DL Flashcards
Hello, my name is Soorya Thavaraj and welcome to my presentation about my project titled ‘Tablet Technology in the Training of Visual Field Technique’.
Glaucoma is one of the most prevalent ophthalmological conditions in the world.
Approximately 57.5 million people suffer worldwide, just from one form of Glaucoma. It is the leading cause of irreversible blindness worldwide and is one of the most pressing healthcare concerns in the world
In glaucoma, an early and accurate diagnosis is key. It can often progress to the latter stages without any obvious signs or symptoms and many patients unknowingly have chronic glaucoma.
However, if it is diagnosed early it can drastically aid the prognosis of the patient. One study has shown that a simple treatment such as lowering the intraocular pressure by 30-50% can halt the evolution of glaucoma if done in the early stages.
However, it is important that any diagnosis made is accurate.
False-negative diagnoses can lead to irreversible optic damage, false-positive diagnoses can lead to subjecting the patient to harmful glaucoma medications.
Automated perimetry is one of the main investigations used to diagnose Glaucoma. It involves the patient placing their head on this headrest and being subjected to several white circular stimuli. They then press a button whenever they see a stimuli; which will in turn assess their visual field.
The results produced look like this. There are 4 main reliability indices: False Positives, False Negatives, Fixation Losses, Test Duration. There are also 2 key parameters used in diagnosing visual field defects, Mean Deviation and Pattern Standard Deviation
However, perimetry does have one key issue.
The learning effect is a well-documented phemenonon about automated perimetry. It is when patients who have never taken a visual field test before will show a distinct improvement in their results when they repeat the test, normally down to a lack of understanding the first time. For example let’s imagine a patient coming in for their first visual field test in the first circle. They might get quite poor results because they didn’t understand the mechanics or what to do. However they will go home with an increased understanding of how automated perimetry works and what is required of them. There is often then a 1-2 year wait between visual field tests. However, when they come back into the clinic, they will have a better understanding of visual field tests and likely produce more reliable results – leading to a more accurate diagnosis
Cycle repeats every time a new patient works in
We’re trying to accelerate this process with the use of an application called VisualFieldsEasy.
Even though the correct diagnosis is eventually achieved, this process can take 1-2 years in which time the eye could’ve deteriorated to an irreparable level
As a result, any way in which this process can be sped up would be very useful.
VisualFieldsEasy is an iPad application which attempts to mimic automated perimetry. Although previous studies have been conducted to check whether it can replace automated perimetry, no study has looked at potentially using it as a training tool.
Our project aimed to have a novel look at whether the app could speed up the learning process and lead to more reliable results at an earlier point
We hypothesised that it would cause a significant improvement.
Our research process had three phases, recruiting patients, visual field tests being done and the data analysis. Patient recruitment started with Exclusion and Inclusion Criteria being Assessed. If the patient was eligible they were given an information sheet detailing the whole research process and if they wanted to take part, informed written consent was obtained. They were then randomsied on excel and assigned to either the training group or the control group. Training Group underwent training via VisualFieldsEasy app
Control group underwent no training
All patients then had a regular visual field test
Software integrated into automated perimetry machine would then produce the data
This was analysed in Microsoft Excel using t-tests and Mann-Whitney U tests after being assessed for normality
The normality of the data was the first thing I analysed. The QQ plots of the 4 main outcomes, False Positives, False Negatives, Fixation Losses and Test Duration, are shown to highlight the lack of a normal distribution. These were all checked with Shapiro Wilks as well to ensure they weren’t normally distributed.
Mann-Whitney U tests were then done to compare the parameters between group A and B. Here are the four reliability results presented in violin plots. As you can see they look quite similar and when we conducted significance tests between the two group, no significance difference was found when comparing any of the parameters between the trained group and the control group.
However, when we compared the two eyes there was a significantly higher number of false positives in the left eye than the right in Group A which may help to explain our results and which I’ll talk about later
We also tested the Mean Deviation and Pattern Standard Deviation to see if there was any effect on those but there was no significant difference in either parameters.
One of the main reasons for our results was Perimetric fatigue. It is a well-documented phemomenon where the sensitivity of visual field testing has been shown to decrease with subsequent tests.
This might explain why in Group A the left eye which is tested second was shown to have more false positives than the right eye; interestingly this is only in Group A suggesting that the training has exaggerated this fatigue. Therefore the fatigue effect might have cancelled out the learning effect and hence no significant improvement was found in the training group.
2
The second reason for the results is the application VisualFieldsEasy. Several of the key components of automated perimetry weren’t tested with VFE, two key ones being keeping your head still and the manual dexterity of pressing a button. As a result, VFE might not be as perfect a training platform as thought and hence why the learning process wasn’t accelerated to the desired extent.
The third reason is the small sample size. Our n was only 30 and as a result it might not have been enough participants to exhibit a clear learning effect and it can still be influenced by random variation.
Finally, several confounders were not accounted for. Confounders such as the ones mention on the slide were not identified until after data analysis and as a result were not accounted for.
There were numerous strengths of the study but here are a few that I’ve selected.
Our study was a new concept with no previous studies to the best of my knowledge investigating the potential use of tablet technology as a training method for automated perimetry,
Our study also addresses one a key healthcare concern with optic problems, for example, the irreversible blindness that Glaucoma can cause, being a major worry for many patients.
Finally our methodology, was a strong point of our study. By choosing a patient-centered method where we initially approached patients whilst in the waiting room ensured, more time out of their day wasn’t wasted and elevated patient’s willingness to participate.
There were a few limitations with our study. Firstly we were unable to regulate the environment both the training and the visual field test was done in and especially how loud it was. Studies have shown that the level of noise can have a direct impact on the reliability of visual field tests.
Secondly the research was only conducted at 2 sites, both in central London and as a result, this may limit the ability of the data to be generalised for wider populations
Work-up bias was another limitation. It is where an examiner being aware of a research project can sub-consciously influence their behaviour and how they act. In this case, different technicians knew about the project to varying degrees and they might have subconsciously changed how much or little detail they give to the patient when instructing them
The last two are two I’ve mentioned before, our small sample size and not being able to account for confounders.
There are several approaches for the next phase of our study.
Firstly, one of the main issues around our project was the sample size which was largely due to the time constraint of the project. Simply, increasing the time period would allow more patients to be recruited.
A second approach would be to change a part of the study design. For example, letting the patients undergo the training at home and at their own ease and then coming in to the clinic to assess if that has helped. However, this is reliant on the patient and their self-regulation.
Finally, instead of an iPad, it may be more efficient to show patients a training video describing perimetry instead. A study investigating this was conducted in 2003, which showed some significance although with a limited sample size.
Here is a timeline of some key points in my project. Talk through slide
Here is the same slide but now with a graph showing my workload and stress levels throughout this period to help show how I was feeling throughout this project
So at the start, I was very interested in my project after the literature search I had been doing and excited about the potential findings I might make. Similarly when patient recruitment started, although my workload increased a lot with me going in every day to recruit patients, my stress level didn’t follow as I was enjoying this phase. However, it soon shot up and for the next month it alternated with my workload. This is because the place I was mainly based Western Eye Hospital is quite a small hospital and there wasn’t much space. With obvious priority being given to clinics and hospital staff, there was often limited time and space to recruit patients. Although all the staff were extremely friendly, this meant I struggled to recruit a lot of patients which was very frustrating, for example some days going in simply to be turned away. Then with data analysis, I initially struggled to retrieve my data as I didn’t have access to Medisoft, a ophthalmological software that stored visual field data. As a result there was a brief lul in my workload before a massive rush to get everything finished at the end.
Going into the project I had quite a few expectations about how I thought it would go. Some of them were proven right and some were proven wrong. I was really looking forward to how focussed on the patient this project would be and it’s something that came to fruition with me thoroughly enjoying the patient interaction aspect of my project.
At the start of my project, I naively thought there would be adequate space and time for me to recruit multit in order to recruit them but I quickly realized that this was far from the truth. I had limited time with my project directly speeding up or slowing the day’s clinic.
Finally, when my project was initially allocated to me I was skeptical about research in the sense as I was doubtful I would enjoy it. However, this project has definitely awoken an interest in research and especially ophthalmology that wasn’t previously there.
Overall, this project was a very positive and rewarding experience.
