Detectability of small objects in PET/computed tomography phantom images with Bayesian penalised likelihood...

Area of Science:

• Nuclear Medicine
• Medical Imaging
• Radiology

Background:

• Accurate detection of small active lesions in PET/CT is crucial for diagnosis and treatment monitoring.
• Ordered subset expectation maximization (OSEM) is a standard reconstruction algorithm, but its performance with subcentimetre lesions can be limited.
• Bayesian penalised likelihood reconstruction (BPLR) offers potential improvements in image quality and lesion detectability.

Purpose of the Study:

• To evaluate the subcentimetre lesion detection performance of Bayesian penalised likelihood reconstruction (BPLR) compared to ordered subset expectation maximization (OSEM) in PET/CT phantom studies.
• To assess observer performance in identifying active and non-active lesions using both reconstruction methods.

Main Methods:

• Phantom images with active and non-active lesions (3-7 mm diameter) at varying contrast ratios were acquired using PET/CT.
• Images were reconstructed using both BPLR (Q.Clear) and OSEM (VUE Point HD HDAC).
• Three experienced observers evaluated lesion detectability and assigned confidence ratings, with data analyzed using JAFROC software.

Main Results:

• BPLR demonstrated statistically significant visual detection gains for active lesions (weighted JAFROC FOM = 0.77) compared to OSEM (FOM = 0.74).
• Improvements were particularly notable for 5 mm active lesions at an 8:1 lesion-to-background ratio.
• No significant differences were observed in the detection of non-active lesions between BPLR and OSEM.

Conclusions:

• BPLR provides significant detection gains for small active lesions (≤7 mm) in PET/CT imaging.
• The enhanced contrast-to-noise ratio with BPLR supports its use for visualizing small active lesions.
• BPLR does not offer an advantage in identifying true negative (non-active) lesions.