Detection of posttraumatic pneumothorax using electrical impedance tomography-An observer-blinded study in pigs with blunt chest trauma.

INTRODUCTION:Posttraumatic pneumothorax (PTX) is often overseen in anteroposterior chest X-ray. Chest sonography and Electrical Impedance Tomography (EIT) can both be used at the bedside and may provide complementary information. We evaluated the performance of EIT for diagnosing posttraumatic PTX in a pig model. METHODS:This study used images from an existing database of images acquired from 17 mechanically ventilated pigs, which had sustained standardized blunt chest trauma and had undergone repeated thoracic CT and EIT. 100 corresponding EIT/CT datasets were randomly chosen from the database and anonymized. Two independent and blinded observers analyzed the EIT data for presence and location of PTX. Analysis of the corresponding CTs by a radiologist served as reference. RESULTS:87/100 cases had at least one PTX detected by CT. Fourty-two cases showed a PTX > 20% of the sternovertebral diameter (PTXtrans20), whereas 52/100 PTX showed a PTX>3 cm in the craniocaudal diameter (PTXcc3), with 20 cases showing both a PTXtranscc and a PTXcc3. We found a very low agreement between both EIT observers considering the classification overall PTX/noPTX (κ = 0.09, p = 0.183). For PTXtrans20, sensitivity was 59% for observer 1 and 17% for observer 2, with a specificity of 48% and 50%, respectively. For PTXcc3, observer 1 showed a sensitivity of 60% with a specificity of 51% while the sensitivity of observer 2 was 17%, with a specificity of 89%. By programming a semi-automatized detection algorithm, we significantly improved the detection rate of PTXcc3, with a sensitivity of 73% and a specificity of 70%. However, detection of PTXtranscc was not improved. CONCLUSION:In our analysis, visual interpretation of EIT without specific image processing or comparison with baseline data did not allow clinically useful diagnosis of posttraumatic PTX. Multimodal imaging approaches, technical improvements and image postprocessing algorithms might improve the performance of EIT for diagnosing PTX in the future