Iris Project and Beyond Hung Ton and Daniel Cao

Hung Ton

Who am I? 5th year optometry student Interests in Graphics Design, Engineering and a bit in

Music Also do photography as a hobby Been doing Research since 2014 as part of a school

supported Research has been focused on pigmentation and

colour, goes hand in h and with my other interests.

Daniel Cao

Daniel - Enjoys photography similarly to Hung and is well versed in software regarding photographic image manipulation

Good understanding of limitations and possibilities in digital image capture through love of photography

Hobby ties in with research project to incorporate related and novel image manipulation techniques and digital imaging with optometry. Por que lo nos dos?

Research Road Map

Iris project started off born out of frank experimentation

Wanted an indirect measure of the eyes pigmentation

Other forms of indirect measures were also

Now we want to take a step further and look towards using the same concept for different structure-function relationships

Iris Colour Poster

• Proof of concept • Demonstrating the application of

graphics analysis

Investigating Clinical

Applications

• Proposed today as part of my 5th year project

• Look deeper towards analyzing the iris

• Preliminary work towards other parts of the eye

The future…

• More expansive suite of different • Hopefully looking towards

publishing a methods paper• Ideal situation: working in

collaboration with companies or other pathologists.

Iris Colour recap

Iris Colour project was a way to compare the amount of pigmentation between individuals, indirectly.

Initially we used other techniques such as:

Measuring the colour melanocyte cell pellets

measuring the colour of melanocytes in bright field photos

These yielded pretty good results but we wanted a method that was quicker and perhaps can be done in vivo.

Abridged version of Iris project

1. We knew about Tyndall Effect, a phenomenon which causes a shift in perceived Hue depending on density of medium – in our case it was the iris stroma

2. We selected irises from a range of different subjective gradings and sorted them dark to light

3. Uses graphics technology to assess hue and showed there was a trend in dark to light was a shift from a shorter to longer wavelengths

What does graphics technology provide to Eye Care

However graphics technology goes beyond looking at the hue in the eye.

Potential clinic applications of these ideas include: Detection and Screening using processing to quickly

segment features Monitoring Lesions assess for growth, changing in

morphology To further understand how we can do this, we need to

look at some of the concepts in graphics technology In order to put this in perspective, here are some

examples I’ve created earlier. Analysis is not limited to only these processing steps,

there’s many more and they can of course be combined in sequential order.

Hue

Used in the Iris Colour Project. Useful in Ocular Pathology as many features are of a distinct hue from

the background. Here we have a fundus background, applied a median filter The naevus is then picked up from the background given its relatively

green hue compared to the background which is relatively red.

Saturation Here I imposed a

saturation mask over a red blank.

Because the feature is of a different saturation to the background it will show up as a decreased level.

Luminosity

Filters

Using filters to exclude out the features Then the exclusion criteria to make a selection Diabetic retinopathy Clinical focus: digital/automated monitoring of

diabetic retinopathy

Red Filter In Action

Different types of selections:Additive, Subtractive, Overlap and Invert Select

Much like criterias, we can combine selections (additive)

E.g. I want to isolate both haemorrhages and exudates we can exclude portions of selection

we’re catching too much noise from choroid vessels we need to remove adnexa/lids from iris photos

We can also chose an area of overlap I want to chose areas of haemorrhages but only within

a certain disc diameter measurement from the posterior pole

Lastly, we can invert selections Say I’ve selected vessels, I can invert my selection to

select only fundus background to assess retina/choroid.

Multiple Selections

Histograms Represents the distribution of pixels to a certain scale Can be luminosity (distribution of darker vs lighter

pixels) Can be hue, saturation as well. Great for quantification.

Histograms Luminosity

Histograms Continued

But probably for more subtle cases

“Fingerprints”

Fantastic for identification purposes Pupil size standardized by using flash to ensure

constriction. Note the yellow left peak in the right selection

Circularity

Takes segmented elements and analyses their morphology.

Can be good for iris nodules, hard drusen, lid papillae.

Where to now?

So that was an exhibition into what some graphics technology can provide

Want to determine a sensible scope for the 5th year research project