 Hello everyone, my name is Emilie Karton, I work in Antwerp at the university and the university hospital as a PhD researcher and I'm very happy to present our recent work on diagnosing somatosensory tinnitus. So of course tinnitus is very often caused by some form of damage to the cochlea but obviously the presence and the severity of the tinnitus can be greatly influenced at many stages along the auditory pathway going from the brain stem all the way up to the primary auditory cortex. And an aspect that is somewhat underappreciated here is the influence that our somatosensory system can have on this auditory pathway. We note specifically that at the level of the brain stem there are cells in the cochlear nucleus that receive direct input from our somatosensory system and in fact in animal research these primary somatosensory projections have been proven. Translating these findings to tinnitus patients we see that many patients suffer from what is called somatosensory tinnitus meaning that they experience a strong influence from neck or jaw complaints on their tinnitus. Estimates vary but right now it's thought that this group contains somewhere between five up to 25 percent of all tinnitus patients. And of course for these patients it would be reasonable to think that if you treat these neck or jaw complaints that tinnitus severity would also decrease. And here I want to show you some results from my colleagues that were published last year where a group of patients with tinnitus and temporal mandibular complaints or jaw complaints were offered a form of orophacial treatment. And you clearly see that after treatments tinnitus complaints also decrease significantly and this just was an effect that remained stable over time also at a longer term follow-up time point. And in fact we saw that over half of these patients achieved a clinically significant benefit regarding their tinnitus complaints. So it is very much possible to treat these patients with somatosensory tinnitus with some kind of physiotherapy. The problem at this point is identifying those patients. So the diagnosis of these somatosensory tinnitus is very complex at the moment. And in fact a few years ago there was a large consensus meeting. There were several international experts present to review the evidence and to agree on a set of 16 diagnostic criteria for somatosensory tinnitus. These 16 criteria can be divided into three groups. First of all tinnitus modulations to the ability to modulate your tinnitus sound with certain movements. Second of all tinnitus characteristics for instance whether or not the tinnitus sound is very variable. And lastly accompanying symptoms which of course include neck or jaw pain but also tension in the muscles and even teeth clenching. So this consensus meeting was obviously a step in the right direction but still this diagnosis is very complex and often requires a trained specialist. So we asked ourselves whether it would be possible to simplify this diagnosis of somatosensory tinnitus. And to do this we launched an online survey on the website tinnitus help which you may know is a very popular platform for tinnitus patients online. The survey was launched in September 2019 as an open survey and patients gave their informed consent to use their anonymized data. We included questions on 12 out of those 16 diagnostic criteria I discussed earlier. Just for your information the remaining four criteria are based on certain manipulations or maneuvers so could not be included in an online survey. And then a whole list of questions on potentially influencing factors such as demographic variables but also properties of tinnitus itself or potential comorbidities. All in all our survey included 24 42 questions sorry in total. We had quite an enthusiastic response to this survey and in the end our data set contained data from almost 8,000 participants who filled out this survey completely. First of all we distinguished those participants with somatosensory tinnitus from those without it. We based our decision on whether or not there had been a diagnosis of somatosensory tinnitus made by their physician and whether the degree to which they themselves experienced any influence from neck or jaw problems on their tinnitus. I will quickly already show you the distribution we saw in this data set that about 10 percent of all participants received somatosensory tinnitus. Next we of course kept about approximately one fifth of the data apart to test our final model and on the remaining training data set we put from the five-fold cross validation to optimize the complexity and the parameters of the model. I already showed you the distribution of our data you can see that patients with somatosensory tinnitus are clearly in the minority here. To account for this imbalance in our data set we applied a majority weighted oversampling of our minority class so of the somatosensory tinnitus participants. First a quick note before I go into the results about model selection. We initially plan to use a quite complex model to analyze these data for instance something like a random forest classifier but we fairly quickly realized that we should really be prioritizing interpretability of the model and we even agreed that we would be willing to let's say sacrifice some percentage points in accuracy in favor of having a model that is very easy to interpret and easy to use in clinical practice and this is why ultimately we decided to use a decision tree for our specific research question. I will go over the parameters used in the decision tree in a moment but first I just want to say that we were able to achieve accuracy of 82% and a very well balanced sensitivity and specificity of course with this imbalanced data set this is really important to take into account. I'll just zoom in on our most important parameter here and this was the question of whether tinnitus and neck or jaw complaints increased or decreased simultaneously. Participants had different answering options if they answered that this was never the case you see that the probability of somatosensory tinnitus is still fairly low. Participants could also answer that they were not sure in that case we needed more answers to different questions further down the decision tree but from the moment that participants answered that they had this tinnitus and neck and jaw complaints increasing or decreasing simultaneously at least some days or most days or always we saw that the probability of them having somatosensory tinnitus really sharply increased so this was really our most important predictive factor. Next we also included the parameter of tension in the suboccipital muscles with of course the higher the degree of tension in these muscles the higher the chances of having somatosensory tinnitus. The ability to modulate tinnitus with certain movements and the presence or absence of teeth quenching. You can see that especially these two last parameters are really quite far down the decision tree so they will only be important for a certain subset of patients and in any case they're way less important than this first parameter that I discussed of these tinnitus and neck or jaw complaints increasing or decreasing simultaneously. So to conclude I think we succeeded in developing a model a fairly straightforward model with a reasonably high accuracy and maybe even more importantly a very well balanced sensitivity and specificity and this model is very easy to use and also very very quick to use it only requires four parameters in total. So I think that filling this out for a single individual patient would take maybe one minute max and it's very easy to do and you do not require a trained professional let's say to do it. So I think we succeeded in creating a very user friendly and intuitive tool to really aid this somatosensory tinnitus diagnosis in the daily practice of the clinic and for us as researchers this was also a really valuable lesson about adapting your model or tailoring the complexity of your model towards your specific research purposes. So with that I would like to thank you for your attention and I am of course very happy to take questions.