Home page of the Decision-Support Lab
Part of the QMUL Risk and Information Management (RIM) research group in EECS.
More infoThe decision-support lab is led by Dr William Marsh and is a part of the Risk and Information Management (RIM) research group in EECS. We aim to develop practical techniques for decision-support with probabilistic and causal models (primarily using Bayesian networks). Much of the work is collaborative with application to medicine and in engineering.
To view our projects from the Decision Support Lab and collaborators see below.
Bayesian Network Decision Support
The BNDS server facilitates the uploading, updating, downloading of Bayesian Network models via an dedicated API. Inference is performed using the gRain package in R together with the plumber API.
In addition to models, customisable interfaces can be uploaded to specify the look and feel of the probability outputs associated to the models.
Optimising Shared decision-makIng for high-RIsk major Surgery
The OSIRIS programme is a major project of research, to understand and improve the shared decision making process for patients at high risk of medical complications as they contemplate major surgery.
The OSIRIS Web Portal hosts the forecast models that will be used by clinicians in the associated trial.
Bayesian Networks for Musculoskeletal Conditions
The BeNDi process used a Delphi method to construct a Bayesian Network (BN) for predicting the presence of serious underlying pathologies in patients presenting with lower back pain.
The final model has been uploaded to our server, where you can test out various predictions. The site also includes some questionnaires for testing.
Decision Support Lab Git repositories
A list of repositories from the Decision-Support Lab.
These contain the code bases for the projects listed here, as well as many others related to the group.
Predicting trauma induced complications
Models designed in collaboration with trauma surgeons from the Royal London Hospital.
The models aim to predict the likelihood of trauma induced complications, such as coagulopothy, in patients in both civilian and military settings.
We also have evidence browsers associated to Trauma Induced Coagulopothy (TIC) and Limb viability models.
Uncovering how predictions are made
Modern black-box type models make it extremely hard to understand why certain predictions have been made.
BNs offer ways to illuminate the black box but doing so can be time-consuming. We are developing algorithms to automatically provided reasoned explanations of predictions to aid decision makers. See below for a demo.
Using decision making analysis for clinical practice
Can we make decision support systems which patients and clinical practioners use collaboratively?
Using methods from the field of multi-criteria-decision-analysis we have designed a tool to enhance shared decision making (SDM) in the field of MSK. View the demo below.