Detection of EGFR gene mutations in glioblastoma: Utilizing information complexity in developing AI-based decision...

Area of Science:

• Neuro-oncology
• Artificial Intelligence in Medicine
• Medical Imaging Analysis

Background:

• Glioblastoma (GBM) is an aggressive brain cancer with significant heterogeneity.
• Epidermal Growth Factor Receptor (EGFR) gene mutations are critical prognostic markers in GBM.
• Current diagnostic methods like brain biopsies carry risks and potential sampling errors.

Purpose of the Study:

• To introduce an AI-based decision support system (DSS) for non-invasive EGFR mutation classification in GBM.
• To automate tumor segmentation from MRI scans for improved diagnostic accuracy.
• To evaluate the performance of deep learning models for predicting EGFR mutations using MRI data.

Main Methods:

• Development of a DSS utilizing deep neural networks (Inception ResNet-v2, DenseNet-121, ResNet-50).
• Training and testing models on a GBM dataset with three MRI input types: expert segmented, without segmentation, and without tumor.
• Application of Information Criteria (IC) for model selection to balance performance and complexity.

Main Results:

• DenseNet-121 demonstrated superior performance across all MRI input types, achieving high accuracy (up to 0.952).
• The model exhibited the highest precision and recall, particularly with expert-segmented MRI inputs.
• Multivariate statistical analysis confirmed significant differences in model performances, validating the chosen approach.

Conclusions:

• The AI-based DSS effectively classifies EGFR mutations in GBM patients using MRI data.
• Integrating information criteria enhances the robustness and interpretability of deep learning models in medical imaging.
• This non-invasive approach holds promise for improving GBM diagnosis and patient management, reducing reliance on biopsies.