Anne Mottram

TL;DR: A deep learning approach that predicts the risk of acute kidney injury and provides confidence assessments and a list of the clinical features that are most salient to each prediction, alongside predicted future trajectories for clinically relevant blood tests are developed.

TL;DR: In this paper, a formal problem definition, data pre-processing, architecture selection, calibration and uncertainty, and generalizability evaluation is described for developing deep-learning risk models that can predict various clinical and operational outcomes from structured electronic health record (EHR) data.

TL;DR: This work proposes an extension of Testing with Concept Activation Vectors to time series data to enable an application of TCAV to sequential predictions in the EHR, and evaluates the proposed approach on an open EHR benchmark from the intensive care unit, as well as synthetic data where it is able to better isolate individual effects.

TL;DR: In this paper, a sequential sub-network routing (SeqSNR) architecture is proposed to find related tasks and encourage cross-learning between them. But the SeqSNr architecture suffers from negative transfer - impaired learning if tasks are not appropriately selected.

TL;DR: In this paper, an extension of the concept activation vector (TCAV) method to time series data is proposed to enable an application of TCAV to sequential predictions in the EHR, and they evaluate the proposed approach on an open EHR benchmark from the intensive care unit, as well as synthetic data where they are able to better isolate individual effects.