README

---

Introduction:

This is Python code for (re)producing the results in the paper
"A deep neural network for fast and accurate scatter estimation in quantitative SPECT/CT under challenging scatter conditions,"
Haowei Xiang, Hongki Lim, Jeffrey A. Fessler, Yuni K. Dewaraja, 2020
http://doi.org/10.1007/s00259-020-04840-9
Published in European Journal of Nuclear Medicine (EJNM)

The results show that using deep convolutional neural network (DCNN) can significantly accelerate the scatter estimation process of SPECT/CT while maintaining similar accuracy to previous Monte Carlo (MC) method.

If you use this code, please cite the EJNM paper linked above.

The code is distributed per the Creative Commons Attribution 4.0 license: https://creativecommons.org/licenses/by/4.0/

https://github.com/haoweix/spect-scatter-deep-learning

If you have any questions or need help using this code, please contact [email protected]

---

Contents:

• Example code for preprocessing, training and testing:
  • ipynb/Deep Learning for SPECT scatter estimation v9.ipynb: Instructions included in the python notebook.
• Pre-trained model used in our paper:
  • model/DCNN_SC_v9.hdf5.h5: A DCNN model trained with 8 simulated phantom/virtual patient. This is the model we used to generate most of the results in our paper.

Data:

• Data are separately stored to keep this repository lightweight:

  https://github.com/haoweix/spect-scatter-deep-learning-data

  • That repository stores phantom SPECT projection view data, including training data and testing data:
    • training_data: contains 8 sets of simulations that are evenly seperated into 2 sets (s1 means set No.1 and s2 means set No.2). Each simulation consists of scatter, primary, total projection views, and also attenuation map (mu-map).
    • test_data: contains 6 test data sets.
      • cat_cali_measure_0822.mat: measurement of 'calibration' phantom.
      • cat_liver_measure_w_noise_0826.mat: measurement of 'liver' phantom.
      • cat_liver_w_noise_Apr2019.mat: Monte Carlo simulation of 'liver' phantom.
      • cat_nema_w_noise_Apr2019.mat: Monte Carlo simulation of 'NEMA' phantom.
      • cat_shell_measure_Aug2019.mat: First measurement of 'shell' phantom (2 layers of activity).
      • cat_shell2_measure_08202019.mat: Second measurement of 'shell' phantom (3 layers of activity).
  • Per data sharing rules, real patient data is available at Deep Blue https://doi.org/10.7302/v07v-z854.

---

Copyright 2020-05-05, Haowei Xiang, Hongki Lim, Jeffrey A. Fessler, Yuni K. Dewaraja, University of Michigan