Machine Learning

Home / Applications / Machine Learning

Tools for medicine, healthcare and HEOR research

Disease modeling and evidence-based decision support

 

We apply Bayesian networks and influence diagrams to support decision making under uncertainty and complexity.

We have experience in applications designed for:

 

  • randomized clinical trials
  • real-world datasets
  • patient-reported outcomes
  • population surveys

Features for health economics and outcomes research

  • Prioritizing transparency, interpretability and the human-facing nature of analytics for health research
  • Handling multiple outcomes, large or small dataset size and missingness in a unified model
  • Incorporating subjective domain expertise
  • Providing decision support under uncertainty for diagnosis and treatment plan selection
  • Conducting causal and counterfactual analysis
  • Supporting public education and dissemination of research

Personalized prognosis and health intelligence for precision medicine

 

We apply graphical models, probabilistic neural networks, random forests, influence diagrams and Markov process models

We provide solutions for problems in analytics to extract maximum value from:

 

  • clinical trials and registries
  • health insurance records
  • electronic health records
  • population surveys

Generated results

  • Updatable, interpretable and scalable probabilistic models
  • Combinable multiple formats and types of data
  • Robust risk prediction and subgroup selection
  • Multi-objective treatment plan selection models
  • Incorporation of subjective values from the patient, physician and health insurer
  • Value of information analyses and recommender systems

Biomedical signal processing

 

We work with convolutional/ recurrent neural networks, probabilistic graphical models and Markov process models

We work with firms and health systems on how to integrate data from:

 

  • medical images
  • patient registries
  • hospitalization records
  • electrocardiogram data
  • wearables (pulse rate, sleep activity)

Features for analysis

  • Time series, sequences (eg. waveform data) and images with a focus on interpretability and model criticism
  • Automated medical diagnosis and disease management from point-of-care devices and smartphone-based sensors
  • Anomaly detection in biomedical signals and clinical workflows to produce actionable insights for clinical decision support and intelligent monitoring systems research

Contact us

Canada
1 University Avenue. 3rd Floor.
Toronto, Ontario.
M5J 2P1. Canada
1-800-535-9760

Inquiries
info@lighthouseoutcomes.com