Automatic Deep Learning Lung Segmentation Using U-Net

---

Project Lifespan: 2018

The main motivation of the this project was to explore the possibility of using a vanilla implemenation of the U-Net deep learning architecture, as a means of performing automatic lung segmentation of chest x-rays. The idea is that this will be a precusor and a pre-processing step to more advanced techniques of pulminary disease classification. The dataset used here was provided by the Radiological Society of North America. It consisted of over 25000 patients half of which were controls and the other half had pneumonia and/or other unknown symptom. A sample output can be seen below.

For a detailed breakdown of the implementation please refer to my Kaggle page or feel free to contact me for further information.