VesselNet: A deep convolutional neural network with multi pathways for robust hepatic vessel segmentation

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer-Aided Diagnosis

Background:

• Vessel segmentation is crucial for surgical planning and diagnosis.
• Challenges include small vessel size, low signal-to-noise ratio (SNR), and varying contrast.
• Existing methods struggle with diverse medical image data.

Purpose of the Study:

• To develop an automatic and robust 3D liver vessel segmentation method.
• To improve recognition performance by exploring 3D structures.
• To create a method robust against varying contrast and device values.

Main Methods:

• A multi-pathways deep learning network for binary classification.
• Training on 3D planes (sagittal, coronal, transverse) for comprehensive structure exploration.
• Input transformation to a probability map to handle diverse medical image data.

Main Results:

• The proposed network achieves impressive performance compared to state-of-the-art methods.
• Demonstrated robustness across datasets with varying contrast and device values.
• Generated accurate vessel probability maps for precise segmentation.

Conclusions:

• The novel 3D deep learning network offers a significant advancement in liver vessel segmentation.
• The multi-pathways approach and probability map input enhance segmentation accuracy and robustness.
• This method holds promise for improved surgical planning and computer-aided diagnosis.