A fully automatic deep learning-based method for segmenting regions of interest and predicting renal function in...

Area of Science:

• Nuclear Medicine
• Artificial Intelligence in Healthcare
• Pediatric Imaging

Background:

• Accurate renal region delineation is crucial for pediatric dynamic renal scintigraphy (DRS).
• Manual segmentation is time-consuming and prone to inter-observer variability.
• Deep learning offers potential for automated and accurate image analysis.

Purpose of the Study:

• To develop and evaluate a deep learning (DL) model for fully automatic renal ROI segmentation.
• To assess the DL model's ability to calculate renal function parameters in pediatric ^99mTechnetium-ethylenedicysteine (^99mTc-EC) DRS.

Main Methods:

• Retrospective analysis of 1,283 pediatric DRS datasets.
• Development and evaluation of a Fully Automatic Segmentation of ROIs (FASR) deep learning model.
• Comparison of automated versus manual segmentation for renal blood perfusion rate (BPR) and differential renal function (DRF) using IOU, DSC, ICC, and Pearson correlation.

Main Results:

• The FASR model demonstrated high performance with precision 0.88, recall 0.94, IOU 0.83, and DSC 0.91.
• Excellent correlation between automated and manual methods for BPR (r=0.94) and DRF (r=0.97).
• High intraclass correlation coefficients (ICCs) of 0.94 for both BPR and DRF, indicating strong agreement.

Conclusions:

• A reliable and stable deep learning model for automated renal ROI segmentation in pediatric DRS was developed.
• The DL model accurately predicts renal function parameters, offering a valuable tool for clinical practice.
• This automated approach enhances efficiency and consistency in pediatric renal function assessment.