Glioblastoma Drug Sensitivity Prediction Using AI-Based Radiogenomic and Pharmacogenomic Methods
Faculty Mentor
Robin Elliot
Major/Area of Research
Artificial Intelligence and Pharmacology
Description
INTRODUCTION: Glioblastoma multiforme (GBM) is the most malignant, treatment-resistant brain cancer and thus deserves new innovative personalized medicine approaches. Artificial intelligence (AI) will likely transform with the addition of radiogenomics, a correlation between tumor genetics and imaging biomarkers on the basis of MRI information, and pharmacogenomics, and drug response prediction by genetic screening. By correlating multimodal MRI scan radiomic features with drug-response profiles and genetic signatures, the research proposes an AI system to forecast patient-specific drug sensitivities.
METHOD: The research will apply multimodal AI models to combine both datasets, collection learning methods for pharmacogenomic analysis, and deep learning models like convolutional neural networks for extracting radiomic features. Retrospective datasets (e.g., GDSC and TCGA) will be applied for comparing prediction performance of the model, and prospectively obtained clinical data will be applied for validation. Survival analysis, ROC-AUC values in case of binary classification, and accuracy will be key measurements of evaluation. Clinical benefit in terms of AEUR is anticipated from the creation of an AI-based decision-support device that will possess the potential to guide personalized therapy, decrease toxicity, and enhance patient survival. The detection of new non-invasive imaging biomarkers corresponding to some gene mutations and drug responses is a second objective of the project. Cloud optimization, multi-institution data training, and advanced normalization techniques will altogether alleviate problems of computational complexity, data heterogeneity, and generalizability of the model.
CONCLUSION: This finding could substantially transform precision oncology of brain cancer therapy by integrating radiogenomics with pharmacogenomics and AI. In order to enhance patient outcomes and reduce healthcare expenditure, the proposed AI model can also function as a clinical tool to enhance therapeutic strategies. For practical application, further validation from clinical collaborations and regulatory approval will be needed.
Glioblastoma Drug Sensitivity Prediction Using AI-Based Radiogenomic and Pharmacogenomic Methods
INTRODUCTION: Glioblastoma multiforme (GBM) is the most malignant, treatment-resistant brain cancer and thus deserves new innovative personalized medicine approaches. Artificial intelligence (AI) will likely transform with the addition of radiogenomics, a correlation between tumor genetics and imaging biomarkers on the basis of MRI information, and pharmacogenomics, and drug response prediction by genetic screening. By correlating multimodal MRI scan radiomic features with drug-response profiles and genetic signatures, the research proposes an AI system to forecast patient-specific drug sensitivities.
METHOD: The research will apply multimodal AI models to combine both datasets, collection learning methods for pharmacogenomic analysis, and deep learning models like convolutional neural networks for extracting radiomic features. Retrospective datasets (e.g., GDSC and TCGA) will be applied for comparing prediction performance of the model, and prospectively obtained clinical data will be applied for validation. Survival analysis, ROC-AUC values in case of binary classification, and accuracy will be key measurements of evaluation. Clinical benefit in terms of AEUR is anticipated from the creation of an AI-based decision-support device that will possess the potential to guide personalized therapy, decrease toxicity, and enhance patient survival. The detection of new non-invasive imaging biomarkers corresponding to some gene mutations and drug responses is a second objective of the project. Cloud optimization, multi-institution data training, and advanced normalization techniques will altogether alleviate problems of computational complexity, data heterogeneity, and generalizability of the model.
CONCLUSION: This finding could substantially transform precision oncology of brain cancer therapy by integrating radiogenomics with pharmacogenomics and AI. In order to enhance patient outcomes and reduce healthcare expenditure, the proposed AI model can also function as a clinical tool to enhance therapeutic strategies. For practical application, further validation from clinical collaborations and regulatory approval will be needed.