Projects

Every project pairs students from the IT and medical faculties. Left of each dashed line is the scan. Right of it is what the model produces.

In progress

GAVE

Generalised Analysis of Vessels in Eye.

Distinguishing arteries from veins in the back of the eye and measuring each accurately — from standard colour photographs to dye-enhanced fluorescein angiography (FFA) — across artery/vein segmentation, cross-modal FFA-guided segmentation, and vascular biomarker quantification.

Why it matters

Accurate vessel measurements — widths, arteriovenous ratio, vessel density, and branching complexity — give doctors clinically meaningful biomarkers for diagnosing both eye and whole-body conditions.

Status
In progress
Field
Retinal Imaging
Year
2026
Methods
Segmentation, FFA, Retinal Imaging, Biomarkers
Team
IT Faculty and Medical Faculty

CHIMERA Agent

Combining histology, medical imaging and molecular data for prognosis and diagnosis.

A clinical prostate-cancer decision-making agent that analyses multiparametric MRI reports to estimate the probability of clinically significant cancer, recommends whether a biopsy is warranted, and provides structured reasoning for how the evidence was gathered and resolved.

Why it matters

Prostate cancer is the 4th most common cancer worldwide, yet current pathways drive overdiagnosis and invasive biopsies. Transparent AI decision support could reduce unnecessary, risky procedures.

Status
In progress
Field
Multimodal AI
Year
2026
Methods
LLM, MRI, Multimodal, Clinical Decision
Team
IT Faculty and Medical Faculty

TREAT-MMTB

Efficient AI technologies for multimodal management of tuberculosis.

Deep-learning models that read chest X-rays alongside clinical metadata to detect and segment cavitary lesions and predict the Timika severity score, generalising across patient populations in Korea, Mongolia, Peru, and the Philippines.

Why it matters

A shortage of trained radiologists makes TB screening slow and subjective. In the regions where TB claims the most lives, AI can widen access to fast, consistent diagnosis and treatment.

Status
In progress
Field
Radiology
Year
2026
Methods
Segmentation, Chest X-Ray, Multimodal, Radiology
Team
IT Faculty and Medical Faculty

Completed

COHORT-X

Extracting executable cohort definitions for medical imaging research.

NLP systems that turn messy free-text cohort selection criteria from biomedical literature into structured, computable representations — extracting inclusion and exclusion criteria as triples and resolving medical conditions to ICD-10-CM codes.

Why it matters

There is no benchmark structure for cohort selection criteria, making imaging studies hard to reproduce. Standardised, computable criteria improve reproducibility and fairness in future research.

Status
Completed
Field
Clinical NLP
Year
2026
Methods
NLP, Text Processing, ICD-10-CM, Reproducibility
Team
IT Faculty and Medical Faculty

Watch the video

Ischemic Stroke Detection

Detecting stroke in brain scans with machine learning.

A machine-learning model that identifies blocked blood vessels and damaged tissue in brain scans to detect whether an ischemic stroke has occurred, built on the ISLES22 dataset using HighResNet.

Why it matters

An ischemic stroke occurs when a vessel supplying blood to the brain is blocked. Fast, automated detection supports earlier intervention when every minute counts.

Status
Completed
Field
Neuroimaging
Year
2024
Methods
HighResNet, ISLES22, Brain MRI, Segmentation
Team
IT Faculty and Medical Faculty

Join AIM.

We're looking for students fluent in Python and curious about machine learning, neural networks, and the medicine they could transform. Medical and IT students work side by side.

Applications are open. Questions? Email [email protected] or read about each team.

Apply for the development teamApply for the operations team