Prediction of treatment medicines can assist doctors in making informed prescription decisions for patients according to their EHRs.

Predicting treatment medicines can provide insights for treatment planning and prescription management, thus improving the efficiency of clinical decision making and medical risk management.

Challenge:

  • 异质性：heterogenous nature of EHR data that typically includes laboratory results, treatment records, disease conditions and demographic information.
    • multiple sources and in different formats.
    • longitudinal information (sequences of laboratory examinations and treatment medicines) and static information (patient demographics).
  • 相关性：complex corrrelations among EHR sequences, including inter-correlations between sequences and temporal intra-correlations within each esquences.
    • examination results, treatment prescriptions(retrospectively review).
    • laboratory results and treatment medicines are mutually informative.
  • 时间动态：temporal dynamics of these correlations changing with disease progression.
    • food and drinks, mood.
  • homogeneous data(同质数据): T-LSTM[4], PacRNN[5], ATTAIN[6] (future diagnoses or recommend medications)
  • heterogeneouts events with multi-scale sampling(多尺度采样的异构事件)：HE-LSTM
  • multi-view sequential learning & encoding heterogeneous sequences: GAMENet[9], DMNC[10]
  • correlations of heterogeneous sequences: LSTM-DE[11]

Weakness

  • ignore the sequential interactions in temporal sequences
  • fail to fully exploit past records for improving prediction performance.

Prediction of Treatment Medicines With Dual Adaptive Sequential Networks

Contributions

  • data heterogeneity & time-varying correlations in logitudinal data:
    • dual adaptive sequential networks(DASNet)
    • two decomposed adaptive LSTM(DA-LSTM): capture the inter-correlations ofheterogeneous sequences through a decomposition structure with auxiliary input.
    • an attentive meta learning network(AT-MetaNet): generates parameter weights in DA-LSTM, extract meta knowleged from historical sequences of examination results and treatment medicines.
    • an attentive fusion network (AF-FuNet): fuse the embedding representations of heterogeneous information sources