Skip to Main Content
An official website of the United States government

Recurrent Neural Network for prediction of survival in lung cancer patients using longitudinal CT data

Principal Investigator

Name
Maria J. Ledesma Carbayo

Degrees
Ph.D.

Institution
Universidad Politécnica de Madrid

Position Title
Full Professor

Email
mledesma@die.upm.es

About this CDAS Project

Study
NLST (Learn more about this study)

Project ID
NLST-748

Initial CDAS Request Approval
Jan 8, 2021

Title
Recurrent Neural Network for prediction of survival in lung cancer patients using longitudinal CT data

Summary
Lung cancer is the most common cause of cancer related deaths worldwide. Despite recent efforts in lung cancer screening, 47% to 57% of new cases are diagnosed at an advanced stage. Radiomics is an emerging field that can generate quantitative biomarkers derived from clinical imaging. Outcome prediction in medical imaging are closely related to both spatial domain and course of time. However, there are very few previous works that propose deep-learning architectures to handle and exploit spatio-temporal imaging data.
Recurrent Neural Networks (RNN) and prominent RNN variation architectures, such as Long Short–Term Memory (LSTM), are specific neural network models which have shown great promise when dealing with longitudinal data.
Our goal is to design a novel spatio-temporal deep learning framework based on RNN trained with radiomics or deep features to predict survival in lung cancer patients using longitudinal CT images.

Aims

Aim 1: Develop a Recurrent Neural Network based on radiomics features to predict survival from longitudinal CT images in patients with malignant lung lesion (lung cancer patients)
Aim 2: Develop a Recurrent Neural Network based on deep features extracted from Convolutional Neural Network to predict survival from longitudinal CT images in patients with malignant lung lesion

Collaborators

Benito Farina (Universidad Politécnica de Madrid, Spain)
David Bermejo Peláez (Universidad Politécnica de Madrid, Spain)
Ana Delia Ramos Guerra (Universidad Politécnica de Madrid, Spain)

Related Publications