Purpose or Case Report: Pediatric abdominal trauma can be challenging and often necessitates rapid and accurate identification of injuries for optimal patient care. This study aimed to assess the potential workflow improvements in diagnostic accuracy achieved with semi-automated analysis of Dual Energy CT (DECT) image reconstruction in the setting of abdominal solid organ injury, as compared to conventional Multi-Detector CT (MDCT) images. The data sets from dual energy reconstructions have superior tissue differentiation and can be used to train a simple artificial intelligence (AI) program to recognize these patterns and expedite care.
Methods & Materials: We retrospectively reviewed a cohort of newborn to 17-year-old trauma patients who underwent imaging between January 1, 2016 and October 4, 2023 at a single large pediatric hospital which is also a pediatric level one trauma center. Imaging was performed using DECT as part of routine clinical practice. Data from the iodine map, monoenergetic images, and conventional images was extracted and analyzed by specialized software which allows for quantitative assessment of parenchymal voxel histograms for normal and injured solid organs. Data was analyzed using common image analysis techniques including thresholding, edge detection, histogram analysis, and morphological operations such as erosion, dilation, noise reduction, feature enhancement, and object extraction. The data is then compared between the conventional, mono-energetic, and iodine map images. These monoenergetic images, iodine map, and conventional CT images were then used to train multiple simple comparative AI models to discriminate between normal and abnormal abdominal solid organs. These models are then compared for sensitivity, specificity, and receiver operator characteristics.
Results: Analysis of DECT techniques, specifically the iodine map, results in a histogram with bimodal distribution in positive cases of solid organ injury with injured and normal parenchyma having clearly separate peaks. Additionally, the AI model trained with the iodine map data demonstrated preliminarily superior differentiation of injured versus non-injured parenchyma.
Conclusions: Dual Energy CT exhibits an advantage over conventional MDCT with additional layers of data that can result in improved accuracy in the diagnosis of pediatric abdominal trauma. This data has potential for artificial intelligence to rapidly identify solid organ injury and potentially improve patient outcomes.