Inheritance-Guided Hierarchical Assignment for Clinical Automatic Diagnosis

Clinical diagnosis, which aims to assign diagnosis codes for a patient based on the clinical note, plays an essential role in decision-making. Considering that manual could be error-prone and time-consuming, many intelligent approaches text mining have been proposed perform automatic diagnosis. However, these methods may not achieve satisfactory results due following challenges. First, most of are rare, distribution is extremely unbalanced. Second, existing challenging capture correlation between codes. Third, lengthy note leads excessive dispersion key information related To tackle challenges, we propose novel framework combine inheritance-guided hierarchical assignment co-occurrence graph propagation Specifically, joint prediction strategy address challenge unbalanced distribution. Then, utilize convolutional neural networks obtain semantic representations medical ontology. Furthermore, introduce multi attention mechanisms extract crucial information. Finally, extensive experiments MIMIC-III dataset clearly validate effectiveness our method.