Treatment selection using prototyping in latent-space with application to depression treatment

Abstract

Machine-assisted treatment selection commonly follows one of two paradigms: a fully personalized paradigm which ignores any possible clustering of patients; or a sub-grouping paradigm which ignores personal differences within the identified groups. While both paradigms have shown promising results, each of them suffers from important limitations. In this article, we propose a novel deep learning-based treatment selection approach that is shown to strike a balance between the two paradigms using latent-space prototyping. Our approach is specifically tailored for domains in which effective prototypes and sub-groups of patients are assumed to exist, but groupings relevant to the training objective are not observable in the non-latent space. In an extensive evaluation, using both synthetic and Major Depressive Disorder (MDD) real-world clinical data describing 4754 MDD patients from clinical trials for depression treatment, we show that our approach favorably compares with state-of-the-art approaches. Specifically, the model produced an 8% absolute and 23% relative improvement over random treatment allocation. This is potentially clinically significant, given the large number of patients with MDD. Therefore, the model can bring about a much desired leap forward in the way depression is treated today.

Omics approaches open new horizons in major depressive disorder: from biomarkers to precision medicine

“…With this strategy, the authors achieved a remission rate of 35.5% over random treatment allocation, achieving better results when compared to the existing state-of-the-art method (i.e. CFRnet, Vulcan, case-based recommender (CBR)) ( 227 ).…”

Artificial Intelligence in Depression – Medication Enhancement (AID-ME): A Cluster Randomized Trial of a Deep Learning Enabled Clinical Decision Support System for Personalized Depression Treatment Selection and Management

“…Extensive feasibility and ease of use testing of this CDSS was previously performed in both simulation center and in vivo feasibility studies [16][17][18][19] . With in silico testing demonstrating that the AI component should help improve remission rates 6, [20][21][22] and in vivo testing demonstrating that the platform was feasible, easy to use and likely safe [16][17][18][19] the current study was undertaken with the main objective of determining the e cacy of the platform in improving depression treatment outcomes in patients with moderate to severe depression, as well as to assess platform safety.…”

AI-assisted prediction of differential response to antidepressant classes using electronic health records

“…Lastly, in contrast to modeling predictions under the naturalistic, observational setting, modeling for interventional recommendations requires additional consideration of potential confounds that could arise from non-random treatment assignments. While a handful of studies have modeled prediction of antidepressant response using data specifically collected for research such as brain imaging or EEG, sample sizes have been typically modest and these approaches can be costly and difficult to scale 9 – 42 . Ideally, a clinically useful model would enable accurate prediction of antidepressant response, comparative predicted response for alternative treatment choices, and control for potential confounding – in particular, confounding by indication – that is, pretreatment factors associated with both the propensity to choose an antidepressant and treatment response.…”