Detection of Lung Cancer Powered by Machine Learning (2017 - 2019)

We developed a machine learning algorithm to identify cancerous lung nodules in CT scans with 97% accuracy. Based on annotations on each 3D scan made by radiologists, a complex image processing pipeline was created, where the lung is segmented from the rest of the scan and is divided into cubes so that the location of the nodule can be identified. Each section of the lung is used to train a neural network that identifies whether that section of the lung contains cancerous nodules.