Segmentation of future lung cancer from LDCT screening scans.
Principal Investigator
Name
Daniel Truhn
Degrees
Priv.-Doz. Dr. med. Dipl.-Phys.
Institution
Department of Diagnostic and Interventional Radiology of the University Hospital Aachen (Germany)
Position Title
Research Group Leader
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-1207
Initial CDAS Request Approval
Mar 12, 2024
Title
Segmentation of future lung cancer from LDCT screening scans.
Summary
Our project will focus on the prediction and classification of future lung cancer in LDCTs.
First, we want to develop image registration to reliably find key points between different screening scans with years in between them.
This will make it easier for radiologists to find the same location on older CT scans and compare them.
For this purpose, different approaches from the literature will be evaluated and adapted for the registration of lung images over long time spans.
Second, we want to predict future lung cancer.
To do this, we will use our registration algorithm to map the location of nodules onto previous screening scans where they are not as clearly visible.
This mapping then allows us to create a holistic deep learning algorithm that takes a screening scan and predicts future lung cancers and finds very subtle signals for them.
First, we want to develop image registration to reliably find key points between different screening scans with years in between them.
This will make it easier for radiologists to find the same location on older CT scans and compare them.
For this purpose, different approaches from the literature will be evaluated and adapted for the registration of lung images over long time spans.
Second, we want to predict future lung cancer.
To do this, we will use our registration algorithm to map the location of nodules onto previous screening scans where they are not as clearly visible.
This mapping then allows us to create a holistic deep learning algorithm that takes a screening scan and predicts future lung cancers and finds very subtle signals for them.
Aims
-Long-term registration of the lung.
-Segmentation of signs of future lung cancer.
-Prediction of future lung cancer in screening CT scans.
Collaborators
Daniel Truhn, Diagnostic and Interventional Radiology of the University Hospital Aachen (Germany)
Patrick Wienholt, Diagnostic and Interventional Radiology of the University Hospital Aachen (Germany)