Premature birth complications have different causes and vary in different parts of the world with sepsis as one of the leading causes of these complications. The body releases anti-inflammatory substances when an infection is detected and this, in turn, could damage healthy organs, especially when they are not fully developed. Preterm babies are susceptible to diseases due to their underdeveloped organs and immune systems. Hence, it is extremely important to treat sepsis as soon as the baby is diagnosed. Neonatal sepsis is a dangerous nonspecific disease in babies, and it is a clinically very difficult and challenging task to diagnose. Late or incorrect treatment of infants' sepsis can lead to death which is one of the most causes of mortality rate in neonates. In the traditional treatment of sepsis, the needed time and accuracy for diagnosis are still very concerning, considering the number of involved risks in late diagnosis or mistreatment of sepsis cases. Thus, the need for having a fast and reliable algorithm with high accuracy to predict sepsis before clinical recognition would help the doctors to treat the neonates in time and to reduce the mortality rate related to sepsis. This paper presents a fast, accurate, and reliable thermographic Bio-Remote Sensing approach to predicting sepsis in neonates and discusses the significance of combining the Thermal Imaging technique with Machine Learning (ML). At the same time, it provides a more practical and desirable solution for physicians by minimising the traditional diagnosis time and maximizing the accuracy of the prediction needed to detect sepsis in neonates.

Neonates can show sudden rapid body movements when they are in pain, need care, or need to be fed. They can also be very quiet and immovable or move very slowly when they are asleep or being fed. Monitoring a neonate's body temperature for a long time provides physicians and nurses valuable information about the health condition of the baby. Thermographic technology is a remote and very safe way to measure an accurate neonate's body temperature to monitor his/her vital signs. However, the tracking of an elastic thermographic profile of a subject with a random and erratic movement in the short- and long-term is a challenging task. The combination of the real-time thermographic detection and tracking system provides a safe and more robust non-invasive method to measure the vital signs and monitor the physiological changes of the neonates over time. However, this method can also be used for other target age groups.

One of the important measured vital signs in neonates is the body temperature. The traditional measurement uses adhesive pads, but medical staff are hindered by connectors attached to the infant. Remote infrared thermal imaging techniques provide a non-intrusive and safe method to measure body temperature. By means of the thermography technology, it is possible to monitor the variations and trends in the body temperature, which is more reliable, faster, less stressful than traditional methods. Measuring body temperature of a moving neonate remains a challenge. Moreover, factors like humidity, thermal lens forming through the incubator portholes, thermal noise from inside and outside the incubator, camera position and limited Field of View through the incubator portholes, etc. could disrupt a reliable measurement. This study will focus on developing a technique that measures neonates' body temperature accurately in an incubator. By eliminating unwanted external factors, continual measurement of a Region of Interest (ROI) become more feasible from which trends become available for the techniques like Artificial Intelligence, Machine Learning or Deep Learning. Moreover, this method reduces stress and discomfort for the infant. The outcome of this study is more accurate and the temperature profile of a geometric shapes or ROI over time provides a valuable input to the physicians or nurses to provide higher quality care.