Intelligent system for infants' pain detection: pain intensity estimation using deep learning approach.

Area of Science:

• Medical technology
• Artificial Intelligence
• Neonatal care

Background:

• Traditional pain assessment tools for infants have limitations.
• Accurate pain detection is crucial for effective pain management in neonates.
• Artificial Intelligence offers novel approaches for objective pain assessment.

Purpose of the Study:

• To develop a deep learning model for estimating infant pain intensity (0-9) from facial expressions.
• To address limitations of traditional pain assessment methods in the NICU.
• To improve pain management for non-verbal infants.

Main Methods:

• A regression Convolutional Neural Network (CNN) model was developed.
• Transfer learning with a pre-trained VGG16 model was employed for fine-tuning.
• Facial expressions of full-term infants during NICU procedures were analyzed.

Main Results:

• The CNN model achieved a Mean Absolute Error (MAE) of 0.494 and Mean Squared Error (MSE) of 0.435.
• The model demonstrated good performance in estimating pain intensity.
• Transfer learning and fine-tuning helped improve model performance and avoid overfitting.

Conclusions:

• The proposed AI model accurately detects infant pain based on facial expressions.
• This technology can significantly aid healthcare professionals in providing effective pain management.
• Future work includes developing a portable, user-friendly web application for the system.