Clinically Applicable Deep Learning Pipeline for Diagnosis of Mediastinal Neoplasms
Principal Investigator
Name
Hengrui Liang
Degrees
Ph.D
Institution
The First Affiliated Hospital of Guangzhou Medical University
Position Title
None
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-868
Initial CDAS Request Approval
Jan 12, 2022
Title
Clinically Applicable Deep Learning Pipeline for Diagnosis of Mediastinal Neoplasms
Summary
Due to the unclear visual clues and complex anatomy, it is difficult to build an accurate lesion segmentation model for mediastinal neoplasms from CT images. Due to the inter-type of confusion, it is difficult to design a neoplasms type classification model. In this work, we will address the above difficulties by constructing a computer-aided mediastinal neoplasms diagnosis system, which would help radiologist to accurately detect and classify mediastinal mass. Based on NLST CT dataset, we are going tp test the tumor detection performance in a large-scale screening cohort of our model.
Aims
1, To build an accurate lesion segmentation model for mediastinal neoplasms from CT images
2, To design and develop a mediastinal neoplasms type classification deep learning model
3, To test and validate the performance in a large-scale screening cohort (NLST CT image cohort)
Collaborators
Jianxing He; The First Affiliated Hospital of Guangzhou Medical University
Qionghai Dai; Tsinghua University
Yuchen Guo; Tsinghua University
Ruijie Tang; Tsinghua University
Zeping Yan; The First Affiliated Hospital of Guangzhou Medical University