Deep architecture for Heparin dosage prediction during continuous renal replacement therapy

Continuous renal replacement therapy (CRRT) is the mainstream approach currently for blood purification. The process needs anticoagulation to prevent blood coagulation. Heparin, as a widely used anticoagulant, requires the doctor to give an appropriate dosage. In this paper, a new method for Heparin dosage prediction is proposed based on deep learning. The proposed deep architecture consists of two parts, i.e., a deep belief network (DBN) at the bottom and a regression layer at the top. The DBN is employed for unsupervised feature learning. It can learn effective features for the prediction task without using the recorded dosage, only using the clinical examination indexes. To incorporate task learning in the deep architecture, a regression layer is used above the DBN which uses the recorded dosage for a supervised learning. Experiments on test datasets show good performance of the deep architecture and it achieves at least 10% higher accuracy at 5% coincidence rate than other traditional prediction approaches.