Necrotizing enterocolitis (NEC) is a disease that leads to inflammation in the intestinal tissue of premature babies. In this paper, we present a novel automated image acquisition and processing system that integrates infrared and RGB-D sensors for NEC detection. Intersensor calibration and data registration are introduced to ensure the consistency of depth, color and infrared images captured by the multispectral sensor. Segmentation of a baby’s torso area is automatically achieved over the infrared imagery while relying on depth and color data to entirely retrieve the region of interest. Analysis of thermal distribution over the whole area reduces the risk of missing key information due to manual intervention. Preliminary results obtained with this multispectral imaging approach for NEC diagnosis are encouraging.
Investigator, CHEO Research Institute