Oxygen extraction fraction mapping at 3 Tesla using an artificial neural network: A feasibility study

Area of Science:

• Biomedical Imaging
• Neuroscience
• Medical Physics

Background:

• Oxygen extraction fraction (OEF) is a key biomarker for assessing tissue viability.
• Magnetic Resonance Imaging (MRI) allows noninvasive OEF estimation via the blood-oxygenation-level-dependent (BOLD) effect.
• Current quantitative OEF mapping using least-squares regression (LSR) requires lengthy acquisition times, hindering clinical adoption.

Purpose of the Study:

• To evaluate artificial neural networks (ANNs) as an alternative to LSR for OEF mapping.
• To assess the potential of ANNs to reduce acquisition times and improve OEF mapping accuracy.
• To present in vivo OEF mapping results comparing LSR and ANN methods.

Main Methods:

• In vivo MRI data acquired at 3T using a gradient-echo sampled spin-echo (GESSE) sequence from five healthy volunteers and one brain tumor patient.
• ANNs were trained using simulated BOLD data.
• OEF mapping was performed using both conventional LSR and the proposed ANN method for comparison.

Main Results:

• In healthy subjects, mean OEF was 36±2% (LSR) and 40±1% (ANN).
• ANN method reduced OEF variance within subjects from 8% to 6%.
• In the patient, ANN showed fewer artifacts in surrounding tissue compared to LSR, while both identified an OEF hotspot in the tumor.

Conclusions:

• ANN-based OEF mapping offers reduced variance and potential for shorter acquisition times.
• The ANN method shows promise for facilitating the clinical integration of OEF mapping.
• ANNs may provide a more robust and efficient approach to quantitative OEF analysis in clinical settings.