Reducing scan time in ¹⁷⁷ Lu planar scintigraphy using convolutional neural network: A Monte Carlo simulation study.

Area of Science:

• Nuclear Medicine
• Medical Imaging
• Radiotherapy

Background:

• ¹⁷⁷Lu-based peptide receptor radionuclide therapy (PRRT) is crucial for personalized cancer treatment.
• Accurate dosimetry is essential for effective ¹⁷⁷Lu-PRRT.
• Reducing scan time in ¹⁷⁷Lu scintigraphy can improve patient comfort and throughput.

Purpose of the Study:

• To develop and evaluate a convolutional neural network (CNN) for reducing scan time in ¹⁷⁷Lu planar scintigraphy.
• To assess the feasibility of using CNN-denoised images for personalized dosimetry in ¹⁷⁷Lu-PRRT.

Main Methods:

• A DenseNet-based CNN model was trained and tested using Monte Carlo simulations.
• Input images with 10-90% of total photon counts were used to generate full-count images (CNN labels).
• Statistical comparison (t-test) of pixel intensities between full-count and CNN-denoised images was performed.

Main Results:

• No significant differences were observed for objects >13 mm in the Derenzo phantom.
• Statistically significant differences were found for 10-mm rods when using 10-60% photon counts.
• Significant differences were noted in kidney dosimetry using 10% photon counts in the NCAT phantom.

Conclusions:

• The proposed CNN method can reduce ¹⁷⁷Lu scintigraphy scan time by up to 70% for objects >13 mm.
• This approach facilitates personalized dosimetry for ¹⁷⁷Lu-PRRT in clinical settings.
• The method shows promise for improving efficiency in nuclear medicine procedures.