Learning to Diagnose with LSTM Recurrent Neural Networks

摘要: Clinical medical data, especially in the intensive care unit (ICU), consist of multivariate time series observations. For each patient visit (or episode), sensor data and lab test results are recorded patient's Electronic Health Record (EHR). While potentially containing a wealth insights, is difficult to mine effectively, owing varying length, irregular sampling missing data. Recurrent Neural Networks (RNNs), particularly those using Long Short-Term Memory (LSTM) hidden units, powerful increasingly popular models for learning from sequence They effectively model length sequences capture long range dependencies. We present first study empirically evaluate ability LSTMs recognize patterns clinical measurements. Specifically, we consider multilabel classification diagnoses, training classify 128 diagnoses given 13 frequently but irregularly sampled First, establish effectiveness simple LSTM network modeling Then demonstrate straightforward effective strategy which replicate targets at step. Trained only on raw series, our outperform several strong baselines, including multilayer perceptron trained hand-engineered features.