Engineering a Model of Disability

Overview The desired outcome from this proposed line of research is a tool which is a readily interpreted representation of the multi-dimensional human condition in the context of neurological disease. The key objectives of the project and the questions we will attempt to answer, are:
1. Is it possible to define, gather and process multi-dimensional healthcare-related data in order to predict functional outcomes across physical, psychological and social domains in the field of neurological rehabilitation? 2. Can a supervised machine learning approach incorporating Bayesian inference be used as the basis for the required predictive model? 3. What other pattern recognition techniques might be applicable and is it possible to also use unsupervised approaches? 4. Does the model provide useful information for people suffering from neurological disorders by offering them a basis for adjustment to anticipated reduced functional abilities? 5. Does the model provide useful guidance to health services, solicitors and insurance companies in allocating scarce resources optimally?
We believe that the novel physical sciences/engineering methodology that will be carried out during the project is likely to be the following:
a. One of the challenges will be in the collection of the requisite amount of data -- the project will require the establishment of a platform which operates (probably via the Cloud) in a non-invasive and easy-to-use fashion. b. The processing of the data will necessarily investigate current state of the art algorithms in both machine learning and data fusion. Ideally we can adopt a Bayesian approach to both to form a novel integrated analysis system. c. Feature extraction may be an important part of the analysis, and choice of an optimal set of relevant features will perhaps have translational uses in other learning/classification problems.
The principal domains will be functional and will relate to biological, psychological and social parameters affecting a person disabled by a neurological condition. Functional parameters in this context means what the person cannot do, e.g. stand, walk, remember, plan, speak, feel, wash, dress, eat or work. The model will provide a consistent currency for analysis of individuals and populations and will help disabled people to know whether is it reasonable, given the stage of their disease and the availability of rehabilitation techniques, to feel and function as they do. It will help to provide patients with a better understanding of their condition and will enable more targeted treatment by making objective assessment more accessible and explicit. By representing cohorts within the population it will also provide clinicians with a tool for managing competing demands for resources. This is particularly important in a UK healthcare system which is under severe pressure, especially in primary care, and is showing signs of failing under the weight of demand from patients with chronic conditions. In particular it is anticipated that the model will prove useful in discriminating between the following groups:
i. people who do not have a neurological condition and are ``functionally normal" ii. those with``hard pathology" and consequent functional impairment e.g. Multiple Sclerosis, Complex Regional Pain Syndrome, or Acquired Brain Injury, iii. those with``functional problems" where no hard pathology can be identified despite extensive (and often completely unnecessary) investigation, but where functional abnormalities present e.g Chronic Fatigue Syndrome and Fibromyalgia. Motion capture and analysis will produce datasets which it is anticipated will be fundamental to the representation of function. Other contextual data about the biological, psychological and social impact of disabling neurological conditions will be sourced directly from patients via a web interface where they will be guided by structured open questions to provide descriptions rich in detail.