Infrared Imaging and Classification of Neonates with Necrotising Enterocolitis

Abstract

Necrotising Enterocolitis (NEC) is a devastating intestinal disease associated with high rate of mortality and long-term morbidity. It can be successfully treated if diagnosed early, but there is no reliable way to detect NEC in early stages. Infrared imaging can detect tissue inflammation and thus can offer an early diagnostic tool for NEC. We enrolled infants with no clinical or radiographic signs of NEC, and a group consisting of infants with evidence of at least Bell’s Stage 2 NEC. The infants underwent bedside infrared imaging for 60 seconds. Our dataset consisted of 20 normal infants and 9 infants with NEC. In infants with NEC, the upper-to-lower (UL) region temperatures differed significantly, whereas there was no significant difference in the UL quadrant temperatures in normal infants (p = 0.0037 Wilcoxon Rank-Sum and Kruskal-Wallis tests). We found no significant differences in left-to-right (LR) quadrant abdominal temperature profiles for both the normal and the NEC group. The performance of the Decision Tree classifier was highest when using the medians of UL quadrant temperatures; the mean specificity, sensitivity, and standard deviations for ten trials with medians and means were respectively: 90&+/-12%; 78%+/-18%; 88%+/-14%; 69%+/-12%. We conclude that abdominal infrared imaging of preterm infants yields thermal profiles that can be analyzed and classified using statistical methods and decision trees to identify infants with and without NEC. Future work includes automating the analysis, conducting a prospective study to attempt detecting NEC at earlier stages, and assessing other image analysis approaches to enhance the overall performance of our methodology.