INTELLIGENT HEALTHCARE WITH WEB-BASED REAL-TIME PAIN LEVEL DETECTION USING LIGHTWEIGHT RESNET-15

Abstract

The evaluation of pain is an essential component of professional decision-making, especially for patients who cannot adequately communicate their symptoms. This research introduces a web-based application for the automated detection of pain levels, employing a deep learning model based on ResNet-15, a lightweight and streamlined variant of the ResNet architecture for enhanced performance. The device analyzes facial expressions captured through live video streams or pictures to predict pain severity levels in real time. The convolutional layers of the proposed model are employed to extract profound visual information associated with facial indicators of pain. The algorithm is trained using publicly accessible datasets of facial expressions annotated with pain scores. Healthcare practitioners can remotely or in-clinic monitor patient pain using an accessible tool enabled by a web interface that promotes seamless user interaction. The experimental findings indicate that the classification of various pain levels can be achieved with notable accuracy while maintaining minimal computational complexity, rendering it a feasible alternative for implementation in resource-constrained environments. This study integrates deep learning and web technologies to provide non-invasive, real-time pain assessment, hence advancing intelligent healthcare systems.