Temporal Convolutional Neural Networks for Diagnosis from Lab Tests
This work addresses the problem of early diagnosis of treatable diseases from sparse lab data for healthcare applications, representing an incremental improvement with a novel method for a known bottleneck.
The authors tackled early disease diagnosis from irregular lab tests by introducing a multi-resolution convolutional neural network that processes imputed data and binary observation matrices, achieving significantly better predictive performance than existing baselines on a dataset of 298K individuals over 8 years for 171 diseases.
Early diagnosis of treatable diseases is essential for improving healthcare, and many diseases' onsets are predictable from annual lab tests and their temporal trends. We introduce a multi-resolution convolutional neural network for early detection of multiple diseases from irregularly measured sparse lab values. Our novel architecture takes as input both an imputed version of the data and a binary observation matrix. For imputing the temporal sparse observations, we develop a flexible, fast to train method for differentiable multivariate kernel regression. Our experiments on data from 298K individuals over 8 years, 18 common lab measurements, and 171 diseases show that the temporal signatures learned via convolution are significantly more predictive than baselines commonly used for early disease diagnosis.