README

Code for the paper "Epidural anesthesia needle guidance by forward-view endoscopic optical coherence tomography and deep learning"[1]. The following pieces of python code and jupyter notebooks were used for the paper. The following architectures were used:

• Resnet 50
• InceptionV3
• Xception

Dataset

The dataset used can be found in [2]

Prerequisites

The language used is Python. We used Tensorflow 2.3.

Structure:

• Classification

  • 0a-Read_images.ipynb
    It process the images for tissues: fat, ligament, flavum ,and spinal cord from JPEG to numpy binary files

  • 0b-Read_new_class.ipynb
    It process the images for epidural space (empty) from JPEG to numpy binary files

  • 0c-Join_datasets.ipynb
    It concatenates the numpy binary files from 0a and 0b.

  • 1-Processing_results.ipynb
    Processing results from history to obtain number of epochs, validation, etc.

  • 2-Processing predictions.ipynb
    Process predicition probabilities to obtain ROC curve.

  • 3-Grad-CAM-S7.ipynb
    GradCAM apply to subject S7 using outer loop model

  • Inception_5cat/: Code for classification for 5 categories

    • inception_5cat_batch/
      • inceptionV3_arg_simult.batch: batch file for Summit
      • inceptionV3_arg_test.batch: batch file for Summit used for debugging
    • inception_5cat_python/
      • inceptionV3_arg.py

  • ResNet50_5cat/: Code for classification for 5 categories

    • resnet50_5cat_batch/
      • resnet50_arg_simult.batch: batch file for Summit
      • resnet50_arg_test.batch: batch file for Summit used for debugging
    • resnet50_5cat_python/
      • archResNet50_arg.py

  • Xception_5cat/: Code for classification for 5 categories

    • xception_5cat_batch/
      • xception_arg_simult.batch: batch file for Summit
      • xception_arg_test.batch: batch file for Summit used for debugging
    • xception_5cat_python/
      • archXception_arg.py

  • InceptionV3_binaries/: Code for binary classification

    • inceptionV3_bin1_batch/:
      • inceptionV3_arg_simult.batch: batch file for Summit
      • inceptionV3_arg_test.batch: batch file for Summit used for debugging
    • inceptionV3_bin1_python/
      • archInceptionV3_arg.py
    • inceptionV3_bin2_batch/
      • inceptionV3_arg_simult.batch: batch file for Summit
      • inceptionV3_arg_test.batch: batch file for Summit used for debugging
    • inceptionV3_bin2_python/
      • archInceptionV3_arg.py
    • inceptionV3_bin3b_batch/
      • inceptionV3_arg_simult.batch: batch file for Summit
      • inceptionV3_arg_test.batch: batch file for Summit used for debugging
    • inceptionV3_bin3b_python/
      • archInceptionV3_arg.py
    • inceptionV3_bin4_batch/
      • inceptionV3_arg_simult.batch: batch file for Summit
      • inceptionV3_arg_test.batch: batch file for Summit used for debugging
    • inceptionV3_bin4_python/
      • archInceptionV3_arg.py

  • ResNet50_binaries/: Code for binary classification

    • resnet50_bin1_batch/
      • resnet50_arg_simult.batch: batch file for Summit
      • resnet50_arg_test.batch: batch file for Summit used for debugging
    • resnet50_bin1_python/
      • archResNet50_arg.py
    • resnet50_bin2_batch/
      • resnet50_arg_simult.batch: batch file for Summit
      • resnet50_arg_test.batch: batch file for Summit used for debugging
    • resnet50_bin2_python/
      • archResNet50_arg.py
    • resnet50_bin3b_batch/
      • resnet50_arg_simult.batch: batch file for Summit
      • resnet50_arg_test.batch: batch file for Summit used for debugging
    • resnet50_bin3b_python/
      • archResNet50_arg.py
    • resnet50_bin4_batch/
      • resnet50_arg_simult.batch: batch file for Summit
      • resnet50_arg_test.batch: batch file for Summit used for debugging
    • resnet50_bin4_python/
      • archResNet50_arg.py

  • Xception_binaries/: Code for binary classification

    • xception_bin1_batch/
      • xception_arg_simult.batch: batch file for Summit
      • xception_arg_test.batch: batch file for Summit used for debugging
    • xception_bin1_python/
      • archXception_arg.py
    • xception_bin2_batch/
      • xception_arg_simult.batch: batch file for Summit
      • xception_arg_test.batch: batch file for Summit used for debugging
    • xception_bin2_python/
      • archXception_arg.py
    • xception_bin3b_batch/
      • xception_arg_simult.batch: batch file for Summit
      • xception_arg_test.batch: batch file for Summit used for debugging
    • xception_bin3b_python/
      • archXception_arg.py
    • xception_bin4_batch/
      • xception_arg_simult.batch: batch file for Summit
      • xception_arg_test.batch: batch file for Summit used for debugging
    • xception_bin4_python/
      • archXception_arg.py

  • ResNet50_binaries_test/: Code for binary classification for cross-testing

    • resnet50_bin1_test_batch/
      • resnet50_arg_test_simult.batch: batch file for Summit
      • resnet50_arg_test.batch: batch file for Summit used for debugging
    • resnet50_bin1_test_python/
      • archResNet50_arg_test.py
    • resnet50_bin2_test_batch/
      • resnet50_arg_test_simult.batch: batch file for Summit
      • resnet50_arg_test.batch: batch file for Summit used for debugging
    • resnet50_bin2_test_python/
      • archResNet50_arg_test.py
    • resnet50_bin3b_test_batch/
      • resnet50_arg_test_simult.batch: batch file for Summit
      • resnet50_arg_test.batch: batch file for Summit used for debugging
    • resnet50_bin3b_test_python/
      • archResNet50_arg_test.py
    • resnet50_bin4_test_batch/
      • resnet50_arg_test_simult.batch: batch file for Summit
      • resnet50_arg_test.batch: batch file for Summit used for debugging
    • resnet50_bin4_test_python/
      • archResNet50_arg_test.py

For Binary models:

bin1: Fat vs Ligament
bin2: Ligament vs Flavum
bin3: Flavum vs Epidural space (aka empty) 
bin4: Epidural space vs Spinal cord

For Python files, for cross-validation they should be run like this:

archResNet50_arg.py test_subject val_subject

e.g

archResNet50_arg.py 1 2

For cross-testing, they should be run like this:

archResNet50_arg.py test_subject n_epochs

e.g

archResNet50_arg.py 1 7

Video

As explained in future paper: "In order to simulate the use of the binary models in a realistic scenario, a stream of images was used as an input. The quantity of images of each tissue is proportional to the width of each tissue . The number of images was 100, 700, 100, 100, and 150 for fat, ligament, flavum, epidural space and spinal cord, respectively. To switch from one binary model to another, we used the simple criteria of having more than 35 images out of the last 50 for Class 1 (the deeper tissue type in these two classified tissues). Figure 8 showed some images from a video that can be found in the Github repository. These images are from Subject 7. Each image showed three important pieces of information. First, the proportion of the last 50 images that were predicted to belong to Class 1, e.g., Class 1 was ligament in the first Classifier and was flavum in the second Classifier. Initially, when the number of images was less than 50, the denominator shows the total number of images. Additionally, the color of fraction followed traffic lights colors. It changed from green to yellow at 26 and from yellow to red at 35. The second information was the current classifier. Last information was the truth and predicted label. The switch of binary classifier occurred when the number of images predicted as Class 1 reaches 35. The fraction did not appear anymore when the last classifier was reached. "

Fast speed

video_epidural_S7_v3_fast.mov

Slow speed

video_epidural_S7_v3_slow.mov

Paper

[1] Wang, C., Calle, P., Reynolds, J.C. et al. Epidural anesthesia needle guidance by forward-view endoscopic optical coherence tomography and deep learning. Sci Rep 12, 9057 (2022). https://doi.org/10.1038/s41598-022-12950-7

References

[2] Chen Wang, Qinggong Tang, Nu Bao Tran Ton, Paul Calle, Justin Reynolds, & Chongle Pan. (2021). Endoscopic OCT for epidural anesthesia [Data set]. Zenodo. http://doi.org/10.5281/zenodo.5018581

Contact

Paul Calle - [email protected]
Project link: https://github.com/thepanlab/OCT-Epidural