3D Fusion W-Net and HoloLens-Based Interactive Visualization for Lung Airway Segmentation

Lung health is crucial to human well-being, with lung diseases exhibiting high incidence and mortality rates worldwide. Accurate segmentation of lung airways is vital for diagnosis, especially highlighted by the recent coronavirus pandemic. Despite extensive research, challenges persist due to the complex structure of lung airways. This study proposes a novel dual-encoder network combining Convolutional Neural Networks (CNNs) and Transformer networks for precise lung airway segmentation. Evaluations on a private dataset from Shandong University and the public LIDC-IDRI dataset demonstrate superior performance over existing methods. We also introduce a system utilizing Microsoft HoloLens 2 for 3D holographic visualization of lung airways, enhancing medical diagnostics and education. This user-centric pipeline offers immersive, interactive, and collaborative experiences for medical professionals. In summary, this study presents an advanced segmentation network and demonstrates the integration of Mixed Reality and deep learning in medical applications, potentially improving lung disease diagnosis and treatment.