Senior Research Fellow
Room 229 - 2nd floor - School of Engineering of Warwick, Coventry, CV4 7AL
Dr. Haleem joined University of Warwick as a Senior Research Fellow in August 2020. He has been based in Applied Biomedical Signal Processing and Intelligent eHealth Lab as his research contributions will be primarily towards GATEKEEPER Project. He would be mainly involved in design, development and supervision of novel deep learning based architectures for predicting health trajectory. He is a data scientist who has worked with multiple interdisciplinary research groups in various domains such as biomedical engineering, medical imaging, criminology etc. His research interests include (but not limited to) machine learning, big data analytics, time series analysis, text mining, image processing, computer vision etc.
Previous Research Projects
Big Data for Police Operational Analytics
Previously, he has been Senior Research Associate in Crime and Well-Being Big Data Centre, Manchester Metropolitan University, where he has been involved in the project of ‘Policing Operational Analytics’ in collaboration with Greater Manchester Police, which aims to analyse evidence-based police demand utilising their ‘Big Data’. Specifically, work in this project focus on three themes i) Developing the ‘Relational Data Management System’ which can integrate data acquired from different police resources (police reports, GPS data, structured police records) as well as socio-economic variables (demographics, landuse etc.) so as to understand causal factors of crime across the city; ii) Development of 'deep learning' based models which can automatically extract contextual insights from police reports which are written in unstructured format describing nature of the incident (e.g. mental-ill health, knife related crime) and to trigger necessary action from the police.; iii) Multilevel modelling so as to understand relationship between police patrolling with the demand generated and area covariates while utilizing the 'Big Data' resource describing police patrolling activity.
Automatic detection of retinal patterns for retinal disease diagnosis
Dr. Haleem finished his PhD in Computing with thesis entitled ‘Automatic detection of retinal features to assist retinal disease diagnosis’. Conventionally, retinal disease identification techniques are based on manual observations as clinicians perform basic image operations (changing contrast, zooming in, zooming out etc) manually for determining symptoms associated with retinal diseases. In this regard, his PhD research has been mainly focussed on automating steps in the diagnostic process that could help not only in determining any potential symptom more accurately but also has potential minimizing the time of diagnosis, enabling more patients to be screened and more consistent diagnoses can be achieved in a time efficient manner. Here he developed several novel computer aided methods in order to address the challenges associated such as i) Retinal Area Detector for detecting true retinal area in Scanning Laser Ophthalmoscope images obtained from Optos capable of having 200 field of view, ii) 'Adaptive Region-based Edge Smoothing Model' which determines retinal structure boundary based on supervised classification on multidimensional feature space as well as adjusts it according to energy minimization across irregular boundary points; iii) 'Regional Image Feature Model' performing retinal disease classification after fusing data comprised of morphometric and textural pattern at different spatial region in the retinal scan. This project has been funded by Engineering and Physical Science Research Council (EPSRC-UK) and Optos, plc (Nikon) as well as been selected as a finalist in UK ICT pioneer competition 2015 for the most exceptional thesis. A couple of publications have also been published under this project
A Task-Based Approach to Parametric Imaging with DCE-MRI
Dr. Haleem has also been involved in a project based on tumour localization and segmentation in the parametric images. The parametric images can be obtained from mathematical modelling of time series data of MRI (Dynamic Contrast Enhanced MRI). The time series data represents the pharmacokinetic activity of contrast agent as they pass through the vasculature system. He determined the weighted model of the time series data by emphasizing the first half and deemphasizing the second half which eventually improved the tumour segmentation performance while constructing the classification performance on parametric images.
For full list, please check google scholar (link)