Early Detection of Malpositioned Catheters and Lines on Chest X-Rays using Deep Learning

Hospital patients can have catheters and lines inserted during the course of their admission and serious complications can arise if they are positioned incorrectly. Early recognition of malpositioned tubes is the key to preventing risky complications (even death), even more so now that millions of COVID-19 patients are in the need of these tubes and lines. Earlier detection of malpositioned catheters and lines is even more important as COVID-19 cases continue to surge, these steps can be time consuming and are still prone to human error, especially in stressful situations when hospitals are at capacity This paper presents an analysis of the solution to the challenge "RANZCR CLiP - Catheter and Line Position Challenge" on Kaggle hosted by Royal Australian and NZ College of Radiologists which scores 0.972 (AUC). It is a Transfer Learning-based CNN heavily inspired by UNet and EfficientNet. This architecture stands out from the rest due to the compound scaling methods applied to achieve its smaller size and higher speed than the rest.