Full-view in vivo Skin and Blood Vessels Segmentation in Photoacoustic Imaging based on Deep Learning

Libraries Requirement

This project depends on the following libraries:

• Tensorflow 2.2.0
• Keras 2.4.3
• Opencv 4.5.1
• Numpy 1.20.1
• Matplotlib 3.3.4
• Itk 5.1.2

Training

• Run train.py
• You can change the model on the line 80th: model = . In this project I used three kind of model:
  • U-Net
  • SegNet-5 (VGG16 backbone)
  • FCN-8 (VGG16 backbone)

Predict B-scan Image

• Can predict all of B-scan in folder or each B-scan depends on the functions:
  • predict_all_Bscan(): predict all images in folder
  • predict_from_img(): predict desired image

Image Reconstruction

• In volumetric_help_function.py includes 3 functions:
  • img_2_npy(): Combine all B-scan image to 3D numpy file
  • npy_2_nrrd(): Convert numpy to NRRD format
  • cscan_reconstruct(): Reconstruct C-scan (Maximum amplitude image MAP) on 3D data

Result

Input image	Predicted Output	3D rendering by leveraging union of B-scan

Acknowledgments

• Any ideas on updating or misunderstanding, please send me an email: lycaoduong@gmail.com
• If you find this repo helpful, kindly give me a start!

Citation

  title={Full-view in vivo skin and blood vessels profile segmentation in photoacoustic imaging based on deep learning},
  author={Ly, Cao Duong and Vo, Tan Hung and Mondal, Sudip and Park, Sumin and Choi, Jaeyeop and Vu, Thi Thu Ha and Kim, Chang-Seok and Oh, Junghwan and others},
  journal={Photoacoustics},
  volume={25},
  pages={100310},
  year={2022},
  publisher={Elsevier}