Hybrid deep learning network for vascular segmentation in photoacoustic imaging

Area of Science:

• Biomedical imaging
• Medical image analysis
• Optical imaging

Background:

• Photoacoustic (PA) technology offers molecular specificity and high resolution for vessel imaging.
• Vessel imaging is crucial for medical diagnosis, but current segmentation methods face accuracy issues.
• Existing image processing techniques for vessel segmentation often result in under- or over-segmentation.

Purpose of the Study:

• To improve the accuracy and robustness of vessel segmentation in photoacoustic images.
• To evaluate the performance of a hybrid deep learning network for PA vessel segmentation.
• To address the limitations of current segmentation methods in medical imaging.

Main Methods:

• Implementation and comparison of a fully convolutional network (FCN) and U-net for PA vessel segmentation.
• Development and application of a novel hybrid network combining FCN and U-net architectures.
• Quantitative and qualitative assessment of segmentation results on PA vessel images.

Main Results:

• The hybrid network demonstrated superior performance compared to individual FCN and U-net models.
• Significantly increased segmentation accuracy and robustness were achieved using the hybrid network.
• The proposed hybrid approach effectively mitigates under- and over-segmentation issues.

Conclusions:

• A hybrid deep learning network offers a significant advancement for photoacoustic vessel segmentation.
• The developed hybrid network provides a more accurate and robust solution for medical image analysis.
• This approach holds promise for enhancing diagnostic capabilities through improved vessel imaging.