Apanicu 1.0: sistema automàtic per a la valoració del dolor dels nounats ingressats a les unitats de cures intensives

Identifier:  TDX:2603

Handle: https://hdl.handle.net/20.500.11797/TDX2603

Authors:  Casanovas Marsal, Josep Oriol

Abstract:
Introduction: Pain is a recurrent topic in Neonatal Intensive Care Units (NICUs), due to the number of procedures performed in such scenarios. Suffering from intense and continuing pain in the first days of life, can entail consequences in the future. The first measure to deal with pain is its correct assessment. The developed scales include some important subjectivity aspects. The use of artificial intelligence (AI) may allow the continuous monitoring, in real time, of the pain so it can be integrated as the fifth vital sign. The goal of this thesis is to design and implement an intelligent and automatic system to assess pain in neonates hospitalised in a NICU. Material and method: Experimental study, evaluative. A prototype analyses behavioural and physiological parameters. By means of the AI a level for pain is computed, which is displayed on a standard vital signs monitor. 11 measures were conducted with neonates (gestational ages between 25 and 42 weeks) hospitalised in the NICU of Hospital Universitari Joan XXIII in Tarragona. For the human assessment of pain (observer A and B) scales Premature Infant Pain Profile (PIPP) and Susan Given Bell (SGB) were used. The Spearman correlation coefficient was used for the validation of the prototype. The ANOVA test was used for comparing variances and averages. Results: Sample n=11. Gestational age 32.00±7.23 weeks; weight 2046.50±1409.27 g. The 63.6% of the sample were boys (n=7). Pain observer A in PIPP: 2.6±1.4, in SGB: 2.4±1.3. Pain observer B in PIPP: 2.9±1.4, in SGB: 2.2±1.5. Pain in prototype 1.55±0.69. ANOVA between PIPP-prototype (observer A and B) F=3.62 (p=0.03). Measures of value SGB-prototype (observer A and B) F=2.99 (p=0.06). Conclusions: The design of the prototype and its connection was successfully performed, but the computed values are not correlated. By improving the software and creating a version 2.0, better results would be expected.