Seeing Change Over Time: Longitudinal Medical Image Analysis for Early Detection and Clinical Insight

About the Project

This research offers an exciting opportunity to explore how longitudinal medical images can be analysed to reveal subtle changes that unfold over time. Many clinical conditions progress gradually, with early signs detectable only through careful comparison of images acquired at different points in a patient’s care journey. Advances in machine learning, image registration, and temporal modelling now make it possible to study these changes with far greater precision. This PhD invites you to shape a research direction within this emerging field, contributing to tools and methods that could support earlier diagnosis, more personalised monitoring, and improved clinical decision making.

Medical imaging provides a rich window into the progression of diseases, whether through structural changes, functional adaptations, or the development of microscopic lesions that may predict future risk. By analysing sequences of images rather than isolated snapshots, researchers can uncover patterns that are otherwise invisible. Longitudinal analysis helps clinicians understand not just what has changed, but how and when it changed, which can offer crucial insights for prevention and treatment.

The project encourages applicants to define their own focus within the wider theme of longitudinal imaging. Examples of potential directions include:

Early Detection: Developing models that identify subtle, early-stage changes in tissue that might signal the onset of disease. For example, longitudinal fundus imaging offers opportunities to detect micro structural changes in the retina associated with conditions such as diabetic retinopathy. Techniques such as image registration, segmentation, and pixel-level comparison can support early lesion discovery in a manner that single time point analysis cannot achieve.

Disease Progression and Monitoring: Investigating how imaging biomarkers evolve over time and how these trajectories differ between individuals. Approaches may include temporal feature extraction, cross-timepoint alignment, multi-modal fusion, or transformer-based architectures that can learn temporal embeddings from medical image sequences.

Predictive Modelling and Risk Stratification: Exploring how longitudinal patterns can be used to predict future outcomes. Machine learning models that incorporate sequential imaging data can provide enhanced prognostic insights compared to static models, helping clinicians identify which patients may benefit from early intervention, closer monitoring, or tailored treatment strategies.

These examples illustrate the breadth of possibilities within the field. The project does not prescribe a specific medical condition, imaging modality, or algorithmic technique. Instead, you will be encouraged to develop your own research questions, informed by gaps in existing literature, access to datasets, and discussions with supervisors and clinical collaborators. Fundus imaging and image registration methodologies are available as potential starting points, but you are free to expand into other domains such as MRI, CT, OCT, ultrasound, digital pathology, or hybrid approaches.

Your research will include:

• Identifying a clinically meaningful area of longitudinal image analysis and formulating research questions that address unmet needs.
• Developing and implementing computational methods for aligning, comparing, and interpreting medical images captured over time.
• Exploring and evaluating machine learning methods suitable for longitudinal analysis, such as segmentation models, registration frameworks, or temporal architectures including transformer-based designs.
• Validating your approach using quantitative and qualitative measures, informed by feedback from clinicians, domain experts, and end-users.
• Presenting your findings in a way that supports real-world clinical relevance and future translational impact.

The successful applicant will join the Medical Image Quantification and Analysis AI Lab (MIQA-AI) at Kingston University. MIQA-AI is an interdisciplinary research group that brings together expertise in computer vision, medical imaging, statistics, and epidemiology. The team works on large scale image analysis, quantitative biomarker extraction, and AI models that link imaging features to clinical and population health outcomes. Ongoing research includes longitudinal retinal analysis, vascular quantification, and the development of automated tools that support early diagnosis and risk prediction in real-world healthcare settings. More information about the group and its projects is available at https://miqa.kingston.ac.uk/.

This project is interdisciplinary, bridging computer vision, machine learning, medical imaging, and health technologies. It is ideal for candidates motivated by developing computational methods that have the potential to advance early detection, support long-term monitoring, and generate actionable insights for clinical practice.

This research has the potential to contribute new knowledge on how temporal patterns in medical images can be modelled and interpreted. It may influence clinical workflows, support emerging diagnostic strategies, and pave the way for more personalised and predictive healthcare.

You will gain skills in image analysis, machine learning, temporal modelling, evaluation methods, and interdisciplinary research, preparing you for a career at the forefront of medical imaging and AI-driven healthcare innovation.

Applicants must have an Upper Second or First Class or Masters degree with a Computing, AI, Engineering or Mathematical background, or professional experience that has exposed them to logical thinking and problem-solving