Probabilistic clustering of signatures

Image from Licence Details: Probabilistic clustering of signatures

Scientists at the University of Oxford have developed machine learning tools to identify the phenotypic subgroups in patients with complex diseases using circadian hospital-summary patient data. This technology enables us to infer clusters of signatures in relation to disease and co-morbidities.

Classification based on phenotypic subgroups in patients could be used to aid effective selection of treatment plans or more efficiently detect a heightened risk of adverse medical events or abnormalities. Various approaches are known for classifying the health of a patient using measurements of vital signs, but most of them are either heuristic or assume that data is time-invariant and independent. Furthermore, such vital sign measurements commonly contain different numbers of observations, and at irregularly sampled times.

Oxford researchers have developed a range of machine learning technologies that overcome these problems. They have developed an end-to-end pipeline composed of data collection and processing techniques to generate the signatures and personalised data modelling which can serve as preventative and precision medicine. Such a pipeline could significantly improve patient outcomes and boost hospital performance.

The invention is an integrated platform for pattern recognition of patient signatures, patient phenotype discovery and patient risk stratification. There is currently no means for analysing circadian signatures of patient data in such a robust manner. This clustering method is particularly well-suited to real-world clinical data, whereby it can handle the missing aspects that typify such real-world data, and which is a key component of the tools.

Request more information
about this technology

Back to Technologies Available

Sparks Background Image

Ready to get in touch?

Contact Us
Sparks Background Image
© Oxford University Innovation