Biography
I am a mathematiciam, statistician and computer scientist based in the Disease Dynamics Unit with over 30-years experience in Artificial Intelligence (AI).
Following a degree in chemistry, I specialised in NMR spectroscopy. I undertook a PhD in statistical and computational chemistry at Leeds University (i.e., the computational diagnosis of endocrine metabolites) and was then appointed as a Research Fellow at St Thomas' Hospital/King's College London, where I undertook AI and statistical research into a variety of medical areas, including clinical microbiology, intensive care, and ophthalmology.
I taught AI as a University Lecturer in computer science at the University of East London. I also wrote the first paper on Machine Learning to be published in "The Lancet" and the seminal book "Clinical Applications of Artificial Neural Networks" (published in 2001).
More information is available from https://dybowski.org/
Research
My research interests lie within the areas of (a) stochastic models of biomedical phenomena, (b) machine learning (particularly deep learning), and (c) normative clinical reasoning.
A. Neural-based computation.
I have been involved in the development of neural networks for medicine for 20 years. I am interested in two aspects of neural-based computing for biology and biomedicine : (a) convolutional neural networks for deep learning, and (b) neural networks for sequence analysis, including neural Turing machines.
B. Logics for clinical decision making.
I have a long-standing interest in the development of clinical diagnostic decision-support systems both via the use of machine-learning techniques (including deep learning) and through the use of formal methods for the fusion of knowledge with data for decision support (e.g., by means of probabilistic graphical models). I am currently focusing on a theoretical comparison of probabilistic logic and various alternative logics that have been proposed for clinical decision making, including possibility theory, Dempster-Shafer theory, and Jøsang's subjective logic.
C. Within-host dynamics of bacterial infections with respect to antibiotic intervention.
I am interested in mathematical and computational techniques that, in conjunction with observed data, reveal the dynamics of within-host infections caused by Salmonella enterica. Application of Bayesian inference is of particular interest in this regard, such as the use of nested sampling and approximate Bayesian computation. The intention is to extend the modelling process to include the known pharmacodynamics and pharmacokinetics of a range of antibiotics with respect to S. enterica in order to ascertain optimal antibiotic regimens. In this context, I am also looking at the potential of probabilistic programming.
Publications
Dybowski, R. (2017) “The state of deep learning for the detection of diabetic retinopathy”, Proceedings of the Second Swedish Diabetic Summit (SSDS), Gothenburg: Sahlgrenska Academy, p 41.
Price, D.J., Breuze, A., Dybowski, R., Restif, O. (2017) “An efficient moments-based inference method for within-host bacterial infection dynamics”, PLoS Computational Biology, 13 (11): e1005841.
Dybowski R., Restif O., Price D.J., Mastroeni P. (2017) "Inferring within-host bottleneck size: a Bayesian approach", Journal of Theoretical Biology, 435: 218-228.
Rossi O., Dybowski R., Maskell D.J., Grant A.J., Restif O., Mastroeni P. (2017) "Within-host spatiotemporal dynamics of systemic Salmonella infection during and after antimicrobial treatment", Journal of Antimicrobial Chemotherapy, dkx294.
Dybowski R., Restif O., Goupy A., Maskell D.J., Mastroeni P., Grant A.J. (2015) "Single passage in mouse organs enhances the survival and spread of Salmonella enterica", Journal of the Royal Society Interface, 12:20150702.
Coward C., Restif O., Dybowski R., Grant A., Maskell D., Mastroeni P. (2014) "The effects of vaccination and immunity on bacterial infection dynamics in vivo". PLoS Pathogens, 10(9): e1004359.
Dybowski R., McKinley T.J., Mastroeni P., Restif O. (2013) "Nested sampling for Bayesian model comparison in the context of Salmonella disease dynamics". PLoS One, 8(12): e82317.
Mahroo O.A., Dybowski R., Wong R., Williamson T.H. (2012) "Characteristics of rhegmatogenous retinal detachment in pseudophakic and phakic eyes". Eye, 26(8): 1114-1121
Farnaud S., Wang Z., Dybowski R., Evans R.W., Odell E.W. (2006) "Towards a tree-induction approach to antimicrobial activity analysis of cationic peptides". Biochemistry and Cell Biology, 84(3): 394.
Dybowski R., Roberts S. (2005) "An anthology of probabilistic models for medical informatics". In Husmeier D., Dybowski R., Roberts S. (eds.) Probabilistic Modeling in Bioinformatics and Medical Informatics. London: Springer Verlag, pp 297-349.
Dybowski, R., Weller P.R. (2001) "Prediction intervals for the visualization of incomplete datasets". Computational Statistics, 16(1): 25-41.
Dybowski, R., Roberts, S. (2001) "Confidence intervals and prediction intervals for feed-forward neural networks". In Dybowski R, Gant V. (eds.) Clinical Applications of Artificial Neural Networks. Cambridge: Cambridge University Press, pp 298-326.
Ramoni, M., Sebastiani, P., Dybowski, R. (2001) "Robust outcome prediction of intensive-care patients". Methods of Information in Medicine, 40(1): 39-45.
Dybowski, R. (1998) "Classification of incomplete feature vectors by radial basis function networks". Pattern Recognition Letters, 19: 1257-1264.
BOOKS
Husmeier D., Dybowski R., Roberts S. (eds.) Probabilistic Modeling in Bioinformatics and Medical Informatics. London: Springer Verlag, 2005.
Dybowski R., Gant V. (eds.) Clinical Applications of Artificial Neural Networks. Cambridge: Cambridge University Press, 2001.
