The problem

The goal was to develop a model to relate eye measurements (made by Lein’s confocal system) to blood glucose levels measured by a gold standard measurement device. There have been several suggestions in literature that blood glucose level and specific eye parameters may be correlated. Since Lein’s confocal system is capable of measuring these parameters with very high accuracy and precision, any relationship between blood glucose and eye parameters will become more pronounced.

Lein conducted a clinical trial at the Royal Berkshire Hospital (Reading, UK) to further investigate the relationship between blood glucose and eye parameters. The initial analysis of the results from the trial showed a very significant correlation between the blood glucose and the eye measurements. However, Lein wished to investigate the data further to improve the correlation and develop a model.

"Shayan’s signal processing expertise and insights provided a complementary approach to our data analysis work. His achievements have moved our work forward significantly and will support our future work in this area," said the industrial supervisor Graeme Clark, Lein Applied Diagnostics.

The approach

Various approaches were used to improve the statistical model relating blood glucose to measured eye parameters. Improving the data quality acquired by the confocal system was one of the main areas covered during the internship. Variability in the scans was reduced by approximately 30% using digital filtering and interpolation in conjunction with an outlier removal algorithm.

Furthermore, a significant improvement (35%) in reducing the variability between sets of data (i.e. averaged parameters produced by the system became more precise using post-processing techniques developed based on observations made during our experimental work at Lein) was achieved. Finally, a way to improve the statistical model relating blood glucose level to specific eye parameters using an appropriate coherent averaging method was discovered.

Through this work, Lein Applied Diagnostics also gained a deeper understanding of the most destructive sources of errors affecting their measurements which will help when conducting more trials in the future.

"The internship offers an insight into a different working environment and has allowed Shayan to develop generic skills that will be extremely useful during the remainder of his PhD and his subsequent career," said academic supervisor Martyn Hill, University of Southampton.

References

[1] Rother, K.I., Diabetes Treatment -- Bridging the Divide. N Engl J Med, 2007. 356(15): p. 1499-1501
[2] Pierro, L., R. Brancato, and E. Zaganelli, Correlation of corneal thickness with blood glucose control in diabetes mellitus. Acta Ophthalmologica, 1993. 71(2): p. 169-172