City Research Online

Abstract

The purpose of this project was to investigate the utility of an automated means of quantifying the conjunctival bed. To simplify the task of any image processing system, it is important that the original image capture stage is optimised, this is an aspect that is often overlooked on the assumption that sophisticated computational techniques can be used to enhance the image. Hence, the concept of exposure density was used to achieve optimal images of bulbar conjunctival vessels using a digital camera (Kodak DCS 100 camera). A database of filters predicted a 6.5 times increase in vessel contrast when recording on the green channel of the charge coupled device (CCD) camera with a Wratten 99 (green) filter over the illumination optics, compared to recording without a filter.

With knowledge of the optical transfer characteristics of the imaging system a vascular model was derived. The tubular model, corrected for optical distortions, was fitted to densitometric profiles across conjunctival vessels repetitively imaged under different optical configurations. Although vessel contrast did not increase by the predicted amount, a -30% increase in the amplitude was observed in comparison to images recorded on the green CCD alone. Hence, this became the method of choice when imaging vessels of the conjunctiva.

Often automated methods of image segmentation are used without quantification of what is actually being measured, however, this is complicated when no accepted gold standard of measurement exists. Manual methods of determining widths using electronic callipers from projected digitally created photographs were used as the gold standard as they demonstrated a good intra session repeatability and range of measurement when 101 sample vessel widths were measured (95% confidence interval from +10.12 to -9.29pm, ranging from 14.4 to 140.0pm). To best agree with the automated measure of width, the algorithm was run at sigma (a) = 3, to give a 95% confidence interval of inter method repeatability of +9.41 to -8.48|im. However, it was acknowledged that this algorithm overestimates small vessel widths, and underestimates larger widths.

The use of an automated approach of vessel recognition results in a vast amount of data concerning vessel axis and vessel edge locations. Five indices were derived to describe the vascular bed including mean vessel width, width variance, tortuosity, tortuosity variance and density, for vessels as a whole and for sub-groups of vessels classified on the basis of size. These indices were a novel way of describing the conjunctival vascular complex. The inter session repeatabilities of these indices were investigated on 31 normal patients and were acceptable in all cases. Also the diurnal variation in these indices on this population showed negligible changes.

The angiopathic consequence of diabetes are well known in a variety of organs. However, these have never been adequately quantified in conjunctival vasculature. Seventeen Type I (TI) diabetics, 36 Type II (Til) diabetics, and 60 normals were analysed. Although several indices showed a difference between normals and diabetics for all vessels and sub-categories of vessels, by far the most remarkable was the dramatic change in density at a capillary level (vessels less that 25pm in diameter). A -57.12% (95% confidence interval from -71.96 to -36.76%, PcO.0001) reduction in capillary density was found in TI diabetics compared to normals and a reduction of - 17.5% (95% confidence interval from -41.02 to 16.64%) in Til diabetics compared to normals, however this was not statistically significant (P=0.273). A similar phenomenon was found in venular density (vessel 25 to less than 40|im in diameter). Hence, diabetes principally exerts its affect on the microvasculature of the conjunctival bed. In addition a statistically significant association between mean arterial pressure and vascular density was found, even though our sample did not contain anyone diagnosed with hypertension. A -13.82% (95% confidence interval from -23.71 to -2.65%, P=0.017) reduction in capillary density per lOmmHg rise in mean arterial pressure was established. Hence raised mean arterial pressure exerts an effect on smaller vessels of the conjunctiva.

Publication Type:	Thesis (Doctoral)
Subjects:	R Medicine > R Medicine (General) T Technology > TK Electrical engineering. Electronics Nuclear engineering
Departments:	School of Health & Psychological Sciences > Optometry & Visual Sciences School of Health & Psychological Sciences > School of Health & Psychological Sciences Doctoral Theses Doctoral Theses

Export

Downloads