Validation of pre-trained deep learning algorithms on NLST data.
Principal Investigator
Name
Ammar Jagirdar
Degrees
M.Tech
Institution
Qure.ai
Position Title
Product Manager
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-793
Initial CDAS Request Approval
May 7, 2021
Title
Validation of pre-trained deep learning algorithms on NLST data.
Summary
We developed deep learning algorithms to detect lung nodules and predict malignancy risk on X-rays and CTs of the Chest. Most deep learning algorithms tend to perform well when tested on data similar to the training set. The NLST data is large, unique and previously unseen by our algorithms. It serves as a valuable archive to test our Chest X-ray and Chest CT nodule detection and malignancy risk prediction algorithms against radiology opinions and confirmed diagnoses. Our project would aim to test the performance of pre-trained deep learning algorithms on NLST data.
Aims
- Testing for generalization of detecting nodules and other findings on the Chest X-ray.
- Testing for generalization of detection nodules and other findings on Chest CT.
- Test the ability to predict malignancy risk of a nodule on Chest X-ray and Chest CT.
- Evaluating nodule detection and malignancy risk prediction accuracy of pre-trained deep learning algorithms on Chest X-ray and Chest CT by measuring AUC, sensitivity and specificity, false positive rate, FROC-AUC, LROC-AUC
- Comparison of efficacy of deep learning algorithms for detecting nodules across Chest X-rays and Chest CTs.
Collaborators
Preetham Putha, Qure.ai