Applying olfactory bulb volume in the clinic: Relating clinical outcome measures to olfactory bulb volume using convolutional neural networks

ABSTRACT

The olfactory bulb (OB) plays a key role in olfactory processing;its volume is important for diagnosis, prognosis and treatment of patients with olfactory loss, e.g. due to a Covid-19 infection, neurodegenerative diseases or other causes. So far, measurements of OB volume have been limited to quantification of manually segmented OBs, which makes its application in large scale clinical studies infeasible. The aim of this study was to evaluate the potential of our previously developed automatic OB segmentation method for clinical measurements of OB volume. The method employs convolutional neural networks that localize the OBs and subsequently automatically segment them (Noothout et al., 2021). In previous work, we showed that this method accurately segmented the OBs resulting in a Dice coefficient above 0.8 and average symmetrical surface distance below 0.24 mm. Volumes determined from manual and automatic segmentations were highly correlated (r=0.79, p<0.001) and the method was able to recognize the absence of an OB. Here, we included MRI scans of 181 patients with olfactory loss from the Dutch Smell and Taste Center. OB volumes were computed from automatic segmentations as described above. Using a multiple linear regression model, OB volumes were related to clinical outcome measures. Age, duration and etiology of olfactory loss, and olfactory ability significantly predicted OB volume (F(5, 172) = 11.348, p<0.001, R2 = .248). The results demonstrate that our previously described method for automatic segmentation and quantification of the OB can be applied in both research and clinical populations. Its use may lead to more insight in and application of the OB in diagnosis, prognosis and treatment of olfactory loss. We aim to extend our research to other populations of patients with olfactory loss.