Artificial Intelligence Based Pain Assessment Technology in Clinical Application of Real-World Neonatal Blood...

Area of Science:

• Neonatal care
• Medical technology
• Pain management

Background:

• Accurate neonatal pain assessment (NPA) is crucial for effective pain management but challenging for healthcare professionals.
• Existing methods for NPA can be subjective and time-consuming.
• The development of objective and efficient NPA tools is a significant need in neonatal intensive care units.

Purpose of the Study:

• To evaluate the clinical practicability and accuracy of an artificial intelligence-based neonatal pain assessment (AI-NPA) tool.
• To compare the AI-NPA system's performance against on-site nurse assessments during real-world procedures.
• To determine the consistency of AI-driven pain scores and grades with established clinical standards.

Main Methods:

• A prospective study involving 232 newborns undergoing blood sampling in neonatal wards.
• Utilized a self-developed automated AI-NPA system and compared its results with nurses' on-site NPAs (OS-NPAs).
• Employed the Neonatal Infant Pain Scale (NIPS) for evaluation, with Spearman correlation and agreement analysis for statistical validation.

Main Results:

• The AI-NPA system achieved high accuracy, with 88.79% for NIPS pain scores and 95.25% for pain grades, compared to the gold standard OS-NPA.
• Excellent agreement was observed, indicated by kappa values of 0.92 for pain score and 0.90 for pain grade (p < 0.001).
• The automated system demonstrated strong consistency with expert clinical judgment.

Conclusions:

• The automated AI-NPA system is highly consistent with on-site nurse assessments for neonatal pain during blood sampling.
• AI-NPA offers significant advantages in repeatability, efficiency, and cost-effectiveness.
• The AI technique holds considerable promise for widespread adoption in precise and efficient neonatal pain management.