Currently my project has multiple parts, first I was working on a fibre bundle image viewer called Versi-colour where I learned the processes behind image enhancement for fibre bundle images, this also included learning the Qt framework in C++. Next, I moved onto machine learning in python to develop a light machine learning model that exploits fibre bundle rotation to generate synthetic multi-frame image bundles that could be resolved to give a 1.48 times improvement on structural similarity index measurement (SSIM) compared to linear interpolation, the current most used method for fibre bundle image enhancement. This improvement was recorded on a lab computer to be carried out in 0.03seconds per image. Currently I am now working on a fibre bundle training website that allows users to explore what tissue looks like through a fibre bundle using a synthetic fibre bundle mask and mouse tracking. The progress of the werbsite has reached the stage where I am looking for feedback based on the design and features of the website. The aim of this work is to help improve Clinicians understanding of what happens when they view tissue through a fibre bundle and how reconstructed image compare when using different enhancement techniques.
Updated the Versicolour system
Integrated sCMOS camera into test system and designed a new acquisition program
Testing Multi-frame autoencoder. machine learning architecture to reconstruct fibre bundle images
Constructing a web-based training application to help improve clinician understanding of fibre bundle images.
I’ve enjoyed meeting other students and seeing what they are doing. I’m exploring a collaboration with another PhD doing machine learning. For me the biggest highlight has been learning from the post-doc who worked on Proteus before me. He introduced me to the Versicolour system, coding practices and the physics behind the system.