Computer-assisted decision support for the usage of preventive antibacterial therapy in children with febrile...

Area of Science:

• Pediatric infectious diseases
• Medical imaging analysis
• Artificial intelligence in healthcare

Background:

• Urinary tract infections (UTIs) are common in children.
• Controversy exists regarding preventive antibiotics for febrile pyelonephritis.
• No prior studies utilized deep learning for this specific diagnostic challenge.

Purpose of the Study:

• Investigate the need for preventive antibiotics in children under 2 with febrile pyelonephritis.
• Utilize deep learning technology on renal static imaging data.
• Develop a computer-assisted decision support system for diagnosis.

Main Methods:

• Collected a dataset of 176 children (64 without preventive antibiotics, 112 with).
• Employed classic deep learning models, including AlexNet, for analysis.
• Evaluated model performance using accuracy, sensitivity, and specificity.

Main Results:

• Deep learning models demonstrated feasibility in screening for preventive antibiotic needs.
• AlexNet achieved 84.05% accuracy, 81.71% sensitivity, and 86.70% specificity.
• The study successfully achieved graded diagnosis for treatment requirements.

Conclusions:

• Deep learning offers a novel approach for computer-assisted diagnosis in febrile pyelonephritis.
• This technology can support clinical decisions regarding antibiotic use in pediatric UTIs.
• Further research can refine AI tools for pediatric infectious disease management.