inTrigue Data Science components
Volv has a number of additional capabilities that make up the inTrigue methodology, enriching the insights and learnings produced by the core predictive modelling.
inTrigue Data Science components
Volv has a number of additional capabilities that make up the inTrigue methodology, enriching the insights and learnings produced by the core predictive modelling.
AI & Data Science
Volv’s approach is unique and provides state of the art performance defining cohorts more accurately…
We are specialists at managing real world data both structured and unstructured to build robust predictive models for better patient outcomes.
Developing computational models capable of detecting rare disease patients in population-scale databases such as electronic health records (EHRs) is challenging for several reasons, perhaps the most daunting of which being the limited number of already-diagnosed, ‘labelled’ patients from which to learn. Often this number is zero.
We overcome this obstacle with a novel, lightly-supervised algorithm that leverages unlabelled and/or unreliably-labelled patient data – which is typically plentiful – to facilitate model induction.
Importantly, the Volv algorithm is safe. Adding unlabelled/unreliably-labelled data to the learning procedure produces models which are usually more accurate and guaranteed never to be less accurate than models learned from reliably-labelled data alone.
Our methods are shown to substantially outperform state-of-the-art models in patient-finding experiments involving two different rare diseases and a country-scale EHR database.
Additionally, we can transform high-performance 'black box' models generated in this way into simpler models which, while still accurate, are human interpretable.
Model external medical knowledge
Volv models the semantics of target rare diseases and their symptoms and comorbidities via representation learning with large medical corpora to render a highly dimensional computational model.
Build robust predictive models with no labels
Volv uses combinations of unsupervised and lightly supervised machine learning methods in a highly dimensional feature space, which would typically require 50,000 labels to produce the same results. Often volv has no labels at all to work with.
Validated prevalence
Throughout the inTrigue process we maintain the ability to get external validation of our results and methods. Specialist dlinicians are involved in the assessment of the patients being flagged as at risk, which delivers empirical and validated prevalence results.
Novel Biomarker Discovery
inTrigue learns new things about the disease like symptoms, comorbidities, natural history and will also define new predictive biomarkers that can be cognitive, digital and physical.
Clinically Interpretable Models
It is desirable to construct prediction models which are both accurate and interpretable. It is essential that clinicians understand the basis for the predictions and recommendations of decision-support systems. Volv is able to transform its complex 10,000 feature models into human interpretable models that maintain similar accuracy.
Transfer learn and deploy models in diverse healthcare systems
Volv uses unique transfer learning techniques to adapt the models in different data sets so that they can be deployed and maintained in multiple heterogenous settings.
