why?
The seminar by Alex Xu about techniques and technologies used for staining - honestly one of the most engaging seminars I’ve ever attended and I learnt more than I could imagine! -Gaura Jha
If you were to summarize your summer internship experience in one sentence, what would it be?
It was challenging and enriching - I learnt more about pathology and machine learning than I could have online or in a course, and the actual hands-on experience helped me make mistakes and learn from them! -Gaura Jha