ACTIVITY CONCENTRATION ESTIMATION IN AUTOMATED KIDNEY SEGMENTATION BASED ON CONVOLUTION NEURAL NETWORK METHOD FOR...

Area of Science:

• Nuclear Medicine
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Accurate dosimetry for 177Lutetium (Lu) DOTATATE treatments relies on kidney segmentation.
• Manual segmentation of kidneys from computed tomography (CT) is time-consuming and prone to misregistration errors with single-photon emission computed tomography (SPECT) images.

Purpose of the Study:

• To develop and validate a convolution neural network (CNN) for automated kidney segmentation.
• To improve the accuracy and efficiency of dosimetry calculations in 177Lu-DOTATATE therapy by aligning CT-segmented regions of interest (VOI) with SPECT images.

Main Methods:

• A CNN was trained using SPECT/CT images from 137 patients undergoing 177Lu-DOTATATE treatment.
• Automated kidney segmentation was performed using the CNN.
• Activity concentrations in automated and manual segmentations were compared using correlation and Bland-Altman analyses on a testing cohort of 20 patients.

Main Results:

• Activity concentrations in automated and manual kidney segmentations showed strong correlation (r > 0.96, p < 0.01).
• CNN segmentation demonstrated higher accuracy than manual segmentation without VOI adjustment, as per Bland-Altman analysis.
• The CNN provided a fast, robust, and accurate method for assessing kidney activity concentrations in SPECT images.

Conclusions:

• The developed CNN enables accurate and efficient automated kidney segmentation for 177Lu-DOTATATE therapy dosimetry.
• This AI-driven approach offers a reliable alternative to manual segmentation, potentially reducing errors and improving workflow.
• The CNN performs comparably to manual methods while significantly enhancing speed and robustness.