The Panacea Project looks to develop and interate a highly accurate machine learning model capable of predicting the probability that any one patient has any disease (using ICD9 codes). It looks to merge concepts of public/population health with personalized medicine through the Reciprocal Perspective framework.
Implements the following technologies:
- Latent Dirichlet Allocation for Modelling Unstructured Clinical Notes
- Stacked Denoising Autoencoders for Dimensionality Reduction
- Multi-Layer Perceptron to Predict the Likelihood accross all ICD9 Codes
- Reciprocal Perspective for Post-Processing the Predictions to Improve Prediction Accuracty