Computing quantitative indicators of structural renal damage in pediatric DMSA scans.

Area of Science:

• Medical Imaging
• Computational Biology
• Pediatric Nephrology

Background:

• Renal scarring from urinary tract infections is a common pediatric concern.
• Accurate quantification of renal damage from technetium-99m dimercaptosuccinic acid (DMSA) scans is crucial for patient management.
• Current visual assessment of DMSA scans can be subjective and observer-dependent.

Purpose of the Study:

• To propose, implement, and validate a computational framework for observer-independent quantification of structural renal damage.
• To develop an automated system for lesion detection and segmentation in DMSA scans.
• To correlate image-derived quantitative indicators with clinical data and long-term outcomes.

Main Methods:

• A computational framework was developed using expert-guided and automatic approaches on pediatric DMSA scans.
• Quantitative indicators were derived from lesion size, intensity, and histogram distribution.
• Correlation analysis was performed with clinical data (CRP, white cell count, fever) and follow-up DMSA scans.

Main Results:

• The automated system achieved high accuracy in classifying DMSA-positive scans (AUC=0.92, sensitivity=81%, specificity=94%).
• Lesion size correlated with fever and C-reactive protein levels (p<0.05).
• Histogram-derived measurements effectively detected permanent renal damage (AUC=0.86, sensitivity=100%, specificity=75%).

Conclusions:

• The computational framework offers a promising, objective method for quantifying renal damage from DMSA scans.
• This automated approach can complement visual diagnosis and non-imaging clinical indicators.
• The framework has potential for improved detection of permanent renal damage in pediatric patients.