Catheter Positioning tool

The following project leverages the applications of Semantic Segmentation in accurately identifying the nerve structure in Ultrasound Images and thereby providing a region of interest for the safe administration of the catheter.

Project Execution

• Clone the project from https://github.com/pranay-ar/Catheter-Positioning-Tool.git'
• Create a virtualenv by executing the following command: virtualenv -p python3 env.
• Activate the env virtual environment by executing the follwing command: source env/bin/activate.
• Enter the cloned repository directory and execute pip install -r requirements.txt.
• Download the data from the source and arrange the training and test data into train and test folders in the same directory.
• All the utility functions required in training and preprocessing the images can be found in utils.py.
• Hover over to the model.ipynb to understand the flow of training the U-Net Model.
• After training the model, you can visualise the results by running the test.ipynb.

Performance Analysis

• Task at hand:

• Training Loss:

• F1 Score:

References

• U-Net Architecture: U-Net
• Canny Edge Detection: Canny Edge
• Dataset: Kaggle

Authors

• Pranay Reddy
• Vipul Goel