neural-segmentation

Identification of nerve structures in ultrasound images of the neck, via Convolutional Neural Networks and Unsupervised Learning Techniques

About the Challenge

Ultrasound Nerve Segmentation - Accurately identifying nerve structures in ultrasound images is a critical step in effectively inserting a patient’s pain management catheter. Surgery inevitably brings discomfort, and oftentimes involves significant post-surgical pain. Currently, patient pain is frequently managed through the use of narcotics that bring a bevy of unwanted side effects.

Dataset description

Each dataset test/train case is described by an grayscale ultrasound image of dimensions 580px x 420px.

1. Train Set The train dataset contains 5635 cases, each case is associated to a binary B&W mask of size 580px x 420px, which establishes the position of the nerve structure in the image.

2. Test Set The test set is comprised by 5508 images, without any binary masks defined.

Dataset Download

• Train Dataset (1.1GB) - MD5 Checksum: fba272da39a6bcfe7489ae6a802924bf
• Test Dataset (1.1GB) - MD5 Checksum: 1a6cb8739900b7df85f368729141274f

Repository clone and dataset download script

Open a terminal in the folder that contains this repository clone and execute:

git clone https://github.com/ai-society/neural-segmentation.git
cd neural-segmentation
./init.sh