Deep Learning-Based Segmentation of Post-Mortem Human's Olfactory Bulb Structures in X-ray Phase-Contrast Tomography

Area of Science:

• Neuroscience
• Medical Imaging
• Computational Biology

Background:

• The human olfactory bulb (OB) exhibits a laminar structure, but indistinct boundaries complicate cell population segregation.
• Standard 3D visualization tools lack the resolution for accurate morphometric analysis of OB layers.
• X-ray phase contrast tomography (XPCT) provides high resolution for brain imaging but presents challenges in manual delineation of OB microanatomy.

Purpose of the Study:

• To develop a fully automated segmentation pipeline for human OB morphological layers using XPCT data.
• To enable accurate morphometric analysis of OB layers for research into smell impairment.
• To provide a tool for assessing OB layer-specific degeneration in olfactory disorders.

Main Methods:

• Proposed a novel pipeline for tomographic data processing of human OB XPCT images.
• Utilized convolutional neural networks (CNNs) for segmenting native, unstained human OB images.
• Achieved virtual segmentation of the entire OB and precise delineation of individual neuronal cell layers.

Main Results:

• Successfully segmented XPCT images of the human olfactory bulb with high accuracy.
• Demonstrated the capability of the CNN-based pipeline for automated layer delineation.
• Enabled accurate virtual segmentation of OB structures, crucial for further analysis.

Conclusions:

• The developed pipeline offers an effective tool for automated segmentation of human OB layers from XPCT data.
• This approach facilitates accurate morphometric analysis essential for understanding OB changes in smell impairment.
• The tool aids in investigating OB layer-specific degeneration in patients with olfactory disorders.