Exploring the generalization ability of chest XRay and CT pathology detection algorithms
Principal Investigator
Name
Bogdan Bercean
Degrees
M.S. Engineering
Institution
S.C. Mindfully Technologies S.R.L.
Position Title
Head of AI
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-1063
Initial CDAS Request Approval
May 30, 2023
Title
Exploring the generalization ability of chest XRay and CT pathology detection algorithms
Summary
The team will establish evaluation metrics to quantify the performance of the algorithms in terms of accuracy, sensitivity, specificity, and positive predictive value. Additionally, they will assess potential biases, limitations, and ethical considerations associated with the use of AI in medical imaging, particularly in the context of generalization and potential disparities in algorithm performance across different patient groups. NLST datasets will be combined with other publicly available datasets to increase the size and diversity of the validation set.
The project's outcomes will contribute to advancing the field of medical imaging AI by providing insights into the generalization ability of chest X-ray and CT pathology detection algorithms. The findings will inform researchers, clinicians, and developers about the limitations and challenges in deploying these algorithms on real-world clinical data. Ultimately, the research aims to enhance the reliability and effectiveness of AI-based pathology detection systems, ensuring safe and accurate diagnostic support for healthcare professionals.
The project's outcomes will contribute to advancing the field of medical imaging AI by providing insights into the generalization ability of chest X-ray and CT pathology detection algorithms. The findings will inform researchers, clinicians, and developers about the limitations and challenges in deploying these algorithms on real-world clinical data. Ultimately, the research aims to enhance the reliability and effectiveness of AI-based pathology detection systems, ensuring safe and accurate diagnostic support for healthcare professionals.
Aims
Measure the generalization ability of an XRay pathology detection algorithm on a mix of public and NLST data
Measure the generalization ability of an CT nodule detection algorithm on a mix of public and NLST data
NLST data will be used to pretrain a CT nodule detection algorithm to determine this technique's efficiency
Test various methods of improving the algorithm's generalization performance
Collaborators
Cristian Avramescu - S.C. Mindfully Technologies S.R.L.