Background

The olfactory bulb is one of the first regions of insult in Parkinson’s disease (PD), consistent with the early onset of olfactory dysfunction. Investigations of the olfactory bulb may, therefore, help early pre-motor diagnosis. We aimed to investigate olfactory bulb and its surrounding regions in PD-related olfactory dysfunction when specifically compared to other forms of non-parkinsonian olfactory dysfunction (NPOD) and healthy controls.

Methods

We carried out MRI-based olfactory bulb volume measurements from T2-weighted imaging in scans from 15 patients diagnosed with PD, 15 patients with either post-viral or sinonasal NPOD and 15 control participants. Further, we applied a deep learning model (convolutional neural network; CNN) to scans of the olfactory bulb and its surrounding area to classify PD-related scans from NPOD-related scans.

Results

Compared to controls, both PD and NPOD patients had smaller olfactory bulbs, when measured manually (both p < .001) whereas no difference was found between PD and NPOD patients. In contrast, when a CNN was used to differentiate between PD patients and NPOD patients, an accuracy of 88.3% was achieved. The cortical area above the olfactory bulb which stretches around and into the olfactory sulcus appears to be a region of interest in the differentiation between PD and NPOD patients.

Conclusion

Measures from and around the olfactory bulb in combination with the use of a deep learning model may help differentiate PD patients from patients with NPOD, which may be used to develop early diagnostic tools based on olfactory dysfunction.

中文翻译：

---

嗅球周围环境可帮助区分帕金森氏病和非帕金森氏嗅觉功能障碍

背景

嗅球是帕金森氏病（PD）的首批损伤区域之一，与嗅觉功能障碍的早期发作一致。因此，对嗅球的研究可能有助于早期运动前诊断。当与其他形式的非帕金森氏嗅觉功能障碍（NPOD）和健康对照进行专门比较时，我们旨在研究与PD相关的嗅觉功能障碍的嗅球及其周围区域。

方法

我们对15位诊断为PD的患者，15位病毒后或鼻窦NPOD患者和15位对照参与者进行了T2加权成像，基于MRI的嗅球体积测量结果。此外，我们将深度学习模型（卷积神经网络； CNN）应用于嗅球及其周围区域的扫描，以将NP相关扫描与NPOD相关扫描进行分类。

结果

与对照组相比，当手动测量时，PD和NPOD患者的嗅球都较小（均为p <.001），而PD和NPOD患者之间没有发现差异。相反，当使用CNN区分PD患者和NPOD患者时，准确率达到88.3％。嗅球上方的皮质区域围绕嗅沟延伸并进入嗅沟，这似乎是PD和NPOD患者区分的一个感兴趣区域。

结论

嗅球周围和周围的措施以及深度学习模型的使用可能有助于将PD患者与NPOD患者区分开，后者可用于开发基于嗅觉功能障碍的早期诊断工具。