Decision Support for Intoxication Prediction Using Graph Convolutional Networks

Every day, poison control centers (PCC) are called for immediate classification and treatment recommendations of acute intoxication cases. Due to their time-sensitive nature, a doctor is required to propose a correct diagnosis and intervention within a minimal time frame. Usually the toxin is known and recommendations can be made accordingly. However, in challenging cases only symptoms are mentioned and doctors have to rely on clinical experience. Medical experts and our analyses of regional intoxication records provide evidence that this is challenging, since occurring symptoms may not always match textbook descriptions due to regional distinctions or institutional workflow. Computer-aided diagnosis (CADx) can provide decision support, but approaches so far do not consider additional patient data like age or gender, despite their potential value for the diagnosis. In this work, we propose a new machine learning based CADx method which fuses patient symptoms and meta data using graph convolutional networks. We further propose a novel symptom matching method that allows the effective incorporation of prior knowledge into the network and evidently stabilizes the prediction. We validate our method against 10 medical doctors with different experience diagnosing intoxications for 10 different toxins from the PCC in Munich and show our method’s superiority for poison prediction.