Attenuation correction using deep Learning and integrated UTE/multi-echo Dixon sequence: evaluation in amyloid and...

Area of Science:

• Neuroimaging
• Biomedical Engineering
• Radiology

Background:

• Positron Emission Tomography (PET) imaging of amyloid and tau pathologies are crucial biomarkers for Alzheimer's disease (AD).
• Accurate quantification in amyloid and tau PET imaging is challenged by attenuation correction (AC) errors, particularly in cortical regions near bone.
• Magnetic Resonance (MR)-based AC methods are being developed to improve PET quantitation accuracy.

Purpose of the Study:

• To present a novel MR-based attenuation correction (AC) method for amyloid and tau PET imaging.
• To combine deep learning with a novel ultrashort time-to-echo (UTE)/multi-echo Dixon (mUTE) sequence for enhanced AC accuracy.
• To evaluate the performance of the proposed method against existing AC techniques.

Main Methods:

• A deep learning approach was developed for MR-based AC, utilizing a novel mUTE sequence.
• The method was evaluated on 35 subjects who underwent both 11C-PiB and 18F-MK6240 PET scans.
• Comparisons were made with Dixon-based atlas methods and other deep learning techniques using Dice coefficients and PET error analysis.

Main Results:

• The mUTE-based deep learning method achieved the highest Dice coefficients for bone regions (0.87-0.94).
• Regional SUV and SUVR errors were below 2% for all deep learning methods, significantly outperforming the atlas method (around 6%).
• Surface analysis showed the mUTE-based deep learning method minimized errors in cortical regions compared to the atlas method.

Conclusions:

• Deep learning combined with mUTE imaging provides accurate attenuation correction for amyloid and tau PET/MR.
• This advanced AC method holds promise for improving the diagnostic and monitoring capabilities of PET imaging in Alzheimer's disease.
• The mUTE-based deep learning approach offers superior accuracy and reduced errors in cortical regions relevant to AD pathology.