Segmentation of 3D CT scans of lung nodules.

Under Dr. Dhiraj Sangwan

As part of a group research internship, I worked on the segmentation of 3D CT scans of lung nodules to detect whether a mass is cancerous or benign. For this task, we used the LIDC-IDRI dataset. We initially expanded the dataset by rotating and flipping it, then used a Generative Adversarial Network to inject benign lung nodules in clean CT scans to boost the available data. Using various architectures, including Mask R-CNN, Faster R-CNN and Cascade R-CNN, I gained first-hand experience working with a powerful computer via remote access. This experience led me to enroll for a Data Science minor to explore the sheer power of machine predictions.