*4 Year MRC PhD Programme* Crowdsourced by Health-Care Professionals: combining psychometrics and deep learning for Citizen Science approaches in health education, health-care information or evaluation

This project is proposed under the auspices of two cross-school initiatives.

iCIGHT (informatics, Clinical Imaging, Genomics and Healthcare Technologies) (Trucco, Doney, Mountain)
DCHARR, the Dundee Centre for Health And Related Research (Croudace and Moncur)
Supervisors:

1 Professor Emanuele Trucco, Computing, School of Science and Engineering, and iCIGHT

2 Professor Tim Croudace, Director, DCHARR and School of Nursing and Health Sciences (3, all year 1)

3 Dr W Moncur, Duncan of Jordanstone College of Art & Design

Health education and health-care research makes extensive use of instruments for assessment and evaluation. Crowdsourcing has become a popular and promising approach to improve instrument development as well as appearing in situations involving digital annotation (1, 2). The latter was investigated in a recent study on crowdsourcing for assessing diabetic retinopathy in fundus camera images, involving two of the proposers of this project and in collaboration with UCL (1).

We are offering the opportunity to work in an inter-disciplinary project team to pioneer crowdsourcing among Dundee healthcare professions in education or evaluation settings.

We intend to leverage our experience in psychometrics (educational statistics), computing (deep learning) and design of digital interaction to evaluate opportunities, prototype and perform critical evaluations: candidates will be able to motivate their topic area (education/evaluation) but options span a wide remit for this project.

Using social digital research approaches and existing crowdsourcing platforms the project will also compare technical methods for datasets comparing deep / machine learning with adaptive psychometrics (3). Issues of platform, software and digital design will be integrated, i.e. the student will tackle design issues in the collection of crowdsourced information, as well as technical procedures which could be applied once crowdsourced datasets are collected. Here, we plan to leverage the Amazon Turk platform as done in previous projects (1).