Automatically fitting CI's220px-Flag_of_the_Netherlands_svg.png

During my involvement in the Hearing Minds projects I worked in developing a decision support system for the fitting of cochlear implants. To do so, I first worked in modelling the relationship between parameter changes in the cochlear implant and the resulting changes in the measured outcomes. This is a complex relationships where many factors come into play, such as the last measured outcomes, the change in the intensity coding function, the patient's past performance, etc. Along with the expert clinician I built a probabilistic model and I worked in analysing the dataset derived from Eargroup's practice database to feed the model, its structural assumptions and its parameters. The model was validated against real cases from the practice database and the expert clinician's judgement.

After building the model, I worked in developing the algorithm that would find the configuration of parameters of the cochlear implant that would render the best outcome results for each patient, taking into consideration the last measured outcomes and the patient's parameter values. The fact that there are almost infinite combinations of values for the dozens of parameters in a cochlear implant forces us to come up with solutions that are smarter than the brute force search. I worked in developing a heuristic that made it possible to find a near optimal solution in a matter of seconds. During the last months of my research I worked in a iterative fashion making small changes both to the model and the search algorithm to make small incremental improvements to the decision support system.