A carregar...
Publicação
Predicting the development of diabetic nephropathy to simulate treatments using deep learning approaches
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 665.36 KB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
Diabetic Nephropathy is a chronical condition leading towards kidney failure. Since the disease is diagnosed in terms of stages, learning disease progression has important implications for health risk prediction. This paper leverages state-of-the-art deep learning models including the Stage-Aware Neural Network Stage Net, working with large-scale patient Electronic Health Records (EHR). It addresses the sparsity of clinical data using an approximate nearest neighbor method to impute missing values. The results show that the Stage Net out performs other machine learning models in predicting the Nephropathy stage. A counterfactual analysis is performed to simulate possible treatments and their associated effect on the disease progression.
Descrição
Palavras-chave
Healthcare informatics Electronic health Records Disease progression prediction Treatment simulation