To develop and evaluate a novel, fully-interpretable neural network model for predicting furosemide dosage in neonates following congenital cardiac surgery, addressing the critical issue of fluid overload management.
Key Findings:
The TGFNN-R model provides interpretable predictions that align with clinical decision-making processes, demonstrating a reduction in variability in dosing.
Fuzzy sets enhance the model's ability to reflect the complexity of clinical data, leading to more personalized treatment approaches.
Interpretation:
The TGFNN-R model represents a significant advancement in the use of machine learning for clinical decision support, particularly in managing fluid overload in neonates post-surgery, with potential to improve patient outcomes.
Limitations:
The study is limited to a single center, which may affect the generalizability of the findings and introduce selection bias.
Further validation is needed in diverse clinical settings to confirm the model's efficacy.
Conclusion:
The TGFNN-R model shows promise as a tool for improving diuretic dosing accuracy in neonates after congenital heart surgery, potentially reducing variability in clinical practice and warranting further research for broader application.
