Differentiation between normal and abnormal kidneys using ^99mTc-DMSA SPECT with deep learning in paediatric patients.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Pediatric Nephrology

Background:

• Accurate differentiation of normal versus scarred kidneys in children is crucial for timely intervention.
• Technetium-99m dimercaptosuccinic acid (99mTc-DMSA) SPECT imaging is a standard diagnostic tool.
• Interpreting renal SPECT images can be subjective and time-consuming.

Purpose of the Study:

• To evaluate the feasibility of deep learning (DL) for distinguishing normal from abnormal pediatric kidneys.
• To assess the performance of DL models using various 99mTc-DMSA SPECT image formats.
• To determine the potential of DL in improving the diagnostic accuracy of renal scarring in children.

Main Methods:

• Retrospective analysis of 301 pediatric 99mTc-DMSA renal SPECT examinations.
• Training DL models on 3D SPECT, 2D MIPs, and 2.5D MIPs (transverse, sagittal, coronal views).
• Comparison of DL model performance against consensus readings by nuclear medicine physicians.

Main Results:

• The DL model trained using 2.5D MIPs demonstrated superior performance compared to 3D SPECT or 2D MIPs.
• The 2.5D DL model achieved an accuracy of 92.5%, sensitivity of 90%, and specificity of 95%.
• These results indicate a high capability of the DL model in differentiating normal from abnormal renal SPECT findings.

Conclusions:

• Deep learning shows significant potential for accurately differentiating normal from abnormal pediatric kidneys using 99mTc-DMSA SPECT.
• The 2.5D MIPs approach for DL model training yielded the best diagnostic performance.
• DL-based analysis could enhance the efficiency and reliability of diagnosing renal abnormalities in pediatric patients.