Inferring disease subtypes from clusters in explanation space.

Identification of disease subtypes and corresponding biomarkers can substantially improve clinical diagnosis and treatment selection. Discovering these subtypes in noisy, high dimensional biomedical data is often impossible for humans and challenging for machines. We introduce a new approach to facilitate the discovery of disease subtypes: Instead of analyzing the original data, we train a diagnostic classifier (healthy vs. diseased) and extract instance-wise explanations for the classifier's decisions. The distribution of instances in the explanation space of our diagnostic classifier amplifies the different reasons for belonging to the same class-resulting in a representation that is uniquely useful for discovering latent subtypes. We compare our ability to recover subtypes via cluster analysis on model explanations to classical cluster analysis on the original data. In multiple datasets with known ground-truth subclasses, particularly on UK Biobank brain imaging data and transcriptome data from the Cancer Genome Atlas, we show that cluster analysis on model explanations substantially outperforms the classical approach. While we believe clustering in explanation space to be particularly valuable for inferring disease subtypes, the method is more general and applicable to any kind of sub-type identification.

Recruitment and Ongoing Engagement in a UK Smartphone Study Examining the Association Between Weather and Pain: Cohort Study.

BACKGROUND: The huge increase in smartphone use heralds an enormous opportunity for epidemiology research, but there is limited evidence regarding long-term engagement and attrition in mobile health (mHealth) studies. OBJECTIVE: The objective of this study was to examine how representative the Cloudy with a Chance of Pain study population is of wider chronic-pain populations and to explore patterns of engagement among participants during the first 6 months of the study. METHODS: Participants in the United Kingdom who had chronic pain (≥3 months) and enrolled between January 20, 2016 and January 29, 2016 were eligible if they were aged ≥17 years and used the study app to report any of 10 pain-related symptoms during the study period. Participant characteristics were compared with data from the Health Survey for England (HSE) 2011. Distinct clusters of engagement over time were determined using first-order hidden Markov models, and participant characteristics were compared between the clusters. RESULTS: Compared with the data from the HSE, our sample comprised a higher proportion of women (80.51%, 5129/6370 vs 55.61%, 4782/8599) and fewer persons at the extremes of age (16-34 and 75+). Four clusters of engagement were identified: high (13.60%, 865/6370), moderate (21.76%, 1384/6370), low (39.35%, 2503/6370), and tourists (25.44%, 1618/6370), between which median days of data entry ranged from 1 (interquartile range; IQR: 1-1; tourist) to 149 (124-163; high). Those in the high-engagement cluster were typically older, whereas those in the tourist cluster were mostly male. Few other differences distinguished the clusters. CONCLUSIONS: Cloudy with a Chance of Pain demonstrates a rapid and successful recruitment of a large, representative, and engaged sample of people with chronic pain and provides strong evidence to suggest that smartphones could provide a viable alternative to traditional data collection methods.

Artefact in Physiological Data Collected from Patients with Brain Injury: Quantifying the Problem and Providing a Solution Using a Factorial Switching Linear Dynamical Systems Approach.

INTRODUCTION: High-resolution, artefact-free and accurately annotated physiological data are desirable in patients with brain injury both to inform clinical decision-making and for intelligent analysis of the data in applications such as predictive modelling. We have quantified the quality of annotation surrounding artefactual events and propose a factorial switching linear dynamical systems (FSLDS) approach to automatically detect artefact in physiological data collected in the neurological intensive care unit (NICU). METHODS: Retrospective analysis of the BrainIT data set to discover potential hypotensive events corrupted by artefact and identify the annotation of associated clinical interventions. Training of an FSLDS model on clinician-annotated artefactual events in five patients with severe traumatic brain injury. RESULTS: In a subset of 187 patients in the BrainIT database, 26.5 % of potential hypotensive events were abandoned because of artefactual data. Only 30 % of these episodes could be attributed to an annotated clinical intervention. As assessed by the area under the receiver operating characteristic curve metric, FSLDS model performance in automatically identifying the events of blood sampling, arterial line damping and patient handling was 0.978, 0.987 and 0.765, respectively. DISCUSSION: The influence of artefact on physiological data collected in the NICU is a significant problem. This pilot study using an FSLDS approach shows real promise and is under further development.