The potential for clinical application of automatic quantification of olfactory bulb volume in MRI scans using convolutional neural networks

Abstract

The olfactory bulbs (OBs) play a key role in olfactory processing; their volume is important for diagnosis, prognosis and treatment of patients with olfactory loss. Until now, measurements of OB volumes have been limited to quantification of manually segmented OBs, which is a cumbersome task and makes evaluation of OB volumes in large scale clinical studies infeasible. Hence, the aim of this study was to evaluate the potential of our previously developed automatic OB segmentation method for application in clinical practice and to relate the results to clinical outcome measures. To evaluate utilization potential of the automatic segmentation method, three data sets containing MR scans of patients with olfactory loss were included. Dataset 1 (N = 66) and 3 (N = 181) were collected at the Smell and Taste Center in Ede (NL) on a 3 T scanner; dataset 2 (N = 42) was collected at the Smell and Taste Clinic in Dresden (DE) on a 1.5 T scanner. To define the reference standard, manual annotation of the OBs was performed in Dataset 1 and 2. OBs were segmented with a method that employs two consecutive convolutional neural networks (CNNs) that the first localize the OBs in an MRI scan and subsequently segment them. In Dataset 1 and 2, the method accurately segmented the OBs, resulting in a Dice coefficient above 0.7 and average symmetrical surface distance below 0.3 mm. Volumes determined from manual and automatic segmentations showed a strong correlation (Dataset 1: r = 0.79, p < 0.001; Dataset 2: r = 0.72, p = 0.004). In addition, the method was able to recognize the absence of an OB. In Dataset 3, OB volumes computed from automatic segmentations obtained with our method were related to clinical outcome measures, i.e. duration and etiology of olfactory loss, and olfactory ability. We found that OB volume was significantly related to age of the patient, duration and etiology of olfactory loss, and olfactory ability (F(5, 172) = 11.348, p < 0.001, R² = 0.248). In conclusion, the results demonstrate that automatic segmentation of the OBs and subsequent computation of their volumes in MRI scans can be performed accurately and can be applied in clinical and research population studies. Automatic evaluation may lead to more insight in the role of OB volume in diagnosis, prognosis and treatment of olfactory loss.

Keywords: Convolutional neural networks; Deep learning; Olfactory bulb volume; Olfactory loss; Segmentation.

Fig. 1

Flowchart for Dataset 1 and 2, including input data and final data that was used as input for the method.

Fig. 2

Automatic segmentation of the olfactory bulbs in MRI scans using convolutional neural networks (CNNs). First, the center of each OB is localized. Subsequently, a region of interest (ROI) containing both olfactory bulbs is extracted and used as input for the segmentation of both olfactory bulbs to determine their volumes (figure derived from [Noothout et al., 2021], with permission).

Fig. 3

Automatic segmentation of the left (orange) and right (blue) olfactory bulb in two MRI scans (rows). The first column shows a coronal slice of the image, cropped for visualization purposes. The middle column shows the automatic segmentation result, obtained with the method while the last column shows the reference segmentation (manual segmentation). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 4

Automatic segmentation of the right (yellow) olfactory bulb. The left column shows the posterior coronal slice, where the cut-off needs to be made between the olfactory bulb and the olfactory nerve, cropped for visualization purposes. The middle column shows the automatic segmentation result, obtained with the method while the last column shows the reference segmentation (manual segmentation).

Fig. 5

Automatic segmentation of the left (orange) and right (blue) olfactory bulb in two MRI scans (rows). The first column shows an axial slice of the image, cropped for visualization purposes. The second and third column show the segmentation results, obtained with the segmentation CNN trained with only Dataset 1 and with additional training with Dataset 2, respectively, while the last column shows the reference segmentation. Dice coefficients improved from 0.64 to 0.75 (first row), and from 0.42 to 0.62 (second row) for the left olfactory bulb, while for the right olfactory bulb Dice coefficients improved from 0.74 to 0.83 (first row), and from 0.54 to 0.65 (second row). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. A1

Bland-Altman plot, showing the average difference in measurements between both measurements.