Because premature infants have an immature thermoregulatory function, proper temperature control with an incubator is essential. Currently, the probe is attached to the skin of the infants to measure the body temperature. However, the probe may come off easily and long-term measurement may be difficult because the skin of infants is immature and the limbs also move. Therefore, we try to measure body temperature by using a thermography, which is a non-contact device. It is possible to measure the body temperature of the whole body without giving stress to the infants. In order to perform non-invasive temperature measurement, we need to estimate the position and measurement region of the infants from the images. We perform binary classification during intervention (treatment by medical staff)/ non-intervention by using deep learning. We also construct detection model of six body parts (head/ chest/ upper right limb/ upper left limb/ lower right limb/ lower left limb). Then, after applying clustering to separate the infants region and other areas (reflection of blankets, medical devices, etc.), we extract the body surface temperature of the neck and limb based on the detected body parts. We are proceeding with the analysis using data provided by medical institutions conducting joint research. In the future, we will investigate the relationship between body surface temperature and core body temperature, and estimate the core body temperature so as to control incubator temperature automatically.
premature infants/thermal images/parts detection/temperature extraction/deep learning