Application for accessing NLST pathology images to validate the proposed prognostic modeling method
Principal Investigator
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-795
Initial CDAS Request Approval
May 21, 2021
Title
Application for accessing NLST pathology images to validate the proposed prognostic modeling method
Summary
Our team is working on building models to predict the survival using pathology images. Using deep learning, a common method is to tile the image, followed by feature extraction neural network and aggregation network, finally outputting a risk score. The method is called Multiple Instance Learning which is categorized into weakly supervised learning. To tackle the problem of existing networks, such as weak neural networks used in MIL framework, our team intends to optimize the capability of networks to improve the prediction performance. As a result, the predictive models would be a personalized treatment recommendation tool which could be used in clinical routines. In short, NLST pathology image dataset is needed to validate our method performance. Related works would be written as papers to facilitate the development of this field. Thank you!
Aims
1. Validating our modeling method using NLST dataset of pathology slices.
2. Exploring the relation between pathology image and patient prognosis.
Collaborators
Fei Wu, UESTC
Related Publications
-
GraphLSurv: A scalable survival prediction network with adaptive and sparse structure learning for histopathological whole-slide images.
Liu P, Ji L, Ye F, Fu B
Comput Methods Programs Biomed. 2023 Feb 20; Volume 231: Pages 107433 PUBMED