Analysis of nerve patterns for improved neuropathy diagnosis

The thesis is based on a project where neurologists from Kennedy Lab in the US wanted to know more about the nerve structure in the epidermis, the uppermost part of the skin. When a medical condition such as diabetic neuropathy progresses, epidermal nerve fibres are damaged and die. This translates to the symptoms of neuropathy such as loss of sensation and neuropathic pain. As symptoms can be treated if detected early, diagnosis at the earliest feasible stage is therefore important.

Clustering and mortality

One of the ways to diagnose the condition is to count nerves per unit area since a healthy person has more nerve trees than a diabetic. Not only do nerves die, but their structure also changes resulting in increased clustering. In the thesis, statistical tools are developed to understand more about nerve fibre changes and mortality. This is done through measuring the volume of the epidermis covered by nerves and by examining second-order properties of the underlying processes. A three-dimensional point process model for the nerve structure has been developed as well.

The second part of the thesis concerns methodological development. Konstantinos has collaborated with researchers from Finland and the Czech Republic and extended a certain non-parametric test, called the global envelope test. This is a statistical test suitable for hypotheses involving multivariate or functional data. Through developing permutation strategies, applications of these tests were extended to various areas of statistics. In particular, the test was extended for global inference in quantile regression and for comparisons of two or more distributions.

Academic journey

– I did my bachelor’s thesis in Cyprus, my native country. I had a brother in Sweden and knew the country’s education had a good reputation, as well as being free of costs for EU citizens. So, I did my master’s studies at Chalmers, in the programme of Engineering Mathematics.

It was a bit of a cultural chock at first, coming from a small village where everyone greets each other to a relatively large Swedish city. The weather also differs a lot, though Konstantinos do not like too much heat, and therefore acclimated well. The master’s thesis was made for Volvo Trucks where Konstantinos developed machine learning models to forecast truck arrival times operating on a transport mission. He wanted to continue as a doctoral student and applied for all positions available that year.

Thus, he landed in the quite different area of mathematical statistics but got lots of help from his supervisor. Since the position started just before the pandemic it was a bit strange at first working so much from home. Now, back to normal, Konstantinos likes to meet colleagues in the lunchroom again and to be able to go to conferences. He intends to look for a job in industry, preferably in Sweden, and include research.

Konstantinos Konstantinou will defend his PhD thesis Spatial Analyses of Nerve Patterns and Global Testing Approaches on May 24 at 9.00 in lecture hall Euler, Skeppsgränd 3. Supervisor is Aila Särkkä, assistant supervisors are Umberto Picchini and Ottmar Cronie.