Prediction of Postoperative Complications in Cardiac Surgery

Outline

Problem Definition
Data Set
Methods
Results & Discussion
Conclusion

Problem Definition

Motivation

Huge amounts of data collected at the intensive care unit (ICU)
High

Postoperative Bleeding

Coagulation Problems:
Bleeding due to non-clotting
Treatment: transfusion (blood products)

Problem Definition -

Problem Statement

Predicting the need for surgical re-exploration due to postoperative bleeding

Related Work

Electronic Health Records (EHRs) for prediction
Mortality prediction in real-time at

Data Set

Patients

Bleeding patient: surgical re-exploration within 25 hours after initial surgery
Control group:

Features

Continuous or categorical
Static features: e.g. age, gender, initial surgery type, …

Time Slices

Time window: end of initial surgery until start of surgical

Representation
a
Time Slice Representation
a
Sequence Representation

Problem Definition - Data Set - Methods -

Methods

Clinical Baseline

Decision in favor of a surgical re-exploration, if the bleeding

Machine Learning Approaches

Naive Bayes
AdaBoost (Decision Trees)
Logistic Regression
Support Vector Machines (SVM)
K-Nearest Neighbors

Recurrent Neural Network (RNN)

x: input
s: hidden state
o: output
U, V, W:

Results & Discussion

Evaluation Metrics

P: number of positive time slices
N: number of negative time

Results

Problem Definition - Data Set - Methods - Results & Discussion

Problem Definition - Data Set - Methods - Results & Discussion

Problem Definition - Data Set - Methods - Results & Discussion

Problem Definition - Data Set - Methods - Results & Discussion

Problem Complexity and Limitations

Ground truth unknown
Real-time prediction
Missing or incorrect data
Coarse temporal

Conclusion

Conclusion

All approaches perform significantly better than the clinical baseline
RNN performs with

Thank you!

Problem Definition - Data Set - Methods - Results & Discussion

Problem Definition - Data Set - Methods - Results & Discussion

Problem Definition - Data Set - Methods - Results & Discussion

Problem Definition - Data Set - Methods - Results & Discussion

Problem Definition - Data Set - Methods - Results & Discussion

Problem Definition - Data Set - Methods - Results & Discussion

Problem Definition - Data Set - Methods - Results & Discussion

Problem Definition - Data Set - Methods - Results & Discussion

Feedforward Neural Network (FNN)

Final model:
Hidden layers: 1
Hidden nodes: 20
Activation function: sigmoid
Regularization: