Skip to main content

Sumeia Ahmed A Elkazza

Improving treatment for patients with retinal vein occlusion.

Email: s.a.a.elkazza2@ncl.ac.uk

Supervisors

  • School of Engineering
  • Faculty of Medical Sciences
    • Dr Jeffry Hogg

Project description

Retinal vein occlusion (RVO) has the second highest rate of all retinal vascular disorders. It can affect around 0.5% of adults over age 30 years. The types of RVO depend on the place where the RVO is prevalent. Through a range of different definitions, the two main categories of RVO are:

  • central retinal vein occlusion (CRVO)
  • branch retinal vein occlusion (BRVO)

RVO can result in severe visual acuity loss through macular oedema. It causes damage to the inner retinal cells carrying visual information from the macula. It also damages the macula photoreceptors themselves.

Anti-Vascular Endothelial Growth Factor (Anti-VEGF) is the standard of care in cases with RVO. But many studies show that not all visual gains are maintained beyond the first year. Additionally, patients show different behaviour patterns of response to anti-VEGF.

The Ophthalmologist has no way of knowing who will improve and who will not. This, in turn, can cause several issues. The treatment is very expensive. Also, giving an injection into the eyeball runs a 1:1000 risk of blindness. Thus, it would be better for patients and NHS funds to avoid administering unnecessary injections.

There are many studies on classification and detection of RVO, by processing digital images and highlighting the features. These systems already achieve efficiency and accuracy. But most previous studies have focused only on diagnosis which may not be that useful. The diagnosis is readily made by optometrists or ophthalmologists with traditional clinical methods.

In current clinical practice, there is no means of predicting which patients will respond well to treatment. Besides small-scale human led analyses, the literature does not offer a solution to this problem. We will build a recognition system and implement different algorithms, mainly for feature extraction. This will produce biomarkers of prognosis from OCT scans such as Ellipsoid Zone, Choroidal Thickness Changes, and greyscale.

Additionally, we are investigating a novel algorithm using deep learning techniques. The algorithm will provide phthalmologists with the information to make better decisions based on predicted long-term outcomes.

Interests

Reading, swimming.

Qualifications

  • MSc in Computer Science from Northumbria University