Screening lung cancer from CT scans through deep learning and assistive AI
Principal Investigator
Name
Amulya Mathur
Degrees
M.S.
Institution
Carnegie Mellon University
Position Title
Masters Student (Artificial Intelligence Engineering - Biomedical Engineering)
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-1225
Initial CDAS Request Approval
Apr 2, 2024
Title
Screening lung cancer from CT scans through deep learning and assistive AI
Summary
The project is about lung cancer detection through CT scans, and is based on the Nature magazine paper: D. Ardila et al., “End-to-end lung cancer screening with three-dimensional deep learning on low-dose chest computed tomography,” Nature Medicine, vol. 25, no. 6, pp. 954–961, May 2019. doi:10.1038/s41591-019-0447-x). It uses deep learning to fully screen lung CT-scans to detect lung nodules and predict risk of cancer. The proposed model performs the following steps: (1) Mask R CNN to create bounding boxes for lung segmentation; (2) TensorFlow Object Detection API for Cancer ROI detection; (3) 3D inflated Inception V1 for full volume CT scans implementation. According to the paper, the model achieves a state-of-the-art performance (94.4% area under the curve) on 6,716 National Lung Cancer Screening Trial cases, and performs similarly on an independent clinical validation set of 1,139 cases. This project was implemented using the NLST dataset consisting of 42,290 CT cases from 14,851 patients, 638 of whom developed biopsy-confirmed cancer within 1 year of a LDCT screening.
Aims
This project is part of my graduate course at Carnegie Mellon University: 16725 Methods in Biomedical Image Analysis. The overall aim is to learn the techniques of medical image segmentation and interpretation through deep learning. To implement the code from this online resource paper, I would need access to this dataset, as it was specifically designed to process the images and extract relevant features. By implementing this code, I would be able to learn the methodology used in medical imaging analysis, and implement new techniques for detection such as U-Net model for segmentation. Overall, the project aims to make us students prepared for medical imaging related careers in the future.
Collaborators
Carnegie Mellon University
Professor: John Galeotti