Fully-funded PhD in Deep Learning for Health Text Analysis: Developing Models to Predict Core Outcomes from Multiple Medical Textual Records
288 views
Skip to first unread message
Danushka Bollegala
Feb 20, 2018, 5:28:04 AM2/20/18
to Machine Learning News
LINK: https://www.findaphd.com/search/ProjectDetails.aspx?PJID=95398
Project Description
Applications are invited for a PhD scholarship in "Deep Learning for Health Text Analysis" at the University of Liverpool.
The position will be based in University of Liverpool’s Department of Computer Science and jointly supervised by Dr Danushka Bollegala and Prof. Paula Williamson from the Department of Biostatistics.
About the project:
Electronic patient records and patient narratives from a variety of sources provide useful information about how a particular patient has benefitted, or not, from a healthcare intervention, or what aspects of their condition they would like an intervention to help improve. Extracting this information is extremely useful to determine whether the patient has experienced the expected outcomes of a treatment or not.
Core outcome sets (COS) represent the minimum set of outcomes that should be measured and reported in all clinical trials of a specific condition, and are also suitable for use in routine care, clinical audit and research other than randomised trials. By measuring these core outcomes, we can objectively evaluate the success of a particular treatment, or compare the quality of care in different places.
Different features can be extracted from patient records and narratives for describing the core outcomes such as words used by the patient to describe the medical state, drugs prescribed and various medical tests conducted. However, it is unknown which features are useful for identifying a particular core outcome for a given health condition. Moreover, the vernacular vocabulary used in patient narratives to express outcomes is often different from the technical terms used by the medical experts in patient records. The mapping between the different vocabularies used to describe outcomes is non-obvious. It is often the case that a complex combination of different features is related to a single outcome. It remains a challenging task to disentangle these features and automatically learn them from large text collections.
In this project, we will use deep learning-based natural language processing methods to develop models to predict core outcomes from multiple medical textual records using minimally supervised approaches. Specifically, we model the problem of core outcome prediction as an information extraction problem, where the latent factors related to a particular extraction is an unknown prior that must be discovered from unlabelled text collections. We will resort to weakly supervised approaches similar to the distant learning techniques used in information extraction to overcome the problem of lack of labelled data to train the deep learning models. The core outcome predictors we build will be evaluated for their accuracy to predict outcomes for unseen health conditions.
This is an interdisciplinary PhD project that will enable the student to not only conduct theoretical research in deep learning but also to apply the theoretical knowledge to solve a real-world problem, which could impact millions of patients across the world.
Prior experience:
Candidates must have a good first degree in Computer Science, or closely related subject, and have excellent programming skills. Some prior experience of natural language processing or machine learning is highly desirable. Previous experience of carrying out research would be a distinct advantage though it not essential.
Applications:
Applications can be made by following the University of Liverpool’s standard process, details of which are available here: https://www.liverpool.ac.uk/computer-science/postgraduate/phdstudy/
Applications should list Dr Danushka Bollegala as the potential supervisor and choose the option "Department funded PhD" when asked how you will fund the PhD.
Applications should also be clearly marked as being for the scholarship in "Deep Learning for Health Text Analysis"
Funding Notes
The scholarship will be for 3 years at £20,000 per year covering both fees and living expenses.
a) For UK/EU applicants the current fee is £4,195 per year and the remainder of £15,805 will be for covering living expenses.
b) For the overseas/international applicants the current fee is £18,900 and only £1,100 is left for covering living expenses. Therefore, for the overseas applicants will need to demonstrate that they have sufficient funding from elsewhere to support their living expenses.
Project Description
Applications are invited for a PhD scholarship in "Deep Learning for Health Text Analysis" at the University of Liverpool.
The position will be based in University of Liverpool’s Department of Computer Science and jointly supervised by Dr Danushka Bollegala and Prof. Paula Williamson from the Department of Biostatistics.
About the project:
Electronic patient records and patient narratives from a variety of sources provide useful information about how a particular patient has benefitted, or not, from a healthcare intervention, or what aspects of their condition they would like an intervention to help improve. Extracting this information is extremely useful to determine whether the patient has experienced the expected outcomes of a treatment or not.
Core outcome sets (COS) represent the minimum set of outcomes that should be measured and reported in all clinical trials of a specific condition, and are also suitable for use in routine care, clinical audit and research other than randomised trials. By measuring these core outcomes, we can objectively evaluate the success of a particular treatment, or compare the quality of care in different places.
Different features can be extracted from patient records and narratives for describing the core outcomes such as words used by the patient to describe the medical state, drugs prescribed and various medical tests conducted. However, it is unknown which features are useful for identifying a particular core outcome for a given health condition. Moreover, the vernacular vocabulary used in patient narratives to express outcomes is often different from the technical terms used by the medical experts in patient records. The mapping between the different vocabularies used to describe outcomes is non-obvious. It is often the case that a complex combination of different features is related to a single outcome. It remains a challenging task to disentangle these features and automatically learn them from large text collections.
In this project, we will use deep learning-based natural language processing methods to develop models to predict core outcomes from multiple medical textual records using minimally supervised approaches. Specifically, we model the problem of core outcome prediction as an information extraction problem, where the latent factors related to a particular extraction is an unknown prior that must be discovered from unlabelled text collections. We will resort to weakly supervised approaches similar to the distant learning techniques used in information extraction to overcome the problem of lack of labelled data to train the deep learning models. The core outcome predictors we build will be evaluated for their accuracy to predict outcomes for unseen health conditions.
This is an interdisciplinary PhD project that will enable the student to not only conduct theoretical research in deep learning but also to apply the theoretical knowledge to solve a real-world problem, which could impact millions of patients across the world.
Prior experience:
Candidates must have a good first degree in Computer Science, or closely related subject, and have excellent programming skills. Some prior experience of natural language processing or machine learning is highly desirable. Previous experience of carrying out research would be a distinct advantage though it not essential.
Applications:
Applications can be made by following the University of Liverpool’s standard process, details of which are available here: https://www.liverpool.ac.uk/computer-science/postgraduate/phdstudy/
Applications should list Dr Danushka Bollegala as the potential supervisor and choose the option "Department funded PhD" when asked how you will fund the PhD.
Applications should also be clearly marked as being for the scholarship in "Deep Learning for Health Text Analysis"
Funding Notes
The scholarship will be for 3 years at £20,000 per year covering both fees and living expenses.
a) For UK/EU applicants the current fee is £4,195 per year and the remainder of £15,805 will be for covering living expenses.
b) For the overseas/international applicants the current fee is £18,900 and only £1,100 is left for covering living expenses. Therefore, for the overseas applicants will need to demonstrate that they have sufficient funding from elsewhere to support their living expenses.
Reply all
Reply to author
Forward
0 new messages